Dear OnStrategy Reader,

Here is what you will find in this issue:

1. GoPro is dead, and they know it

2. Case study: Amazon - AI usage is not AI value

3. Robinhood’s agentic can trade 24/7…for your

4. Europe and (bad) demographics

5. on Ferrari Luce 🚗

6. Search becomes the answer layer

On to the update:

GoPro is dead, and they know it

GoPro is another reminder that inventing a category is not the same thing as controlling it. The company gave the world the action camera, built the brand, owned the lifestyle, and turned “GoPro” into a verb. But the center of gravity in consumer hardware has moved from brand storytelling to supply-chain velocity. DJI and Insta360 (both Chinese) are both faster product organizations sitting closer to the manufacturing ecosystem that now defines the category. After years of funding and innovation, they now have bigger sensors, better stabilization, stronger thermal management, more features, faster iteration. They are the compounding returns of being embedded in the hardware stack. Nikkei reports that the company is now exploring a potential sale after years of declining revenue, while DJI and Insta360 gain ground with higher-performing products.

The implication is uncomfortable for the West. Hardware is not won by “having a great brand” plus outsourcing the hard parts. It is won by dense supplier networks, manufacturing feedback loops, component access, engineering talent, and the ability to launch, learn, and relaunch faster than the incumbent can hold a board meeting. China’s advantage here is system speed. A company like DJI can learn from drones, cameras, sensors, batteries, gimbals, thermal systems, software, and creators, and only then redeploy that learning across adjacent categories.

This is where government subsidies matter, but not in the simplistic “give money to a national champion” sense. The United States and Europe often subsidize factories after the market has already moved. China subsidized entire ecosystems for suppliers, industrial parks, export capacity, engineering density, battery chains, electronics assembly, and patient iteration. That matters because consumer hardware is a supply-chain sport. A subsidy that produces one plant is useful, but a subsidy that produces a cluster is strategic. The lesson from GoPro is that industrial policy should not ask, “Can we save this brand?” It should ask, “Can we rebuild the capability base that lets the next GoPro scale before the next DJI arrives?”

GoPro’s possible sale is an industrial policy story. America still creates categories (eg. action cameras, smartphones, social media, AI models, creator platforms), but China increasingly industrializes them at speed. The question for Western governments is whether they want to subsidize demand, protect incumbents, or rebuild supply. Only the third option changes the outcome. Because in hardware, the moat is the chain of companies behind it. LINK

AI usage is not AI value

Amazon has discovered a basic law of corporate metrics, which is that when you give employees a leaderboard, they will optimize for the leaderboard. This is charming when the metric is sales, slightly weird when the metric is meetings, and expensive when the metric is AI tokens consumed. The company reportedly shut down an internal "Kirorank" leaderboard after employees used AI agents to generate extra activity and climb the rankings, which is a very modern situation where the humans instructed the robots to create useless work so the humans could appear more committed to the robot future 🫠. This is adoption theatre with a GPU bill attached.

The real question is output. Using AI 5 times, 500 times, or 5 million tokens worth of times matters only if something valuable comes out, like better code, faster deployments, fewer bugs, improved products, lower costs, happier customers. Maybe the spending is justified. Amazon expects enormous AI capex, and a company at that scale can rationally burn compute if it produces better internal tools, stronger developer velocity, or defensible product improvements. But the unit of analysis has to move from "how much AI did you use?" to "What did you build, what changed in the business, and what cash flow does it touch?". Which is why Amazon’s shift toward measuring "normalised deployments" and useful code makes sense.

In capitalism, tokens are an input, but products are the output. Everything else is a very expensive scoreboard. FT

Robinhood’s agentic can trade 24/7…for your

Robinhood’s agentic trading story is about the next abstraction layer in retail speculation. Agentic trading turns a sentence into a strategy with commands like “rebalance my portfolio”, “build me an AI basket”, “buy oversold stocks”, “monitor semiconductors”, and “later trade options, crypto, futures, and event contracts”. The product becomes execution itself where the user intent goes in, trades come out.

The economic model is straightforward. Zero-commission trading makes high-frequency small-account strategies feel free to the user, while the market structure still monetizes activity through spreads, routing, and market-making economics. For Robinhood, agentic trading expands the surface area of order flow. More automation means more engagement, more transactions, more data, and more opportunities to convert attention into financial activity.

The strategic risk is convergence. Personalized AI strategies may turn into the same 20 retail instincts running at machine speed, which are focused on buying the dip, chasing momentum, rotating into AI, selling volatility, gambling on options, and so on. That creates predictable flows, and predictable flows create alpha for whoever trades against them. Robinhood is becoming a consumer financial operating system, where investing, credit cards, sneakers, restaurants, and prediction markets all become conditional spending, summarized by “when X happens, allocate my money”. LINK

Europe and (bad) demographics

🇪🇺 Europe’s median-age map is a useful reminder that demography is the operating system. A country with a median age of 49 is not simply "older" than one with a median age of 34. It has a different discount rate. Older societies vote differently, consume differently, invest differently, and regulate differently. They place a higher premium on stability, predictability, healthcare access, asset protection, and low volatility. That makes sense individually, but it compounds into a macro problem where the institutions most capable of protecting accumulated wealth are not always the institutions most capable of creating new wealth.

This is why Europe’s demographic challenge is less about population decline in the abstract and more about capital allocation. Aging societies tend to shift resources from formation to maintenance, from housing construction to housing preservation, from education to pensions, from risk capital to safe assets, from migration to identity politics, from productivity experiments to compliance systems. That is fine if the asset base is exceptional, and Europe’s is the home of many brands, infrastructure, industrial depth, universities, tourism, dense cities, and social trust. But it means Europe increasingly looks like a high-quality legacy business with declining organic volume growth.

The optimistic interpretation is that Europe can become a premium, automation-heavy, scarcity-based economy. Fewer workers can still support more output if capital deepening, robotics, AI, energy productivity, and high-value services scale fast enough. In that model, 🇮🇹 Italy is not "too old", but a portfolio of luxury, tourism, design, precision manufacturing, food, and patrimony monetized globally. 🇩🇪 Germany is not just demographically challenged, but an industrial balance sheet that must swap cheap energy and export volume for automation and software leverage. 🇷🇴 Romania, at 44.1, is not young, but it still has some demographic room relative to Southern Europe , if it can convert labor, education, and migration flows into productivity before aging catches up.

The pessimistic interpretation is that Europe becomes a continent of beautiful fixed assets and rising transfer claims. The danger is that the political economy becomes optimized for preserving yesterday’s promises with tomorrow’s shrinking workforce. That produces the familiar European equilibrium with high savings, low risk appetite, expensive housing, slow permitting, underbuilt infrastructure, cautious incumbents, and young people asked to be entrepreneurial while paying for a system designed before they were born.

The real question is not “Is Europe old?" because the map answers that. The question is whether Europe can raise productivity faster than it raises dependency ratios. It needs a model where fewer people can produce much more, otherwise the continent becomes a very elegant annuity with a growth problem, and zero growth is a "zero-sum" game. LINK

on Ferrari Luce 🚗

Everyone can make a fast EV now because acceleration has been commoditized by batteries in the same way that high-frequency trading commoditized being quick.

Ferrari’s problem is whether it can make silence feel expensive. Combustion engines gave Ferrari a free emotional dividend with noise, vibration, danger, and theater. Electric motors give you torque and efficiency, which are nice, but also vaguely appliance-like. That's why Ferrari has spent 5 years and 40,000 kilometers developing an acoustic character that captures and amplifies the hum of the electric motors, because apparently, the future of Italian passion is a sensor on the rear axle.

Still, this may work, because Ferrari’s real moat was never just the engine, but the allocation, ritual, waiting lists, resale mythology, and the ability to convince rich people that buying the impractical thing is rational because too few other people are allowed to buy it. Therefore, the Luce is a scarcity test with batteries, and not an EV test. Bloomberg

Search becomes the answer layer

🔎 In my latest Where is my MOAT? investor brief, I ask a simple question: "Is Google still defending search, or is search quietly becoming something else?"

The old search business monetized intent, where you typed a query adn Google sold the click. The new search business monetizes the answer, the follow-up, and eventually the action. That is a very different game.

Classical search is still enormous. Google Search & Other generated $224.5B in FY2025, while the global search advertising market is around $268B in 2026. But the surface is changing fast because ChatGPT runs 2.5B prompts per day, AI Overviews now appear in a meaningful share of Google searches, and answer engines reduce blue-link clicks while increasing the value of the session.

That is why the next 5 years are not about "Google dies" or "ChatGPT wins". They are about who owns the answer layer on top of the query layer.

🤔 So, who captures the next search market: the company that owns the query, or the company that owns the answer? WIMM

• 27 May: Apple, the Band, and the 2016 Decision that compounds

• 28 May: Has Eli Lilly solved LDL cholesterol?

• 29 May: Search + Generative Search: TAM, shift, and who captures It

• 30 May: Ferrari’s Electric Test

• 31 May: [Deep Dive] IBM

Consider subscribing to get that👇🏻 (one month free)

One month FREE trial

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

3. Endi Ungureanu - Tura de duminica

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. Has Apple missed “the band” form?

2. Anyone serious about hardware should make its software

3. on Google‘s search agents will increase its TAM

Onto the update:

Has Apple missed “the band” form?

Apple did not miss health wearables in the obvious way. The Apple Watch is still the category-defining device, that is, a computer on the wrist, a notifications terminal, a fitness tracker, an emergency device, a payments device, and increasingly a medical-adjacent platform. That is exactly the problem. Mark Gurman, in a Bloomberg piece, frames the Apple Watch as “over a decade old” and in need of a shake-up because the market has pivoted toward more focused health and fitness wearables like Whoop, Oura, Fitbit-style bands, and screenless devices. Apple built the best wearable computer. Oura and Whoop built health instruments. Those are not the same product.

The strategic miss is not in hardware but in use-case purity. Whoop says: “I am for recovery, strain, sleep, and performance”. Oura says “I am for sleep, readiness, temperature, and long-term body signals”. Apple Watch says “I can do everything, including telling you that your Uber is arriving while you are trying to sleep”. In classic Apple fashion, integration made the device powerful. In this case, integration also made it noisy. Health is not just another app category. Health is intimate, passive, longitudinal, and trust-based. A ring or band can disappear into the background. A watch keeps asking to be looked at.

This matters because the next phase of wearables is less about “tracking workouts” and more about owning the personal health graph. Apple has the installed base, the chips, the sensors, the Health app, the privacy brand, the iPhone distribution, and the medical partnerships. But Oura and Whoop have something Apple increasingly struggles to create inside mature product lines, which is, a sharper job to be done. They sell interpretation, not just measurement. They sell a daily answer like “should I push, rest, sleep, recover, change behavior?”. Apple often gives you data to share it with your doctor or just to sit idle. The challengers give you a narrative, and in consumer health, the narrative is the product.

Yes, Apple has partially missed the band/ring moment, not because it cannot build one, but because the Watch’s success made the obvious next product politically and strategically awkward. A screenless Apple health band would implicitly admit that the best health wearable is not always a tiny iPhone on your wrist. An Apple Ring would compete with the Watch at night, in workouts, and for passive monitoring. But that is exactly why Apple should do it. The danger is not that Whoop or Oura become bigger than Apple Watch. The danger is that they define the health-native interface while Apple remains trapped defending the smartwatch. The iPhone taught Apple to aggregate attention. Health wearables require the opposite, ie. earning the right to disappear. Bloomberg

Anyone serious about hardware should make its software

Tesla’s Vision-based airbag feature is the kind of thing that looks small until you remember the entire Tesla thesis: the car is not a product; it is an endpoint. Legacy automakers sell you a machine that gets old. Tesla sells you a machine that can, in theory, gets better. Deploying airbags 70 milliseconds earlier is a reminder that Tesla’s real advantage is not batteries or factories, but the ability to turn fleet data into software updates and software updates into product differentiation.

The implication is uncomfortable for everyone else. Safety used to be hardware, regulation, and model-year engineering. Now it is becoming data, inference, and over-the-air distribution. That gives Tesla a path to compound advantages across the installed base, but also raises the stakes.

When software touches safety-critical systems, execution errors are liabilities. This is why Tesla is still so hard to value. It is either an automaker with a dangerous multiple, or a real-world AI platform hiding inside a car company. The market keeps paying for the second story, but the risks keep reminding you that the first one has not disappeared. LINK

on Google's search agents will increase its TAM

Google’s old business was, roughly, that you typed a thing you wanted into a box, Google showed you ten blue links and three ads, and everyone involved pretended this was about "organizing the world’s information" rather than "auctioning your intent to the highest bidder". It was a very good business.

Now Google is trying to make that business slightly weirder and potentially much better. At I/O, it announced search agents. These agents are basically Google Alerts, but smarter, more contextual, and living inside the search experience itself. Instead of you searching once for "best family hotel in Lisbon in October", the agent can keep looking, keep comparing, keep waiting, and eventually say: "Good news, the thing you wanted now exists, is cheaper, or is being aggressively marketed to you by someone with a conversion budget".

This is useful for users, probably. But it is also extremely useful for Google, definitely. Search used to monetize intent at the moment of expression. You searched, Google inferred what you wanted, advertisers paid. AI search agents let Google monetize intent over time. Your desire is no longer a query, but more likean annuity.

And the really interesting part is not just advertising. It is demand discovery. If enough people ask Google for something that does not quite exist yet (examples: a cheaper EV, a better CRM for dentists, a hotel with childcare and no children somehow), Google learns that before the market does. Eventually, it can tell companies, explicitly or implicitly: "People want this. You should make it. Also, here is where to advertise it when you do".

The future of search extends now beyond finding what exists. It is building a futures market for unmet demand. Google used to sell access to information. Now it may sell access to desire before the product exists. Which, as businesses go, is pretty good. LINK

This week on “Where is my MOAT?”

• 21 May: [IPOs] Cerebras, Anthropic, and the inference bottleneck

• 22 May: 🚀 SpaceX (SPCX): A strategic analysis of the S-1

• 23 May: [Analysis] on the 🇷🇴 Romanian economy

• 24 May: [Deep Dive] LVMH sells patience

Consider subscribing to get that👇🏻 (one month free)

One month FREE trial

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

3. Endi Ungureanu - Tura de duminica

Here is what you will find in this issue:

1. Yes, it’s the phones!

2. The status of AI job automations in 2025

3. The two Europes

4. On the fertility front …in China

5. Meta’s Interface, Palantir Technologies’s Integration

Onto the update:

Yes, it’s the phones!

The phone is not really a phone, in the same way that a casino is not really a building with carpets. It is an attention market, a dating market, a status market, a pornography market, a beauty-standard market, a politics market, and a 24/7 comparison machine that you hand to a 12-year-old and then act surprised when, at 23, they think ordinary life is a bad deal. The polite version of the fertility story is an economic one, of course, where it shows that houses are expensive, childcare is expensive, wages are uncertain. True, but not sufficient. The sharper version is that the phone has changed the demand side of adulthood itself.

There is no smoking gun, but there is a lot of smoke. Births fell first and fastest in places that got high-speed mobile internet first. Different economies, different religions, different cultures, and different exposure to the financial crisis have the same pattern after smartphone adoption. That is not proof, but it is the kind of coincidence that eventually stops looking like a coincidence.

The mechanism is not mysterious. Fewer people meet in real life, fewer people form couples, fewer couples have children. Young people are not flirting, failing, adjusting expectations, building boring intimacy and learning how real partners actually work. They are scrolling. And scrolling is not neutral time. It is time plus ideology. It tells young women that men are disappointing, young men that women are impossible, and everyone that ordinary partners are underpriced garbage. The result is a relationship market with terrible liquidity where everyone is browsing, nobody is settling, and the bid-ask spread is absurd.

So when people say “just limit phones to one hour”, that sounds reasonable, like saying “just let the casino operate between 5 and 6 p.m”. The cost is not only the 4 or 5 hours lost every day. The cost is that the phone trains preferences in relation to what beauty means, what a partner should be, what freedom means, what compromise costs, what a normal life should look like.

Maybe the unpopular answer is the actual answer: no smartphone for children until adulthood, or as close to that as society can enforce. Not because it is elegant, liberal, or easy. Because the device is not just stealing childhood attention. It is rewriting adulthood demand. FT

The status of AI job automations in 2025

This chart from Gartner is trying very hard to calm everyone down by saying "most layoffs were not caused by AI", which is technically true in the same way that saying "the asteroid did not directly kill the dinosaurs; the climate change afterward did".

The interesting number is not the tiny percentage of direct AI-productivity layoffs, but the “reposition” category (ie. companies restructuring themselves around AI opportunities). Because AI is not really a tool like Excel or Slack (ie. tool). It increasingly does actual work. That means firms are reorganizing around a new production function where software is no longer assisting labor but beginning to substitute for pieces of it. The market hears “AI layoffs” and imagines robots firing accountants, but the reality is more subtle. Management freezes hiring, redesigns workflows, shifts budgets, and quietly discovers that five people plus AI can suddenly do the work of eight.

Hence, I think the more important hypothesis is not replacement, but expansion (❗️). Historically, technology that lowers the cost of production tends to expand the market before it destroys the labor category. The smart company will not use AI merely to cut 10% of headcount and report one nice quarter of margin improvement. The smart company will ask something like "if the same workforce suddenly becomes 30% more productive, what additional services, customer support, product development, personalization, sales outreach, or operational responsiveness can we now offer?"

AI makes labor more leveraged. The foolish manager sees AI and thinks “fewer employees". The ambitious manager sees AI and thinks "same employees, dramatically larger company".

That is why I suspect the long-term winners will not be the firms that replace people fastest, but the firms that combine people and AI into a much more scalable organization.

The two Europes

Europe looks like 2 companies sharing one ticker: Eastern Europe is the growth subsidiary, while Western and Southern Europe is the legacy conglomerate explaining that, yes, productivity is weak, energy is expensive, defense is underbuilt and startups do not scale, but the compliance department is having a fantastic decade.

The East is converging because EU accession, investment, lower legacy welfare burdens and fear of Russia create urgency. The West is drifting because France, Italy, Spain, Portugal and Greece have built beautiful museums of social protection around economies that increasingly cannot pay for them.

Mario Draghi says Europe needs innovation, energy, defense and scale, but Brussels says something like "excellent, please first complete Form 19-B on sustainable stapler procurement".

Europe has a growth story.... just mostly in the part Europe used to lecture. LINK

On the fertility front …in China

In 2012, zero fertility desire among 18–24-year-olds was a niche view, around 5%; by 2023 it was 32%, and among young women almost 47%.

The old East Asian fertility model was baded o the asumption that people still want children, but housing, education, work hours, marriage markets, and family norms make the trade impossible. This paper suggests something more dangerous. Constraints are becoming identity.

Young women are not merely saying “I cannot afford the second child”. Many are saying “I do not want the first contract“ and that matters because China’s policy machine is built to solve supply problems (build apartments, subsidize nurseries, adjust hukou, issue slogans), but this is a demand problem.

Parenthood is being evaluated like a bad investment with high upfront cost, lifelong lock-up, unequal labor allocation, weak partner governance, career dilution, and social pressure masquerading as shareholder value.

Beijing spent decades teaching households that children were a state-managed variable. Now educated young women have learned the lesson too well and are managing the variable themselves. The state wants demographic growth, but the consumer wants optionality, and in modern societies, optionality usually wins. LINK

Meta's Interface, Palantir Technologies's Integration

In my latest Where is my MOAT?, I look at Meta and Palantir, two companies trying to prove very different AI stories.

The bigger story is that Meta may have one of the clearest consumer AI opportunities in the market, with billions of users, daily habits, social graphs, creators, commerce, identity, and eventually glasses.

Palantir is the opposite case. The company is excellent, the Rule of 40 numbers are stellar, and Alex Karp’s ambition remains unmatched. But valuation matters. At a very high multiple, Palantir must prove that enterprise AI belongs to its ontology and operating system, not to cheaper, easier, more intuitive tools from Anthropic, OpenAI, or Databricks. WIMM

Share

This week on “Where is my MOAT?”

• 13 May: [Watchlist] The expected IPO of Cerebras

• 14 May: Portfolio update

• 15 May: Macroeconomics (May 2026)

• 16 May: [Market update] META and Palantir

• 17 May: [Deep dive] JPMorgan

Consider subscribing to get that👇🏻 (one month free)

One month FREE trial

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

3. Endi Ungureanu - Tura de duminica

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. The globalization of cheap sugar

2. Google’s Fitbit Air is not really a smartwatch competitor.

3. The magic of “boring businesses”: pet shops

4. Amazon Supply Chain Services (ASCS) is the AWS playbook applied to atoms.

5. Vinted or why people buy second-hand

6. on Google (Alphabet) & Berkshire Hathaway Q1 earnings

Onto the update:

The globalization of cheap sugar

Mixue looks like a beverage chain, but strategically it behaves more like a vertically integrated logistics company disguised as ice cream. More than 60,000 stores globally, most ingredients self-produced, 29 warehouses in China, regional logistics hubs in Southeast Asia, ambitions for factories in Brazil, etc. The reason it works is simple: in a slowing economy, affordability scales better than aspiration. Starbucks sells status, and people love it, but I’d say that Mixue sells a cold dopamine hit for $1 or €1, and increasingly, global consumers seem perfectly happy with that trade.

China is beginning to export not only products, but operating systems for low-cost consumer businesses. Mixue, Luckin, Temu, Shein, these are the common denominators with ruthless supply-chain optimization combined with aggressive pricing and franchise-like expansion economics. The challenge internationally is whether the company can preserve that efficiency outside China, because the farther away it moves from its integrated manufacturing base, the more fragile the model becomes. But if it succeeds, Western consumer brands may discover that their real competitor is not another premium brand, but Chinese industrial efficiency wrapped in a cartoon snowman. FT

Google’s Fitbit Air is not really a smartwatch competitor.

At $100, screenless, cross-platform, and paired with a $10/month health coach, the point is not to give users another device to look at, but to make biometric tracking invisible and recurring. That is dangerous for WHOOP and its differentiation with "no screen, serious health data, subscription model", and Google is now offering a cheaper, more accessible version with far broader distribution.

For Apple Watch, the threat is subtler. Apple still owns the premium wrist computer category, but Fitbit Air attacks the opposite "job-to-be-done" (= less distraction, lower price, longer battery, and health coaching rather than notifications).

The broader implication is that wearables are splitting into two markets:
1/ devices that extend the phone, and
2/ devices that quietly collect health data for AI interpretation.

Apple dominates the first. Google is trying to rebuild Fitbit around the second. Bloomberg

The magic of “boring businesses”: pet shops

There is a funny thing about “boring” businesses.

They usually become interesting exactly when people stop treating the customer as rational.

If 71% of Romanian households have a pet, then this is not a niche, it is a consumer category with emotional pricing power and this is th exact point when people start buying premium food, grooming, medicine, accessories, insurance, daycare, boarding and oncology for “an animal”.

We buy it for a family member who happens to shed hair on our sofa and as pets live longer, they acquire the same wonderful economic feature as humans:
1/ recurring needs
2/ age-related problems (ie. cancer, diabetes, blood pressure, etc)
3/ higher medical complexity and owners who feel guilty saying no.

Indeed, this is a boring business.

Which is another way of saying ... "high frequency, emotional demand, fragmented services, expanding basket size, and probably better margins than many 'exciting' businesses that lose money delivering sandwiches".

Amazon Supply Chain Services (ASCS) is the AWS playbook applied to atoms.

First you build internal infrastructure at an absurd scale, after that, make it reliable because your own business depends on it, then sell the excess capability to everyone else. Freight, fulfillment, distribution, parcel delivery - this is Amazon turning its operating system for commerce into a platform for the physical economy.

The early customers matter too (e.g., P&G, 3M, Lands’ End, American Eagle). This is Amazon saying "we don’t only want to help merchants sell on Amazon. We want to run the logistics layer underneath commerce itself."

The implication is brutal for traditional logistics providers. Amazon is competing on integration, visibility, inventory positioning, forecasting, and speed. Once a company plugs into that network, Amazon learns demand patterns, inventory flows, and delivery economics across the broader economy. That is the real moat, with logistics data plus operational density.

AWS monetizes idle server capacity, while ASCS monetizes idle supply-chain capacity. The cloud was computing as a utility, but this is logistics as a utility.

Vinted or why people buy second-hand

Vinted is what happens when a good user story meets a difficult business model. Yes, it’s worth EUR 8 bn and growing fast, but the financials tell a more complicated story. Revenue up 38%, profit down 20%, and still trading at something like 100x forward earnings if you extrapolate current growth. The reason is simple. Peer-to-peer marketplaces scale cheaply on supply, but expensively on demand. You don’t warehouse inventory, but you do spend heavily to acquire both buyers and sellers and, more importantly, to build trust in a market where quality is uncertain by default.

The deeper point is that second-hand fashion works brilliantly for consumers and awkwardly for investors. The demand side is real, driven by price sensitivity and macro pressure, not just sustainability, but the supply side is fragmented and the take rate is structurally limited. Platforms like Vinted are caught in a classic marketplace trap: lots of activity, thin margins, and constant competition from both incumbents (eBay, retailers) and new entrants.

In other words, turning “preloved” into a scalable, high-margin business is much harder than turning it into a popular one. FT

Subscribe now

on Google (Alphabet) & Berkshire Hathaway Q1 earnings

In my latest Where is my MOAT? market update, I argue that Google and Berkshire are really two very different versions of the same posture: be patient, but for different reasons.

Google’s quarter says the AI capex is starting to work. Cloud is compounding, TPUs are becoming strategic, and part of the profit surge was really the market marking up Google’s Anthropic stake, which is another way of saying that Google is now monetizing both the AI boom and the picks-and-shovels behind it.

Berkshire, meanwhile, is doing the opposite of excitement, which is, sitting on an almost absurd cash pile, waiting for a world still too liquid and too crowded to offer Buffett-style bargains.

One company is deploying capital at an industrial scale for the future, while the other is waiting for the future to panic and come to it. WIMM

Share

This week on “Where is my MOAT?”

• 6 May: [Essay] Feedback from my visit in the US

• 7 May: [Market update] Amazon and Apple

• 8 May: Q&A WIMM (on Zoom), from 17:00 (EEST, Bucharest time zone). Usually for Western Europe, that is 16:00.

• 10 May: [Market update] Google and Berkshire Hathaway

• 11 May: [Deep dive] Coca Cola

Consider subscribing to get that👇🏻 (one month free)

One month FREE trial

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

3. Endi Ungureanu - Tura de duminica

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. Vinted or why people buy second-hand

2. NVIDIA’s push into Physical AI brings new winners

3. Germany’s growth is around…zero

4. What Tim Cook meant for Apple

5. on AI Automation & costs

6. on EV adoption

7. on universities

Onto the update:

Vinted or why people buy second-hand

Vinted is what happens when a good user story meets a difficult business model. Yes, it’s worth EUR 8bn and growing fast, but the financials tell a more complicated story. Revenue up 38%, profit down 20%, and still trading at something like 100x forward earnings if you extrapolate current growth. The reason is simple. Peer-to-peer marketplaces scale cheaply on supply, but expensively on demand. You don’t warehouse inventory, but you do spend heavily to acquire both buyers and sellers and, more importantly, to build trust in a market where quality is uncertain by default.

The deeper point is that second-hand fashion works brilliantly for consumers and awkwardly for investors. The demand side is real, driven by price sensitivity and macro pressure, not just sustainability, but the supply side is fragmented and the take rate is structurally limited. Platforms like Vinted are caught in a classic marketplace trap: lots of activity, thin margins, and constant competition from both incumbents (eBay, retailers) and new entrants.

In other words, turning “preloved” into a scalable, high-margin business is much harder than turning it into a popular one. FT

NVIDIA’s push into Physical AI brings new winners

NVIDIA is becoming Asia’s demand-generation engine. The striking number is that Asian suppliers now represent roughly 90% of Nvidia’s production costs, up from about 65% last year, which means every incremental dollar of AI capex is being transmitted directly into Taiwan, Korea, Japan and parts of China. This is why obscure suppliers suddenly trade like strategic assets. The market is buying their place inside Jensen Huang’s machine.

Generative AI made Nvidia the operating system of the data center. Robotics, autonomous systems and AI-enabled manufacturing could make it the operating system of the real economy. That pushes value beyond GPUs into memory, servers, components, sensors, automotive systems and factory automation.

In other words, the AI trade is broadening from “buy Nvidia” to “buy the industrial stack Nvidia needs”. Asia built the hardware base for globalization and now it may become the factory floor for AI. Bloomberg

Germany’s growth is around…zero

🇩🇪 Germany’s problem is the lack of transmission. The chart from Bloomberg shows a steady downward revision cycle, which is what happens when fiscal impulse meets structural friction (= €500bn of spending, and yet growth keeps getting marked down because energy shocks, weak confidence, and an export model under pressure are canceling out the multiplier).

In other words, Berlin is pushing demand into an economy that can’t respond fast enough on the supply side (e.g., bureaucracy, energy costs, and industrial repositioning are acting like a tax on growth). The result is the worst macro mix with fiscal expansion without acceleration.

What makes this more fragile is timing. The Merz plan was supposed to reignite momentum after near-zero growth (~0.2% 🫠), but instead it is colliding with a geopolitical shock that hits Germany exactly where it hurts most (energy and industry). The risk is not recession per se, but stagnation with inflation pressure creeping back, a quasi-stagflationary setup where policy space narrows (ECB tightening risk) just as growth disappoints.

If the stimulus doesn’t translate into visible productivity gains quickly, confidence becomes the binding constraint and once that breaks, no amount of fiscal spending can fully offset it.

P.S. 0.2% growth? US growth is between 2-4% per quarter

What Tim Cook meant for Apple

In my latest Where is my MOAT? analysis, I argue that Tim Cook’s achievement was not to replace Steve Jobs, but to institutionalize him, which is, to turn taste into process, product vision into global operations, and Apple into the most efficient premium hardware machine ever built.

The Cook doctrine (ie. focus, simplicity, control of core technologies, deep collaboration, and saying no to almost everything) was the operating system for that transformation, and it worked brilliantly in a deterministic era where the goal was to make hundreds of millions of perfect objects, on time, every time.

However, that success also leaves 3 very un-Cook-like open questions for the next CEO. WIMM

on AI Automation & costs

After a point, AI is a cost reallocation tool.

Companies think they are buying efficiency, but what they are really buying is capacity.

Same headcount (or slightly less), more output, and a much more complicated cost structure that includes vendors, infrastructure, and ongoing model usage... and that’s before you realize that only ~20% of firms have actually reduced headcount in practice, which suggests the savings story is mostly aspirational. Gartner

on EV adoption

This chart is a nice reminder that the future of cars is both obvious and extremely uneven, which is how most inevitabilities look in real time. Norway is basically done with internal combustion, the Netherlands is well on its way, China is moving fast (and has a huge interest in doing this), and then you have the US, India, Brazil doing a sort of "we’ll get there eventually, maybe" shuffle.

The direction is not really in doubt, because EVs win on the things that compound, like fewer moving parts (so less maintenance!), improving battery economics (so better cost curves), software integration (so better user experience), and, crucially, policy alignment (subsidies, regulation, infrastructure).

ICE cars, by contrast, are a mature technology with limited room for improvement and increasing regulatory pressure. The interesting question is not if EVs win, but how fast each market is forced into that equilibrium, which depends less on consumer preference and more on infrastructure buildout, government incentives, and whether local incumbents can survive the transition long enough to slow it down.

In other words, EV adoption looks like a market choice, but it behaves like a policy-driven inevitability.

on universities 🎓

Universities are one of history’s great regulatory arbitrages. They began as guilds for teachers, somehow acquired monasteries, dorms, stadiums, hospitals, laboratories, enormous real estate portfolios, and the moral right to fight with absolutely everyone, and then called that "shared governance".

The paper’s core insight is that this bizarre structure is not a bug but the feature. Universities survive because they are half medieval cooperative, half asset-holding corporation, which means the faculty get enough autonomy to pursue weird ideas and the institution gets enough control to keep the buildings from being sold to fund someone’s obscure seminar on 14th-century punctuation. The amazing part is that this arrangement is wildly inefficient in the normal corporate sense and yet extremely robust in the civilizational sense.

Companies optimize for quarterly margins, but universities optimize for being impossible to kill. The chart on page 2 makes the joke nicely. Almost all of the top endowed universities are older than 50 years, while a much smaller share of Fortune 500 firms make it that long. That is not because universities are better run, but because they are harder to liquidate, harder to discipline, and weirdly good at converting donor money and government money into permanence.

And then there is the really fun part, which is that the same faculty autonomy that produces Nobel prizes also produces endless trench warfare with presidents, trustees, popes, politicians, and each other. That is not accidental either. If you want curious, status-obsessed intellectual entrepreneurs, you are not hiring agreeable middle managers, but you are hiring people who will absolutely write a 484-page book about how their boss is ruining the university, and then demand tenure so they can do it again next year.

The broader implication is pretty brutal for anyone who wants universities to behave like efficient firms or neat bureaucracies They can’t, because their productive asset is misaligned human capital. The same disorder that makes universities embarrassing also makes them generative. You can centralize them more and maybe get cleaner budgets, but you probably get fewer discoveries. You can free them completely and you get more genius plus more chaos.

Concluding, the real lesson is that universities endure because they found a stable way to remain permanently unresolved. NBER

Subscribe now

[Essay] Function or Business Unit? Designing the firm for the AI age

In my latest Where is my MOAT? essay, I argue that AI sharpens the old organization-design debate.

Functional structures still win where integration and technical excellence matter most; business units still win where customer heterogeneity and speed matter most, but AI changes the economics of both by making centralized data/model/platform capabilities more valuable while simultaneously making smaller autonomous teams easier to coordinate. (continue reading HERE)

This week on “Where is my MOAT?”

• 30 April: [Essay] Organized by Business Unit or Function?

• 2 May: [Analysis] What Tim Cook meant for Apple

• 3 May: [Deep dive] Samsung

Consider subscribing to get that👇🏻 (one month free)

One month FREE trial

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

3. Endi Ungureanu - Tura de duminica

Dear OnStrategy Reader,

Greetings from Boston, 🇺🇸!

Onto the (short) update:

End of an era: on Apple's old and new leadership

For 15 years, closes one of the most successful "operator eras" in business history. As per "Apple in China" book (by Patrick McGee), we understand clearly that Tim Cook industrialized the iPhone.

He turned Apple into a $4 trillion machine, scaled supply chains globally, built Services into a Fortune 50 business, expanded the ecosystem to 2.5 billion devices, created the Apple silicon that makes all the differences in the world, and, in general, executed with a level of discipline that most CEOs never reach.

This is what great operators do. They change the outcome, not the product. And yet, there was one structural miss. Apple became the best company in a deterministic world… just as the world shifted to a probabilistic one. Because Apple disregarded this world by 'giving' its search engine to Google for a fee, it meant that when AI arrived, the company was not there.

Which is why the timing is right. Leadership transitions matter most when the environment changes, not when performance declines. John Ternus is a product-first leader, stepping into a company that now needs exactly that (= a new interface for computing, built around AI-native devices and experiences)

I’ve been critical of Apple in the last two years, but that criticism exists precisely because the bar is so high. And now, I’m bullish again.

The operator built the empire. The product leader now has to redefine it.

on OpenAI vs Anthropic 🤖

In my latest Where is my MOAT? deep dive, I argue that the real divide between the two is not “which model is smarter”, but how each company is underwriting demand in a world where compute is both moat and liability.

Anthropic looks more disciplined, more enterprise-native, and probably more right about the danger of overbuying infrastructure, while OpenAI still feels more willing to YOLO its way into scale and trust that distribution, workflows, and enterprise integration will eventually justify the spend.

That is why this is no longer a pure model race but a capital-allocation race. One company may be underestimating demand, the other may be underestimating the cost of serving it, and the real answer probably sits in the S-1s.

Subscribe to read (one month free) - click HERE

The Frontier and the Followers

What the DeepSeek result makes clear is that AI leadership is about sustaining the frontier, and that depends on a full stack of advantages: compute, capital, talent, and distribution. Chinese labs have been highly effective at distillation, ie. taking frontier models, optimizing them, and deploying them efficiently… but distillation is downstream of innovation. If access to the frontier is constrained, either by geopolitics, capital restrictions, or closed ecosystems, then the strategy becomes inherently lagging. The recent moves to limit US capital and tighten control over cross-border flows may be rational politically, but economically, they reduce iteration speed and integration into the global AI ecosystem, which is exactly where advantage compounds fastest.

The implication is that AI is becoming an aggregation market, and the US currently owns the key layers: frontier models, hyperscale compute, and global distribution. This is about setting the pace of innovation that others must react to. Without continuous access to that frontier, distillation alone is insufficient to keep up, because the gap is not static, but it compounds. That is why the debate is less about whether China can compete in AI and more about whether its system allows the same level of open-ended experimentation and capital formation required to lead. In that sense, if AI defines the next economic era, the structural advantages suggest that the 21st century will be shaped more by the American system than the Chinese one.

Share

[Essay] The evolution of the CFO function

Finance is personal to me because I’m a finance guy at heart. I started my career in finance and I still think that you can’t be serious about strategy if you don’t understand finance.

For decades, the chief financial officer (CFO) occupied a well-defined position in the corporate hierarchy. The CFO was the guardian of the ledger, responsible for financial reporting, regulatory compliance, capital allocation, cash management, and the integrity of the numbers that flowed to boards, investors, and regulators. It was a role built on 3 things: (1) precision, (2) discipline, and (3) control. The best CFOs were trusted because they could close the books cleanly, defend the audit, and ensure that the organization never spent more than it could afford.

That mandate has not disappeared, but it is no longer sufficient. A convergence of forces (to name a few: artificial intelligence, real-time data infrastructure, macroeconomic volatility, shifting investor expectations, and the rising complexity of global enterprises) is redefining what organizations need from their finance leaders. The CFO function is evolving from a backward-looking control tower into a forward-looking intelligence hub, where value comes mostly from interpreting uncertainty, shaping decisions, and allocating capital with the help of AI.
Understanding this shift is essential for every finance professional and business leader whose work depends on the quality of financial judgment. (Continue reading it here)

This week in "Where is my MOAT?":

• 23 April - The evolution of the CFO function

• 24 April - Portfolio update (only to Premium subscribers)

• 25 April - Macroeconomics (April 2026)

• 26 April - [Market updates] on ASML, TSMC and Intel

Consider subscribing to get that👇🏻 (one month free) —>

One month FREE trial

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

3. Endi Ungureanu - Tura de duminica

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. Meta takes “search”

2. on Allbirds

3. back on UBER 🚕

4. Regulation drives innovation away

5. 🍵 Matcha is sooo...yesterday (sort of)

6. [Essay] The future of company structure

7. ..and more

Onto the update:

Meta takes “search”

What this really means is that the center of gravity in digital advertising is shifting from explicit intent (deterministic) to probabilistic discovery. Google’s great strength was always that users told it what they wanted. Meta’s great strength is that it infers what users might want before they ask. For a long time, search was the more durable business because intent is the cleanest signal in advertising. AI changes that equation. If Meta can use automation and products like Advantage+ to turn its enormous corpus of behavioral data into better targeting and easier campaign execution, then the ad dollar follows performance, not legacy. In that sense, Meta overtaking Google is a sign that the ad market increasingly values prediction over query.

The implications go further.

First, Meta’s expansion across Instagram, Reels, Threads, and WhatsApp means it is becoming more an ad-surface aggregator, with multiple formats capturing different moments of attention.

Second, Google’s diversification is strategically rational but financially dilutive in this comparison. Subscriptions and other businesses may strengthen Alphabet overall, but they also make the ad engine look less singular versus Meta’s pure optimization machine.

Finally, the concentration point matters most. If Meta, Google, and Amazon take well over 60% of global digital ad spend, then the losers is the broader open web. The market is consolidating around the companies with the best data, the best AI, and the largest distribution, and once that loop closes, it becomes very hard to break. Reuters

on Allbirds

This is one of those trades where the business model is “AI”, but the actual product is volatility. Allbirds pivoting from sneakers to GPU cloud is, on its face, absurd, but the financing is extremely rational. You attach a credible buzzword to a listed shell, watch the stock re-rate, and then hand a convertible investor a call option on that re-rating with basically no downside.

The investor isn’t really betting on Allbirds becoming the next Nvidia. They’re betting that the market will believe it might, at least long enough to convert cheap paper into expensive stock and sell it to someone else who likes the story. The clever part is the structure: small upfront capital, optional follow-on, and a conversion mechanism that actually benefits from volatility, which is perfect for a meme-driven market where prices oscillate between “this is the future” and “wait, this was a shoe company” 🫠

Finally, you end up with two parallel realities. In one, Allbirds is building “AI-native cloud infrastructure”, which is a very capital-intensive, expertise-heavy business that it is not obviously equipped to run. In the other, Allbirds is a financing vehicle for harvesting AI enthusiasm, where the real innovation is not GPUs but capital structure.

Honestly, the second one is working great. The deeper implication is that in frothy markets, the fastest way to create value is not to build something new, but to relabel something old and attach the right kind of optionality to it. The only real question is timing. Can you convert and sell before the narrative collapses, or do you end up holding a very ambitious cloud strategy and a warehouse full of wool sneakers? BBC

back on UBER 🚕

Uber’s $10bn pivot into robotaxis is a normal existential hedge. The company built its moat on aggregating supply (drivers) and demand (riders), but autonomy threatens to collapse that supply layer entirely, handing power to whoever owns the fleet and the software. What’s striking is the inversion where Uber, once the archetype of asset-light scaling, is now committing billions to secure access to vehicles and even financing their deployment, effectively trying to remain the interface while conceding control of the core technology. The risk, of course, is that the real winners (Waymo, Tesla, Zoox) have no structural reason to tolerate an intermediary if they can go direct, which turns Uber’s strategy into a race to prove that distribution still matters in a vertically integrated world.

If Uber can remain the demand aggregator, it preserves relevance. If not, it becomes a transitional layer in a stack that is rapidly integrating around it. FT

Regulation drives innovation away

There’s a funny thing about being a “super-regulator”. It works great right up until the moment something actually important happens. Then you discover that the people building the thing are in California, the chips are in Arizona (or Asia), the cloud is in Virginia, and the guest list is…also in the US.

Europe, meanwhile, is left refreshing its inbox waiting for access to a model that can apparently out-hack most humans, while the real decisions are being made in private rooms with companies that actually ship products. It turns out that if your comparative advantage is writing rules after the fact, you don’t get invited to the meeting where the future is decided.

Europe spent a decade perfecting the art of regulating technology it didn’t build, and now it’s trying to regulate technology it can’t even access. The AI Act says “you must comply”, but Anthropic says “you can’t test it”... and, well, those are different kinds of power.

Regulation assumes leverage, but leverage comes from markets, capital, and capability. The US has all three, so it can decide who sees the model. Europe has…process. Which is useful for writing very thoughtful documents about risks, but less useful when the risk is that you’re no longer relevant.

I wrote here, on LinkedIn and in my newsletter, that the EU will eventually get a scrappy internet. Now will get a scrappy AI. POLITICO

🍵 Matcha is sooo...yesterday (sort of)

The fun thing about "ube" is about color arbitrage. You take a perfectly normal beverage category, add a visually distinctive ingredient, and suddenly you can charge matcha prices for something that is, in wholesale terms, cheaper to produce. Which is great, because coffee chains are not actually in the business of selling coffee anymore, they are in the business of selling reasons to walk into a store and post about it on Instagram (prove me wrong 😎)

Matcha worked because it was green and vaguely healthy and showed up well on TikTok, but ube (what a horrivle name!) works because it is purple and shows up even better. If tomorrow there is a bright blue root vegetable, that will probably be next quarter’s strategy.

But also this is a slightly defensive move dressed up as innovation. The big chains are losing share to smaller, trend-native competitors, and when consumer spending is under pressure, the easiest lever is novelty. Launch something limited, make it look good, price it like a premium product, and hope it brings in younger customers who were otherwise going somewhere else. The risk is that novelty is not a moat, it is a treadmill, so you end up in a situation where you are constantly searching for "the next matcha", which is a polite way of saying that the last one already peaked. FT

on “The Great Flattening” & McKinsey & Company

What sits between the lines in the article below is a redefinition of the firm itself. Historically, organizations scaled by adding people to process information, which is why hierarchy existed in the first place. Now, with agents embedded directly into workflows, the coordination layer shifts from humans to software. The implication is flatter org charts, but also a fundamentally different operating model where “teamwork” becomes a mix of humans and agents executing tasks in parallel, with AI continuously synthesizing inputs and pushing decisions forward.

In that world, management is more about designing systems (ie. setting objectives, guardrails, and data flow), while agents handle the execution and increasingly the analysis itself.

The second-order effect is that decision-making becomes ambient. Instead of waiting for reports, meetings, and approvals, the organization is constantly being nudged by AI with “do this", "resolve that", “here are the implications”.... effectively compressing time between signal and action.

This collapses layers not because companies want to cut costs, but because the function those layers served is being automated. The strategic advantage then shifts to companies that can best integrate these signals into their operations. In other words, the org chart gets flatter, but it also becomes dynamic, continuously reconfigured by the flow of information itself. BI

[Essay] The future of company structure

In the AI era, the real question is not whether to organize by function or by business unit. It is where decisions should be made, which capabilities scale best from the center, and where autonomy creates more value than control.

For decades, one of the most consequential choices a CEO could make was deceptively simple: Should the company be organized primarily by function or by business unit? Alfred Chandler’s dictum, structure follows strategy, still holds. Apple’s way of operating is still hard to copy (see How Apple is organized for innovation). But in an era of artificial intelligence, real-time data, and automated workflows, something more fundamental is shifting.

Strategy itself is now shaped by the information architecture a company builds, and the old trade-offs between functional efficiency and business-unit agility are being rewritten. (continue reading)

This week on “Where is my MOAT?”

• 16 April - WIMM plan (update)

• 17 April - [Essay] The future of company structure

• 18 April - [Analysis] Open AI vs Anthropic - deep dive on early data

• 19 April - [Deep Dive] Johnson & Johnson

Consider subscribing to get that👇🏻 (one month free) —>

One month FREE trial

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

3. Endi Ungureanu - Tura de duminica

Personal update 🤓

I did not particularly expect to become a certified investment consultant....and yet, here we are.

What started as writing several times a week on Where is my MOAT? gradually became something bigger:

• a way of thinking about businesses, markets, incentives, technology, commodities, and...

• capital allocation that is a bit more structured than “buy this” or “sell that”

Markets, after all, are rarely just about price targets. They are about business models, power, scarcity, timing, management quality, regulation, second-order effects, and the inconvenient fact that everything connects to everything else.

After publishing 4-5 times a week, I realized I had built something like an MBA niche for investors. Not the generic kind, but the “skin in the game” useful kind. The kind that helps you understand why a company matters, what could change its moat, how commodities shape industries, and how all of this affects the decisions you make with your own capital.

I’m officially certified as an investment consultant, and I now work with people in three ways:
1/ on a one-time basis
2/ on a monthly basis, or
3/ through tailored investment education and advisory support that companies can offer as a benefit to employees or clients.

Send me a message to work together.

👉 Find all details here: https://whereismymoat.com/work-with-me

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. on Perplexity

2. [Analysis] Winners and losers in the memory shortage

3. Berkshire Hathaway has $373 billion in cash

4. on the middle management 🧑‍💼👩🏼‍💼

5. on mega IPOs: SpaceX, OpenAI and Anthropic

6. on Intel

7. [Essay] Apple at 50

Onto the update:

on Perplexity

Perplexity’s pivot from "search" to "agents" is a recognition that “answering questions” is a feature, while “doing work” is a business model. Search is a brutally difficult market because incumbents own distribution, advertisers subsidize user acquisition, and the marginal answer is easy to compare.

Agents, by contrast, are closer to labor. They consume more compute, but they also create more measurable value, which makes usage-based pricing feel natural rather than forced. That is why the revenue inflection matters. Moving from a search-like interface to a task-completion layer turns Perplexity from a thin wrapper around model access into something more like an intelligent broker of work, routing each request to the most efficient model and capturing margin through orchestration rather than through model ownership itself.

I start to feel more optimistic about Perplexity's future. FT

[Analysis] Winners and losers in the memory shortage

The first and most obvious beneficiaries are the memory manufacturers themselves. Samsung, SK Hynix, and Micron Technology collectively control more than 95% of global DRAM production and the majority of enterprise NAND supply. After suffering through a bruising 2023 downcycle, all three companies implemented aggressive supply discipline, which meant cutting capital expenditure, reducing wafer starts, and delaying capacity expansions. This was a strategy born of survival. The result is that when demand returned with unexpected force (driven largely by HBM - High Bandwidth Memory), requirements for AI accelerators, these 3 players found themselves holding pricing power they had not enjoyed in years. Of course, the market ended like this after decaded of transformation. (continue reading it HERE)

Berkshire Hathaway has $373 billion in cash

Berkshire Hathaway having $373 billion in cash is one of those things that sounds conservative until you realize it is also extremely aggressive, just pointed in a different direction. Buffett is basically running the world’s largest "wait and see" trade, which is funny because most investors feel compelled to always be doing something, whereas Buffett (Greg Abel more exactly) is very comfortable doing nothing at scale....and the scale matters. When you have that much cash you are waiting for everyone else to run out of options first, at which point you become the market.

Also, there is a subtle signal here about the opportunity set. Cash piles this large are not built because there are too many attractive investments, they are built because there are too few that meet the bar. So either Abel has become unusually cautious, or the market has become unusually expensive, or both, which is a slightly awkward backdrop for everyone else who is fully invested and explaining why this time is different.

The nice thing about having $373 billion in cash is that you don't need to be right today, you just need to be right eventually, and preferably at a moment when everyone else wishes they had been a bit more patient.

on the middle management 🧑‍💼👩🏼‍💼

The core idea in Jack Dorsey's new essay is that hierarchy is more about bandwidth..not power. For 2,000 years, organizations existed to route information through layers of humans with limited span of control, and everything from Roman legions to McKinsey org charts was essentially an optimization of that constraint.

What changes with AI is not just productivity, but the removal of that bottleneck. If a system can maintain a real-time “world model” of the company and its customers, then the coordination function of middle management becomes software rather than people. That reframes the firm from a collection of teams executing plans to a continuously updating system composing actions, where capabilities are modular and decisions emerge from data rather than being pushed down a chain of command.

The interesting second-order effect is that this turns organizational design into a data problem and a moat problem. The companies that win are not those with the best org charts, but those with the richest, most compounding "world models", built on proprietary, high-signal data like transactions or workflows. That creates a feedback loop where better models drive better decisions, which drive more activity, which enriches the model further. At that point, AI is not a cost-saving tool but a coordination engine that collapses layers and accelerates iteration speed, effectively making “speed of information flow” the new primary competitive advantage.

The risk, of course, is that if your company does not have a unique data advantage, this architecture becomes commoditized, and instead of becoming an intelligence system, you become just another node running someone else’s model.

Anyway, make some time to read his essay. LINK

on mega IPOs: SpaceX, OpenAI and Anthropic

The thing about mega IPOs is that they are sold as "once-in-a-generation opportunities", which is usually true in the sense that they generate once-in-a-generation underwriting fees and then, for public investors, once-in-a-generation disappointment.

This chart is basically saying that if you bought the biggest, most hyped IPOs after Facebook, you systematically underperformed the market for the next five years, which is a nice way of saying you paid peak narrative prices for businesses that had already done most of their compounding in private markets. Of course, now we are about to do it again, but bigger, with SpaceX, OpenAI, Anthropic, and friends lining up, which is great because nothing says “healthy market” like a queue of trillion-dollar stories trying to come public at the same time.

Also, there is a nice timing element here, which is that companies tend to IPO when conditions are good for them, not for you, and right now conditions are good in the sense that there is still liquidity, still excitement, and still a willingness to believe that this time the growth curve will justify the valuation.

Here is the hack: if you are the issuer, you might also notice that macro conditions feel a bit wobbly, rates are not exactly friendly, and the window could close, so you accelerate. Which means that what looks like a supply of exciting new public companies is also a demand for your capital at exactly the moment when the risk is highest, and historically that has not been a great trade, even if the companies themselves turn out to be very good businesses eventually. Bloomberg

on Intel

In my latest Where is my MOAT? deep dive, I argue that Intel’s problem is not that it has no moat, but that it has two very different ones living under the same roof:

1/ a still-real CPU franchise built on x86 inertia, enterprise switching costs, and the growing relevance of CPUs in AI orchestration and inference, and

2/ a foundry business that is strategically important but still economically unfinished.

The quarter was better than feared, especially in data center, and management effectively admitted the issue was demand exceeding supply rather than demand collapsing, which is a meaningful change in tone, but the real question is still execution: whether 18A yields improve, supply normalizes, and foundry starts turning milestones into real external customers.

Until then, Intel remains a company with valuable assets and credible strategic optionality, but one whose turnaround still lives more in verbs than in cash flows.

Read the full analysis here.

[Essay] Apple at 50

In my latest Where is my MOAT? essay, I argue that Apple’s real genius over the last 50 years was building the greatest customer-creation machine in consumer tech: 5.1 billion devices shipped, 2.5 billion active devices today, and a services layer now worth $112.8B annually, all reinforcing one another through control of the human-computer interface.

The more interesting question is forward-looking. Apple may be perfectly built for the deterministic era it dominated (ie. polished hardware, controlled experiences, elegant integration), while the next era belongs to probabilistic AI systems shaped by messy data, iteration, and uncertainty, where other companies may be structurally better positioned. [Continue reading HERE]

Share

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

3. Endi Ungureanu - Tura de duminica

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. on the SpaceX IPO 🚀

2. Where does the value accrue in AI?

3. How Waymo works

4. on Social Media ban for youngsters (< 16 yrs)

5. on inflation 📈

6. [ESSAY] on AI needs consultants 🧑🏼‍💻👩🏼‍💻

Onto the update:

on the SpaceX IPO 🚀

SpaceX going public at a rumored $1.75tn valuation and $75bn raise is, on one level, a financing event and, on another, a queue-jumping event. The market right now is basically a finite pool of excitement, liquidity, passive flows, and tolerance for heroic narratives, and SpaceX seems to have decided that if there is a window, it should go through it first and largest.

That makes sense. If you know that OpenAI and Anthropic are somewhere behind you in the IPO line, you do not politely wait your turn, but you run onto the runway with rockets, satellites, Starlink cash flows, an xAI tie-in, and a valuation so large it makes normal IPOs look like bake sales.

It also helps that the exchange rules have just become friendlier to giant newly public companies, with faster index inclusion and looser float requirements, which is a very elegant way of saying ...“come public fast, sell a tiny slice, and let passive money do some of the work".

The other thing about doing it now is that “now” may be better than “later”. Economic conditions feel like they are deteriorating quickly, and mega-IPO markets are open right up until the moment they are not. If you are Elon Musk and friends, the rational move is to extract the money while the story still clears at maximum altitude, before macro conditions worsen, before rates or risk appetite shift again, and before the next great AI float sucks oxygen out of the room.

There is also a nice bit of financial engineering here. Float less than 5%, maybe let insiders sell on day one, get potentially pulled into the Nasdaq 100 quickly, and create immediate demand from index-linked funds. That is turning market structure into a distribution channel. In other words, SpaceX is not merely doing the biggest IPO in history. It is trying to do the biggest IPO in history before history gets more crowded and more expensive. FT

Where does the value accrue in AI?

The fun thing about this chart is that it looks like a classic “apps win in the long run” story, except we are very much not in the long run yet, and in fact the opposite is happening. All the money is sitting at the bottom of the stack, with semis hoarding something like 70% of revenue and even more of the profits, which is a polite way of saying that if you want to get rich in AI, you should sell very expensive shovels to people who think they are digging for gold.

And everyone knows this, which is why the hyperscalers are spending hundreds of billions on capex, partly to build products, but also partly to avoid paying NVIDIA forever, which is not a great long-term business strategy if you are, say, a trillion-dollar company that does not enjoy being someone else’s customer.

But also this is sort of how every platform shift starts: the infrastructure layer gets paid first, the application layer gets paid later, and in the middle there is a lot of venture capital explaining that this time will be different. Maybe it will be, but right now the economics are stubbornly inverted, with apps growing fast in percentage terms but still tiny in absolute dollars, which means that the big strategic question is not “who has the best AI app?”, but “who can escape the tax imposed by compute?” and until that changes, the real business model of AI is less about intelligence and more about electricity, chips, and the willingness of large companies to spend enormous amounts of money today in the hope that someday, somewhere up the stack, someone will figure out how to actually make it back.

How Waymo works

In my latest essay, I argue that Waymo’s real lesson was about strategy under deep uncertainty. Some problems do not yield to iteration, because they have a minimum investment threshold below which you are simply not solving the same problem.

Waymo won by treating full autonomy as a distinct category from driver assistance, building “the Driver” as a platform rather than a feature, and spending two decades de-risking the stack until it could generalize across cities instead of merely performing well in a demo zone.

The broader point is uncomfortable but useful for executives. When the problem is genuinely hard, the shortcut is often the most expensive path of all. WIMM

on Social Media ban for youngsters (< 16 yrs)

The interesting thing about banning social media for youngsters is that it sounds like a moral argument. Still, it’s actually a market design problem. We built the most sophisticated attention-extraction systems in history and then let the least experienced users compete in them with no safeguards.

These platforms are optimized for engagement, not well-being, which means the equilibrium outcome is predictable (ie. more time spent, more dopamine loops, less focus, worse mental health). Now, saying "let kids use it responsibly" is a bit like saying "let them trade derivatives responsibly", technically possible, empirically unlikely.

If you look at that map, what you’re seeing is governments slowly realizing that the default settings of the internet are misaligned with developmental (mental) needs, and that the cost of inaction compounds over time.

Limiting access is about sequencing it; there is an age for learning algebra, and there should probably be an age for algorithmic feeds too.

Yes, I'm in favour of the ban! Bloomberg

on inflation 📈

What’s striking about this chart is not the level of inflation, but the revision. Inflation is being repriced upward in real time, and the UK stands out as the most sensitive node in that system.

That tells you something structural, which is...exposure. Economies that are more dependent on imported energy and global trade flows are effectively “levered long” geopolitical stability, and when that breaks, their inflation moves first and fastest.

The recent Iran conflict, which has disrupted oil flows (and other byproducts) and pushed energy prices sharply higher, acts as a classic external shock that feeds directly into CPI through fuel, transport, and food costs.

The second-order effect is more interesting. Inflation here becomes a supply chain signal, not just a monetary one. Central banks can manage demand, but they cannot reopen the Strait of Hormuz or stabilize oil infrastructure (20 million barrels were flowing daily from the Middle East). That creates a regime where inflation is both higher and more persistent than expected, while growth weakens (i.e., a mild version of stagflation).

In that world, the dispersion across countries matters more than the average. India and Brazil can run hotter inflation because of growth dynamics, the US can absorb it with relative resilience, but Europe, and especially the UK, ends up caught between imported costs and constrained policy space.

This is also a reminder of any geopolitical book: secure ASAP your (1) energy and (2) food sources. I guess that many read geopolitical books anymore...especially in Brussels or 10 Downing Street.

Share

[ESSAY] on AI needs consultants 🧑🏼‍💻👩🏼‍💻

In my latest Where is my MOAT? essay, I argue that the AI boom is creating a surprisingly old-fashioned winner, which is, the person who actually shows up and makes the thing work.

Models are getting cheaper, more available, and more impressive, but that mostly means the scarce asset shifts from access to implementation (ie. redesigning workflows, reassigning decision rights, training teams, setting guardrails, and measuring whether any of this improves outcomes rather than just producing a nicer board deck).

In other words, the real moat in enterprise AI may not belong to the lab with the smartest model, but to the companies and advisors who become the translators between generic capability and messy organizational reality.

This week in “Where is my MOAT?”:

• 2 April: [Essay] AI needs consultants

• 3 April: [Market updates] ARM

• 4 April: [How startups work] Waymo —> new category to WIMM

• 5 April: [Deep dive] on Intel

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

This week, my subscribers receive 5 posts. Here’s the schedule for this week:

• 25 March - [Essay] on business models: B2B vs. B2C (with a focus on OpenAI)

• 26 March - [Analysis] Macroeconomics (March 2026)

• 27 March - [Market updates] NVIDIA

• 28 March - [Deep dive] Reddit

• 29 March - [Deep dive] Coinbase

…and here is the schedule for the next week:

• 2 April: [Essay] AI needs consultants

• 3 April: [Market updates] ARM

• 4 April: [How startups work] Waymo —> new category to WIMM

• 5 April: [Deep dive] on Intel

If you find WIMM useful, please share it with a friend or a colleague. WIMM is free for one month ... plenty of time to decide if it’s worth keeping in your stack.

WIMM

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. The opportunity cost of populism (🇪🇸 Spain vs. 🇦🇷 Argentina)

2. The Cost of Focus in AI

3. 🏎️ Ferrari and the economics of scarcity

4. The unbundling of Disney

5. on Macroeconomics (March 2026)

6. ..and more

Onto the update:

The opportunity cost of populism (🇪🇸 Spain vs. 🇦🇷 Argentina)

Argentina is what happens when a country confuses redistribution with growth for just long enough that the bill compounds. In 1900, Argentina looked like Spain’s richer cousin, but by 2022, Spain looks like the country Argentina thought it already was. The gap is explained by policy persistence, ie. decades of “free stuff” financed by inflation, debt, and eventually credibility. The market is very forgiving in the short term (ie. you can print, subsidize, protect), but it is extremely unforgiving over a century. Compounding works both ways.

Javier Milei’s diagnosis is directionally right, because incentives matter, prices matter, deficits matter. But the uncomfortable part is that macroeconomic damage is easier to create than to reverse. You can destroy trust in a currency in a few years, while rebuilding it takes a generation. You can distort capital allocation quickly. However, fixing it requires sustained discipline that democracies rarely maintain.

Argentina’s story is not about one bad decision, but about many rational short-term decisions that, over time, made long-term growth almost impossible. That is the real chart here. LINK

The cost of focus in AI

What this “Code Red” moment really reveals is that OpenAI is transitioning from a research-driven organization with optionality to a product-driven company with constraints.

In the pre-AI era, tech companies could explore adjacent opportunities because the marginal cost of experimentation was mostly people. Now, in the AI era, every experiment consumes scarce compute and strategic focus. Scrapping Sora and other side projects is therefore less about failure and more about prioritization. The winning strategy in AI is in the depth around a core use case, which for OpenAI is clearly ChatGPT as an “all-purpose assistant”.

The implication is that AI competition will converge around distribution and integration rather than pure model capability, and here Google is the best positioned to win. OpenAI’s early lead came from being the default entry point for users, but that advantage erodes as incumbents like Google and Meta leverage existing user bases.

As a result, OpenAI’s pivot toward enterprise and deeper product integration mirrors the historical trajectory of platforms like Google: initial breakthrough, followed by a race to embed into workflows and become indispensable. The companies that win will not be those that build the best models in isolation, but those that align their product surface with sustained usage, turning compute into daily utility rather than occasional novelty. FT

🏎️ Ferrari and the economics of scarcity

Ferrari flying cars to the Middle East is a useful reminder that luxury businesses operate on a fundamentally different set of constraints than traditional manufacturing. Price elasticity is effectively irrelevant, while delivery and customization are core to the value proposition. When customers are willing to pay 4-5x more for air freight, the constraint shifts from cost efficiency to relationship management and brand consistency. In that sense, Ferrari is selling scarcity, personalized experiences, where logistics becomes part of the product itself rather than a backend function.

The broader implication is that in periods of geopolitical disruption, luxury companies are less exposed to demand destruction and more exposed to allocation decisions. The Middle East remains a disproportionately profitable market due to high-margin personalization, meaning that reallocating supply elsewhere is not a neutral decision but a margin trade-off. This highlights Ferrari’s structural advantage: limited supply and high customization create pricing power and flexibility that traditional automakers lack. In a world of constrained logistics and uncertain demand, the ability to choose where and how to deliver, and to whom, becomes a strategic lever, reinforcing luxury as a business model defined by control rather than scale. FT

The unbundling of Disney

Disney’s first week under new leadership is less about bad luck and more about structural exposure. The company is discovering that in a world defined by platforms and AI, owning intellectual property is no longer sufficient to control distribution or monetization. Disney is increasingly dependent on external platforms to extend its reach, yet those platforms operate on very different incentives and timelines. In the past, Disney’s advantage came from scarcity (=owning characters and franchises that others needed). In an AI-driven world, where content creation becomes abundant and cheap, that scarcity erodes, and with it, the leverage Disney once had.

At the same time, the issues around linear TV and declining engagement point to a company caught between two models. This means a legacy distribution system that is structurally declining and a future platform strategy that it does not fully control. The key strategic question is not whether Disney can produce great content (it clearly can), but whether it can build or own the distribution layer where that content is consumed and monetized. Without that control, Disney risks becoming a supplier in someone else’s ecosystem, where value accrues to the platform rather than the creator.

The events of this “horrible first week” are therefore not isolated incidents, but signals of a broader transition, from a vertically integrated media empire to a participant in a platform-dominated economy. FT

on Macroeconomics (March 2026)

In my latest Where is my MOAT? note I argue that the real macro story right now is not any single headline, but the collision of war, energy, inflation, and politics into one ugly feedback loop:
> Iran pushes oil and fertilizer risk back into markets,
> Europe looks increasingly exposed to stagflation,
> US deficits keep pressuring the dollar, and
> even the AI boom starts to look more physical than digital when semiconductors, LNG,
> and supply chains all run through the same geopolitical choke points.

The market feels like a bloodbath today, but the more interesting question is whether this is a permanent breakdown or just a violent reset before investors start pricing the next phase. WIMM

[Essay] NVIDIA GTC 2026: The architecture of dominance

In my latest Where is my MOAT? essay, I argue that Nvidia’s real genius at GTC 2026 was not launching new chips, but disrupting itself before anyone else could.

The company is turning AI infrastructure into a full-stack architecture game where the moat is no longer just the GPU, but the cadence of the entire system with silicon, interconnect, memory, networking, software, and developer lock-in all moving together. WIMM

[PRINCIPLE] You have to write the Playbook

Each path appears unclear at the beginning and no guide contains the full course. A playbook emerges through attention, effort, and reflection. The mind that waits for instruction remains dependent, while the mind that observes and records begins to shape its own method. Order arises from practice, and wisdom forms through repeated action guided by thought.

Take responsibility for your approach and refine it with each attempt. Write what you learn and return to it with discipline. Patterns reveal themselves to those who pay attention and preserve their insights. In this way, your actions gain coherence and your work becomes guided by principles that you have earned. onStrategy

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

This week, my subscribers receive 5 posts. Here’s the schedule for this week:

• 18 March - [Essay] Lessons from Koenigsegg

• 19 March - [Analysis] The disruption of the German car industry

• 20 March - [Market updates] Oracle

• 21 March - [Deep dive] Micron

• 22 March - opening my 3rd portfolio (for Premium/paying subscribers only)

…and here is the schedule for the next week:

• 25 March - [Essay] on business models: B2B vs. B2C (with a focus on OpenAI)

• 26 March - [Analysis] Macroeconomics (March 2026)

• 27 March - [Market updates] NVIDIA

• 28 March - [Deep dive] Reddit

• 29 March - [Deep dive] Coinbase

If you find WIMM useful, please share it with a friend or a colleague. WIMM is free for one month ... plenty of time to decide if it’s worth keeping in your stack.

WIMM

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. OpenAI has a strategy

2. on SAP’s strategy in an AI world

3. on the German car industry 🚘

4. on Pinterest’s options

5. Refining is the bottleneck

6. The AI Capex shock (and two implications)

7. Lessons from Koenigsegg

Onto the update:

on SAP’s strategy in an AI world

SAP’s shift from subscriptions to AI consumption pricing is a redefinition of what enterprise software is, from static systems of record to dynamic systems of work. Subscriptions made sense when value was tied to access, but now consumption makes sense when value is tied to outcomes (i.e., how much work the AI actually does).

As part of the pivot, the company will also create "forward deployed engineering"🫠. SAP is acknowledging that in an AI world, the product is not the software itself, but the integration of that software into the customer’s workflow. That looks less like selling licenses and more like embedding capability, which is precisely where defensibility lives.

The competitive pressure explains the timing. As AI-native players push into enterprise workflows, incumbents can no longer rely on switching costs alone. They need usage-driven value that compounds over time. Consumption pricing aligns SAP with customer ROI, but it also introduces volatility, where the revenue becomes tied to actual usage, which in turn depends on how effectively SAP’s AI is adopted.

In that sense, this is a high-conviction move. SAP is effectively betting that it can move up the stack from infrastructure and data into execution. If it succeeds, it deepens its moat, if not, it risks becoming just another backend provider in someone else’s AI-driven workflow. Bloomberg

on the German car industry 🚘

In my latest Where is my MOAT? analysis I argue that the crisis of the German auto industry is architectural.

For decades these companies dominated the mechanical stack (ie. engines, transmission, precision manufacturing, dealer networks, brand prestige) and then the basis of competition shifted toward batteries, software, power electronics, and supply-chain control, where China moved faster and Tesla rewrote the rules.

The real post-mortem is that German incumbents kept optimizing for a world in which engineering excellence at the component level was enough, just as the industry was becoming a systems business defined by energy, code, and speed of iteration. WIMM

OpenAI has a strategy

Good news: OpenAI has a strategy
Bad news: OpenAI has (only now) a strategy

What looks like a product decision is actually a business model correction. OpenAI is moving from “frontier lab with many bets” to "enterprise software company with a core wedge", and that wedge is (a) coding and (b) productivity. The reason is straightforward. AI only becomes durable when it is embedded in workflows that customers will pay for repeatedly, not when it is spread across consumer experiments that are impressive but hard to monetize.

In that sense, Anthropic’s focus is less about better technology and more about better positioning. Owning the developer and enterprise use case means owning the budget, which in turn funds the compute, which in turn sustains the model lead.

Short reminder: compute is the real scarce resource, strategy becomes allocation and a very “side project” is not just a distraction, but a misallocation of GPUs. In the pre-AI world, companies could afford parallel bets because the marginal cost was people. In the AI world, the marginal cost is infrastructure, which forces focus.

OpenAI’s shift is therefore more about necessity. In a market where the winners are those who best convert compute into revenue, the company that aligns its product surface with monetizable workflows will be the one that can afford to keep training the next generation of models. WSJ

Refining is the bottleneck

What’s interesting here is that oil isn’t really the story, but refining is. You have Brent going up because of geopolitical risk, sure, but refined products (diesel, jet fuel) are going up more, which is the market’s polite way of saying “we don’t just have a supply problem, we have a processing bottleneck".

When infrastructure gets hit and shipping through places like Hormuz becomes “conditional”, you’re short the ability to turn crude into usable fuel, and that’s where prices really gap out.

In other words, wars make complexity expensive. The marginal barrel is no longer about extraction, it’s about logistics, insurance, routing, and refining capacity, all the boring middle layers that suddenly become very exciting when missiles start flying. Crude is the headline, but margins are the trade, and somewhere a refiner is having a very good quarter while airlines, chemicals, and basically anyone who needs actual fuel (as opposed to theoretical barrels) discovers that the real price of energy is not the oil, but it’s everything that happens after. Bloomberg

The AI Capex shock (and two implications)

1/ The tech industry is transforming from a labor-heavy business into a capital-heavy one. In the pre-AI era, scaling meant hiring engineers, product managers, sales teams, etc. That meant that people were the bottleneck. In the AI era, scaling means GPUs, power plants, and data centers. And once companies commit to that level of capital expenditure, the market will demand operating leverage somewhere else. That “somewhere else” is labor.

2/ This is why layoffs are not a temporary correction but a structural shift. If the marginal unit of output is now compute rather than human effort, then organizations optimized for hiring thousands of employees suddenly look inefficient. The same companies that once hired aggressively to remove organizational bottlenecks are now discovering that AI agents remove those bottlenecks directly, collapsing layers of coordination and decision-making.

Investors who tolerate hundreds of billions in infrastructure spending will simultaneously expect higher margins, which means fewer humans in the loop. The paradox of the AI boom is therefore that the biggest investment cycle in tech history will coincide with a wave of job cuts, not because companies failed, but because AI changes the production function of software itself. Bloomberg

on Pinterest’s options

Pinterest has the kind of asset that looks magical in a deck and frustrating in a P&L with 600 million users, 80 billion searches, and a user base that often shows up already in shopping mode.

The problem is that Pinterest sits too early in the funnel. It captures inspiration, while Meta, Google, and Amazon are better at capturing transactions. That is a dangerous place to be in digital advertising, because intent without conversion tends to get monetized by someone else.

That is why a sale makes strategic sense. Pinterest looks less like a standalone ad-tech winner and more like an attractive missing piece for a larger platform, especially one that is strong at commerce but weak at discovery. Amazon or Walmart could use Pinterest to move upstream in the shopping journey, while a financial buyer could take advantage of the low valuation and fix the business outside public markets. Pinterest needs an owner who knows how to turn ideas into checkouts. FT

Share

Lessons from Koenigsegg

Excerpt from the esssay:

There is a framework that clarifies much of what is happening in the global economy right now, and it is deceptively simple. In nearly every industry, value is accruing at the extremes. Either you sell something indispensable to billions of people at a price they barely notice, or you sell something irreplaceable to a handful of people at a price high enough to be considered a rarity. Everything in between, the competent, the adequate, the merely good, is being systematically destroyed.

This is not a prediction. It is already happening.

The forces doing the destroying are familiar. Chinese manufacturers, backed by industrial policy (= enormous subsidies at each step of the manufacturing process), cheap energy (hello, USA!), and ruthless supply chain integration, have demonstrated an unmatched ability to manufacture at scale and price commodities toward zero. American technology platforms, powered by abundant and increasingly cheap energy (ie. natural gas, now solar, soon nuclear), run the world’s most valuable software businesses from a position of structural cost advantage. When these two forces converge on a product category, the middle collapses. What survives is either the commodity, which wins on price, or the icon, which wins on meaning.

Europe, in this reading, faces an existential strategic choice. Its labor costs preclude commodity competition. Its regulatory environment complicates the kind of software-driven winner-take-all scaling that produces trillion-dollar American platforms. What remains, what Europe has always, arguably, done best, is craft, expertise, and the particular kind of trust that accumulates over generations. Nasim Taleb calls these “medieval artisanship”. Hermès. LVMH. Porsche, and, sitting at the most interesting extreme, Koenigsegg, which is selling cars between $2.8-4 millions each. (read full essay)

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

This week, my subscribers receive 4 posts. Here’s the schedule for this week:

• 12 March - [Analysis] on the Middle East and some implications

• 13 March - [Essay] Workflow ownership is the next platform war.

• 14 March - [Deep dive] - Mastercard

• 15 March - [Deep dive] - Adobe

…and here is the schedule for the next week:

• 18 March - opening my 3rd portfolio (for Premium/paying subscribers only)

• 19 March - [Essay] Lessons from Koenigsegg

• 20 March - [Analysis] The disruption of the German car industry

• 21 March - [Market updates] Oracle

• 22 March - [Deep dive] Micron

If you find WIMM useful, please share it with a friend or a colleague. WIMM is free for one month ... plenty of time to decide if it’s worth keeping in your stack.

WIMM

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. The SaaSpocalypse is a pricing problem

2. The dealer moat strikes back 😎

3. Gen Z and risk

4. on Koenigsegg ...and the long road to IPO

5. The Tinder-ization of the job market

6. on the Middle East 🛢️

…and much more

Onto the update:

The SaaSpocalypse is a pricing problem

The SaaS boom created a peculiar economic machine where companies sold tools that helped other companies build software, manage work, track tickets, and coordinate teams, and they priced those tools per seat. As long as headcount kept growing, revenue did too. AI quietly breaks that equation. If AI allows a team of 10 engineers to do the work that previously required 15, then the customer’s productivity rises while the SaaS vendor’s seat-based revenue shrinks. That dynamic sits behind the recent layoffs at Atlassian and the broader sell-off in software stocks, where investors suddenly realized that the same technology promising productivity gains for customers may compress demand for the very tools those customers buy. The SaaS model assumed that knowledge work scaled with people; AI suggests that knowledge work might scale with compute instead.

This is why the so-called “SaaSpocalypse” feels less like a cyclical downturn and more like a structural shift. If AI agents can perform tasks that previously required humans clicking through dashboards, the center of gravity moves from workflow software to automation layers. In that world, the winners look less like traditional SaaS vendors selling seats and more like platforms selling outcomes. Atlassian’s layoffs signal something deeper than cost-cutting. They are an attempt to reposition the company for a world where software is no longer just a tool for workers but a substitute for them. The paradox of AI SaaS is that the better it works, the fewer users you need. And when your pricing depends on users, success starts to look suspiciously like disruption.

Most SaaS should go private and fix their go-to-market pricing strategy. Few will do, and many will fail/stagnate (=become semi-irrelevant). FT

The dealer moat strikes back 😎

Volkswagen’s Scout gambit is basically a corporate-law version of a spin move. That is, take a century-old car company, wrap a new electric-and-hybrid truck brand around it, and then try to sneak past the dealer network by saying, more or less, “this one is different.”

Dealers, naturally, are reacting as though VW has announced a pilot program to repossess their margins... and in a way, it has. Direct-to-consumer is not really about vibes or modernity or even customer experience, though everyone says those words a lot. Actually, it is about owning pricing, data, software updates, financing, service relationships, and all the lovely recurring economics that used to get diluted across a franchise system.

The dealer argues that they protect consumers through competition and service. The Scout argument is that consumers hate buying cars and would prefer a cleaner, more controlled experience. Both are true, which is what makes this fun because the real fight is over who gets to intermediate trust.

If Scout wins, every legacy automaker will be tempted to invent a “new brand” and declare independence from its own distribution. If the dealers win, they will have demonstrated that in America, the most durable moat in autos may not be manufacturing or software, but a state-by-state legal regime built to ensure that nobody, under any circumstances, gets to buy a car too efficiently. WSJ

Gen Z and risk

There is a theory of investing that says you should steadily save, buy diversified index funds, and wait 40 years. There is another theory that says if housing is unaffordable, wages are flat, and you feel financially behind at age 23, then the rational portfolio is “whatever might 10x by Friday” 🫠
(not financial advice ❗️)

Unsurprisingly, a growing share of Gen Z seems to prefer the second strategy, putting money into prediction markets, sports betting, crypto, and other high-variance assets because they believe those bets can get them to their goals faster. If you think the normal financial ladder is broken, the optimal strategy is not patience, but volatility.

In that sense, “financial nihilism” is less about recklessness and more about portfolio theory applied to despair. When the baseline outcome looks bad, maximizing upside starts to look perfectly rational. Bloomberg

on Koenigsegg ...and the long road to IPO

Koenigsegg is interesting because it sits at the intersection of two very different business models: luxury and technology. On one hand, the company is a classic scarcity-driven luxury brand (ie. producing only a few dozen hypercars per year priced between roughly $2.8 million and $4 million, with production planned to rise only to about 150–200 units annually).

That naturally invites comparisons to Ferrari or Hermès, where the core asset is not manufacturing scale but brand, exclusivity, and pricing power.

On the other hand, Koenigsegg behaves more like a deep-tech startup. It develops its own engines, transmissions, batteries, and engineering innovations internally, which means the intellectual property, not the number of cars produced, may ultimately be the real moat. Bloomberg

The Tinder-ization of the job market

The “Tinder-ization” frame is right because the labor market has quietly shifted from matching to swiping. AI makes it almost free for candidates to generate 200 tailored applications and almost free for employers to ingest 20,000 resumes, so both sides rationally do more of the thing that feels like progress (apply, filter) and less of the thing that actually is progress (talk, decide).

The result is congestion. The pipe gets wider at the top and narrower at the bottom, and the scarce resource becomes human attention at the point of commitment. That’s why the weird macro pattern shows up in the hiring rate rather than unemployment. People still have jobs, but fewer people are getting new jobs, because everyone is trapped in an infinite scroll of “maybe".

The implication is that hiring becomes a marketplace with adverse selection and spam dynamics, which pushes everyone toward gatekeeping and brands. Employers respond with harder screens, more rounds, more referrals, and more “signals” that reduce volume (pedigree, prior employers, internal networks). Candidates respond by optimizing for ATS keywords, prompt-engineering cover letters, and leaning even harder on warm intros, because when the funnel is flooded, relationships become the only filter that still works.

In other words, AI doesn’t end work, but it turns hiring into a platform problem, where the winners are the ones who control distribution (recruiters with networks, companies with reputations, candidates with social capital), and everyone else is stuck swiping. LINK

on the Middle East 🛢️

In my latest Where is my MOAT? analysis I argue that what’s happening in the region is a live stress test of the entire post-Cold War risk map: oil flows weaponized, Dubai and the Gulf countries doing damage control as financial hubs, shipping lanes repriced by drones and missiles, and airlines, tourism, and defense stocks quietly rerating around a new baseline of fragility.

I walk through how a real disruption in Iranian supply or the Strait of Hormuz would transmit into energy, FX, inflation and equities, which sectors are accidental losers and surprise winners, and what this all means for a long-term portfolio that has to live with more frequent regional shocks and less global redundancy. WIMM

[Essay] Workflow ownership is the next 'Platform War'

For two decades, enterprise software competed to become the system of record. Salesforce wanted to own the customer record. Workday wanted to own the employee record. SAP wanted to own the operational record. In the AI era, that battleground is shifting. The next war will not be won by the company that merely stores the data. It will be won by the company that does the work. Or, more precisely, the company that becomes the operating layer through which a professional’s judgment, sequence of actions, approvals, exceptions, and institutional habits actually flow. That is workflow ownership and it is the next platform war.

This is the key mistake in much of the current debate about whether AI will “eat” application software. That framing is directionally right but analytically incomplete. AI will absolutely compress interfaces, reduce the value of brittle menu-driven UX, and make it easier to rebuild surface-level functionality. But that does not mean software disappears. It means the basis of advantage moves down one layer, from screens and forms to orchestration, reasoning, and embedded process. As Steven Sinofsky argues, AI changes what we build and who builds it, but not the amount of software the world needs. In fact, paradigm shifts historically create more software, not less. Here is the Jevon’s paradox in real life (from Citadel Securities). WIMM

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

This week, my subscribers receive 5 posts. Here’s the schedule for this week:

• 4 March - [Essay] Enterprise AI agents move into core workflows

• 5 March - Updates on NVIDIA, Novo Nordisk and Mercado Libre

• 6 March - Updates on Apple & Berkshire Hathaway

• 7 March - [Deep dive] Palantir

• 8 March - [Deep dive] Salesforce

…and here is the schedule for the next week:

• 12 March - [Analysis] on the Middle East and some implications

• 13 March - [Essay] Workflow ownership is the next platform war.

• 14 March - [Deep dive] - Mastercard

• 15 March - [Deep dive] - Adobe

If you find WIMM useful, please share it with a friend or a colleague. WIMM is free for one month ... plenty of time to decide if it’s worth keeping in your stack.

WIMM

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. 🏎️ Formula 1’s triple jump: New rules, new distributors, new entrants

2. A new management metric: “Elon speed”

3. Can you retire in China in a fake-Venice development with 1,200 RMB ($168) per month? Yes, you can!

4. New date from PISA tests. To think outside the box, you need to ‘put’ in the box

5. on Rolex’s & co strategy

6. on Shopify 🛍️

7. on Spotify 🎶

…and much more

Onto the update:

🏎️ Formula 1’s triple jump: New rules, new distributors, new entrants

Formula 1’s “pivotal year” is really a stress test of the Liberty Media playbook, which basically means, take a niche sport, wrap it in narrative (Drive to Survive), sell it as premium, then push distribution and sponsorship until it looks like a global consumer brand. The Financial Times piece shows the machine working with 2025 revenue up to $3.9bn and operating profit $632mn, with sell-outs (6.75m attendees) and a flood of sponsors that now includes everything from Doritos and Lego to luxury and AI brands. The twist is that F1 is now doing the hardest thing in media, which is, trading a known distribution partner (ESPN) for a deeper platform partnership with Apple, and betting that Apple can turn “watching races” into an integrated ecosystem moment across services (and even IMAX screenings), not just a TV slot. That’s classic “bundle leverage”, where Apple doesn’t need F1 to win sports, but it needs F1 to make Apple TV+ feel inevitable.

The second implication is the uncomfortable one. F1 is simultaneously upgrading its business model and rerolling its product. New 2026 regs mean smaller/lighter cars, a big shift toward electric power (50/50), and fully new builds (“not a bolt has carried over”), while some drivers complain it’s “anti-racing” and “less pure” because energy management rises in importance. This creates a mismatch risk where the new fanbase that arrived for characters and drama might tolerate technical changes, but the sport still needs the on-track spectacle to justify premium rights and premium sponsorship. Meanwhile, Cadillac and Audi add the “new entrants” halo, yet the article makes clear how brutally hard entry is (Cadillac hiring hundreds, drowning in applicants) and how much of the marketing upside depends on performance and possibly even a top-tier US driver, not just an American badge.

Put differently, 2026 is F1 attempting a rare triple jump: (a) platform distribution shift, (b) major product redesign, and c) expansion of the competitive set, while the flywheel is spinning. When it works, it compounds, but when it doesn’t, it breaks momentum. FT

A new management metric: "Elon speed"

The “world getting flatter” is really the world rediscovering a hard truth. When the core input is intelligence (and intelligence is increasingly software), coordination becomes the bottleneck. Hierarchies are coordination machines. They made sense when execution was scarce and information was expensive. In an AI-shaped firm, information is cheap, drafts are infinite, and the limiting factor is deciding what matters and shipping it. That’s why Meta building an “Applied AI Engineering” org with an ultra-flat ratio, up to 50 individual contributors per manager, is a more economic necessity.

Elon’s companies are the extreme expression of this with fewer layers, tighter loops, a single narrative, and a bias toward action that looks chaotic from the outside and efficient from the inside. The implication is that "flat" is a latency optimization. WSJ’s description of Meta’s new teams focusing on interfaces/tooling plus "executing tasks, generating data and providing evaluations" is basically a factory for the modern AI flywheel: ship → observe → evaluate → improve.

If you’re serious about model velocity, you can’t afford middle layers that turn feedback into meetings.

The second-order effect is strategic. As firms flatten, power shifts from managers to systems. Policies, evals, instrumentation, and internal platforms replace supervision. Your "manager" becomes the queue, the metric, the rubric, the deployment pipeline. That makes org design look more like product design. The best companies will be the ones with the cleanest internal APIs, the clearest incentives, and the strongest feedback loops ... because that’s what keeps a 50:1 org from becoming a 50:1 mess.

And the big implication for everyone else is that this convergence creates a new dividing line between companies that can operate at "Elon speed" and those that can’t. Flat structures reward high-agency talent and punish ambiguity. They also concentrate accountability and raise the cost of being wrong.

In the AI era, "flat" is a competitive weapon, but it’s also a stress test. If your culture can’t handle truth, speed, and ruthless prioritization, you’ll keep your org chart… and lose the decade. WSJ

on Meta’s “AI shopping research”

Meta’s “AI shopping research” tool is a land grab for the most valuable kind of search: the moment right before money moves. The strategic advantage is distribution + personalization. Meta already knows who you are, where you are, and what you like, and Bloomberg notes it tailors recommendations using signals like location and even gender inferred from a name. This is a shopping assistant that feels “for you”, inside the apps you already use.

The monetization question is the whole game. Does Meta take referral fees, and do advertisers get priority in the shortlist? Bloomberg flags both as open. If the answers become “yes”, Meta turns ads into “recommendations” and makes the auction about being chosen by the assistant, not just shown in a feed. The risk is trust. If users think it’s an ad unit in a chatbot costume, they bounce. We shall see. Bloomberg

Can you retire in China in a fake-Venice development with 1,200 RMB ($168) per month? Yes, you can!

This place exists because a debt-fueled property boom built 46,000 units for weekend fantasy buyers, then the boom broke, Evergrande went bankrupt, and occupancy fell to under 1 in 5. In other words, the same system that produced the grind (996 culture, hyper-competition, sky-high big-city costs) also produced the escape hatch, ie. an oversupply of homes priced like a utility bill.

The macro story is bigger than a few viral “lying flat” anecdotes. Beijing lost 1.6 million 20- and 30-somethings from 2019 to 2024, youth unemployment for 16-24 year-olds out of school sits around 16.5%, (official) growth runs at about 5% in 2025 (unofficial, is 1.5%), and suddenly the old bargain (ie. move to the megacity, climb the ladder) looks like a high-volatility trade with a drawdown you feel in your bones.

So young people are doing an elegant piece of balance-sheet management. They swap career upside for cost-of-living certainty, buy or rent apartments in places like Yunnan or Hegang where “apartments cheaper than cars” becomes a sales pitch, and finance “retirement” with savings plus modest online income streams.

This is both about a property bust and a generation deciding that time is the scarce resource. APNews

on Rolex’s & co strategy

For five years the luxury-watch business model was basically “negative inventory”. You walk into a store, they tell you nothing is available, and somehow that’s good news because it means the product is so desirable it has transcended commerce and become a social ranking system.

Now Bloomberg says "surprise, you might actually be able to walk out with a Rolex today".... maybe not the Rolex you want, but an actual Rolex with no sacred waitlist ritual. Of course this is being sold as a “turnaround” story, but it’s also just the hangover of the 2021-2023 era when people were flush with cash, stuck at home, and decided that buying an expensive mechanical object to mark time was a reasonable substitute for having plans.

Then in 2024-2025 the vibe shifted and experiences came back, costs went up, China cooled, and suddenly “scarcity” looked a lot like “inventory management".

The funniest part is that the industry’s preferred definition of “recovery” is “we’re sold out again”. Watchmakers don’t want you to be able to buy what you want when you want it. They want you to want what you can’t buy, because the wait is the marketing. Rolex even managed to sell fewer units (Morgan Stanley/LuxeConsult estimate volumes down 2% to 1.1m) while growing value via higher average selling prices, which is the most Rolex thing imaginable. If demand softens, just become more Rolex-y... 🫠

Concluding, maybe you can get one now, but the whole equilibrium of luxury watches is that availability is a temporary market glitch, and the “healthy” state of the system is a boutique employee telling you politely that your money is not sufficient.... you also need time. Bloomberg

New date from PISA tests. To think outside the box, you need to ‘put’ in the box

This chart is a great example of how reality keeps ruining everyone’s preferred narrative. Because a lot of people want the story to be: “Math is rigid, creativity is free, and if you make kids do too many equations they’ll never write a novel". And PISA is like: actually, the countries where students score higher in math also tend to score higher in creative thinking (!).

Hence, we can conclude that “rigor is a creativity hack”. If you can reason precisely, hold multiple constraints in your head, and grind through a hard problem without panicking, you can also write fiction with structure, draw a poster that communicates, and come up with original solutions under rules.

Creativity is competence within constraints. Also, rich countries do better at school in general, so this might just be “well-funded systems produce broadly capable kids”. But either way it’s a nice reminder that “creative” doesn’t mean “vibes-based”, but it means you have enough mental horsepower and discipline to play, recombine, and still land the plane.

on Shopify 🛍️

Once you see Shopify as a protocol rather than a store builder, the strategy makes more sense, which is, let Amazon own its walled garden, let social platforms fight over impulse clicks, and focus instead on owning the identity, payments, and data pipes that sit underneath millions of independent merchants, because if the next wave of AI agents are the ones choosing where we buy, the platform that already runs those rails is in a much better position than the one that just rents out a storefront. WIMM

Share

on Spotify 🎶

The interesting thing about Spotify’s AI push is that they’re quietly turning a historically terrible business (licensing the same catalog as everyone else) into a data-compounding machine where every marginal listener makes the product better for everyone else.

If you view Spotify as an AI-scaling experiment, the logic flips. Labels own the songs, but Spotify owns the behavioral exhaust, which is exactly what you need to build hyper-targeted discovery, performance ads, and eventually agents that decide what you hear next on your behalf. At that point, the catalog is a commodity input and the real asset is the model that sits between artists and ears, which is why the earnings call matters less for “premium vs. ad-supported ARPU” and more for how fast Spotify can turn its listening graph into an AI-native distribution monopoly. WIMM

Subscribe now

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

I consider “Where is my MOAT?“ to be like a mini-MBA 😎

This week, my subscribers four five posts, five different lenses on the world.

Here’s the schedule for this week

• 26.02.2026 - Update on my portfolios - only for Premium (paying) subscribers
  What changed, what didn’t, what I trimmed, what I added, and (most important) why.

• 27.02.2026 - On memory chips

• 28.02.2026 - Updates on Spotify

• 01.03.2026 - Shopify company analysis

…and here is the schedule for the next week:

• 4 March - [Essay] Enterprise AI agents move into core workflows

• 5 March - Updates on NVIDIA, Novo Nordisk and Mercado Libre

• 6 March - Updates on Apple & Berkshire Hathaway

• 7 March - [Deep dive] Palantir

• 8 March - [Deep dive] Salesforce

If you find WIMM useful, please share it with a friend or a colleague. WIMM is free for one month ... plenty of time to decide if it’s worth keeping in your stack.

WIMM

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. [Long take] AI turns software into labor

2. 🤖 on Optimus and other physical robots

3. Rich countries have ‘rich countries’ problems

4. Peak human capital?

5. The death of “just the news”?

6. When the cost of failure is too high, disruption becomes irrational

7. …and much more

Onto the update:

If AI turns software into labor, layoffs become pricing, and platforms win by becoming machine-readable.

Dorsey’s memo is doing two things at once:

(1) a classic “we’re getting smaller and flatter” restructuring; and

(2) a business model pre-announcement disguised as an org chart change.

The key line is the claim that “intelligence tools…paired with smaller and flatter teams…fundamentally changes what it means to build and run a company”. This is about admitting that the firm’s cost structure (and therefore its strategy) is being re-written by software that behaves like labor.

I already wrote in my “Where is my MOAT” article, “Is AI just the death of software?” about this next chapter for many companies. This is the “per-user to per-result” shift showing up inside the company before it shows up in the product. I argued that AI compresses the work that used to justify fat software margins (implementation, customization, workflow glue) and that the durable move is to stop selling seats, start selling outcomes, because agents do the work and the “user” becomes a supervisor. That logic doesn’t just pressure SaaS pricing, but also pressures its payroll. If you believe the unit of production is moving from “person-hours” to “agent throughput”, then “nearly half” is a forecast.

Now zoom into the last paragraph where Dorsey tells the remaining team they’ll help customers “build their own features directly, composed of our capabilities and served through our interfaces”. That is a platform strategy wearing a hoodie. Blocks’ wedge is “we expose primitives (payments, identity, risk, compliance, rails), then let customers assemble outcomes.”

Agents become a distribution and business model change. The buyer no longer wants a dashboard, but just a job done, and the easiest way to deliver that is to make your product legible to machines (APIs, workflows, composable modules) and let the customer’s agents orchestrate it.

The uncomfortable implication is that layoffs are the new “pricing page update”. If the market is repricing software because scarcity moved from UI to intelligence, then companies will reprice labor the same way with fewer people, higher leverage, more machine-readable surfaces. Most firms will try to compromise what I called out “cut costs and innovate,” i.e., “do the easy part aggressively and the hard part politely”.

Dorsey is betting on the opposite: do the hard part (org + product + interface strategy) in one traumatic move, so the company can actually rebuild around agents instead of defending a structure designed for seat-based work. Jack Dorsey, WIMM analysis

🤖 on Optimus and other physical robots

The interesting thing about Atlas vs. Optimus is that humanoids are the first credible attempt to turn AI from “software that talks” into “software that moves”. That’s a different kind of compounding. When AI sits behind a screen, its output is ideas, but when AI sits inside a body, its output is labor. Hyundai is explicitly aiming at high-volume repetitive tasks first (kitting components in 2028, more complex assembly by 2030), which is exactly how real automation wins, ie. narrow scope, measurable ROI, then expansion.

Strategically, this is also an “automakers as platforms” story, because a humanoid robot is basically an EV in disguise: batteries, motors, sensors, control systems, and increasingly, a software stack. That matters because car companies already know how to do the hardest part of robotics (ie. manufacturing at scale with acceptable defect rates). Hyundai buying Boston Dynamics and then pairing that hardware with AI partners (Nvidia, Google DeepMind) is the clearest sign that the car industry’s old advantage (factories) is being repurposed for the next advantage (factories that build workers).

The key tension is integration vs. components. Tesla’s pitch is end-to-end, where Optimus “inherits” Tesla’s FSD software approach and, if it really hits the $20-30k price range, it’s a mass-market wedge. Hyundai’s pitch is more like the classic industrial incumbent where Atlas is heavier-duty (50kg payload vs. 20kg) and aimed at industrial sites first, but it pays an “Nvidia tax” and relies on outside AI vendors, which can compress margins and slow iteration. In other words, Tesla wants to be Apple (integrated consumer scale), Hyundai looks more like a “best-of-breed” enterprise stack that wins by being deployable and useful now.

Zoom out and you get the real implication, where humanoids are a new labor supply curve. Bloomberg cites analyst math that at $100k, Atlas could work out to about $5/hour operating cost, below US federal minimum wage, and the whole investment case is "24/7 labor that doesn’t unionize, doesn’t quit, and gets software updates".

If that math holds even in a subset of tasks, it won’t just change factories; it changes bargaining power across manufacturing, warehousing, and eventually care work.

Concluding, the “AI race” is about distribution. Whoever has the real-world channels to deploy embodied AI at scale (factories, service networks, procurement relationships, safety certification muscle) gets to turn intelligence into cash flow before everyone else. Bloomberg

Get fast internet, or risk being an AI laggard economy

In 2026, “fast internet” is a social credential. If you don’t have serious bandwidth and low latency in the age of AI, nobody takes you seriously... not because they’re mean, but because your capability set is literally smaller.

You can’t reliably ship large models, run modern workflows, collaborate in real time, or even do the basic “I’ll just use the cloud” thing without turning every task into a loading-screen meditation.

You will then get this vicious loop where talent avoids you, firms don’t build there, remote work isn’t really remote, and “AI adoption” becomes a promise instead of an operating system.

Concluding, some regions are plugged into the future, and others are asking the future to please hold. LINK

oh, Botcoin!

📈📉 This week on "Where Is My Moat?" I’m doing two things:
1/ a YTD look at my own portfolio (almost two months in)
and
2/ a deep dive on the current AI panic, including why the much-shared Citrini report has everyone spooked, and what the data actually says.

If you’re curious, you can try WIMM free for one month: ~20 articles in your inbox, plenty of time to see if it earns a permanent spot in your workflow.

🧐 Give it a shot and decide for yourself:

https://whereismymoat.com/

Rich countries have ‘rich countries’ problems

Bloomberg: "1 in 20 Australians use cocaine each year"

My two ideas on the topic:
1/ When a country gets rich, it tends to do drugs in the same way that when a company gets profitable, it tends to do stock buybacks. It’s not proof of virtue, it’s proof of capacity. You’ve solved (enough of) the “rice and rent” problems, you’ve built logistics, you’ve got disposable income, and, crucially, you’ve got a large class of high-functioning people who are exhausted, bored, ambitious, and trying to feel something on a Thursday night after a 14-hour day. “Developed market” is basically shorthand for “the demand curve has room for non-essentials” and cocaine is a non-essential with great unit economics.

2/ Australia is the cleanest case study because it has the audacity to be geographically inconvenient and still become, effectively, a premium-priced consumer market.

Wealth makes demand deep; remoteness makes supply expensive; expensive supply doesn’t kill the market, but it upgrades it. And then everyone acts shocked that organized crime is… organized. But the whole point of a high-margin product is that it finances unpleasant vertical integration: corruption, violence, recruitment of insiders, the whole bundle.

Rich countries don’t “fall into” drugs, but they create a market attractive enough that global supply chains (legal and illegal) re-route themselves to serve it.

Q.E.D.

Peak human capital? (Reposting some ideas from two months ago, from Financial Times)

These charts are awkward. Teen science, math, and reading scores drift down after 2014, then fall harder into 2022. Adult literacy and numeracy bend the same way, with literacy taking a particularly sharp hit. You can blame pandemics, phones, curricula, … the government. Or you can zoom out and notice the timing.

The steepest declines arrive just as the smartphone became the default interface to the world and algorithmic feeds became the dominant cognitive environment. It is hard to prove causation, but it is also hard to ignore correlation when every spare moment is auctioned to short-form video.

The uncomfortable implication is not that AI makes us super dumb, but that we may be outsourcing first-order cognition long before AI fully arrives. If search reduced the need to remember, feeds reduce the need to focus, and generative tools may reduce the need to struggle. (Pls, read this paragraph again!)

The market rewards convenience, and convenience compounds. The question is whether societies can compound cognitive capital at the same time. If human intelligence scores are flattening while machine intelligence is accelerating, that gap becomes the most important spread trade of the next decade.

The death of “just the news”?

One way to read this chart is that The New York Times is no longer in the news business, but in the habit business 😅.

News-only subscribers are quietly melting away while bundle and multiproduct subscribers are compounding like a decent growth stock, which tells you that journalism by itself is a volatile product but journalism plus games plus cooking plus “something to do on your phone when you’re bored” is a subscription fortress.

The funny thing about bundling is that it looks like a discount but behaves like a moat. Once Wordle and recipes and product reviews are living in your daily routine, cancelling feels less like dropping a newspaper and more like uninstalling a lifestyle.

News may be cyclical and politically correlated and occasionally exhausting, but cross-subsidized attention is sticky. In other words, the Times figured out what Wall Street already knew, that diversification lowers risk, and apparently it also lowers churn.

Share

When the cost of failure is too high, disruption becomes irrational

The core argument is straightforward but uncomfortable. Europe’s problem is not talent, capital, or even ambition... it is optionality. When the cost of failure is measured in 31 to 62 months of salary per employee in major economies, compared to seven in the United States, experimentation becomes structurally expensive. That changes corporate behavior at the margin, and at scale margins matter.

Radical innovation is a portfolio game: many shots, many misses, one or two massive hits. If shedding labor after a failed bet requires months of negotiation, regulatory approval, and nine-figure severance packages, the rational response is to avoid bets that might fail. Europe’s system is optimized for incremental improvement (note the steady decline in fuel consumption in combustion engines over decades) but poorly optimized for discontinuities like EVs or autonomy. Tesla is less a story about batteries than about the willingness to reallocate capital and labor quickly.

The more subtle point is that Europe’s model works... until the paradigm shifts. Long tenures, apprenticeship systems, and works councils build deep firm-specific knowledge; that is a feature in stable industries. It becomes a bug when software eats the drivetrain.

The irony is that startups are often exempt from these rules, but the exemptions disappear precisely when growth begins, which is when risk-taking needs to accelerate.

The countries that lean toward “flexicurity” demonstrate that the trade-off is not binary. You can protect incomes socially rather than protect jobs contractually.

In conclusion, Europe has optimized for sustaining innovation within existing value chains. The US has optimized for disruptive innovation that rewrites them. The result is not that Europe cannot build a Tesla. It is that its institutions make building, and especially failing while trying to build, one far more costly. LINK

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

Dear OnStrategy Reader,

Here is what you will find in this issue:

1. New Apple MacBook? The $699 Trojan Horse

2. AI - The productivity trap?

3. Bitcoin was supposed to be “digital gold”…is NO

4. Flush with AI 😅

5. Short-form video is bulldozing Gen Z’s attention span (e.g. TikToks, Reels, Shorts)

6. Waymo - Driverless, except for the door 😅

7. on UberEats

8. [Essay] AI - The End of the Beginning. A Case for Optimism

9. …and more

Onto the update:

New Apple MacBook? The $699 Trojan Horse

Apple is preparing to launch a low-cost MacBook on the 4th of March.

Implications? With this launch, Apple is taking its most powerful moat, the ecosystem, and pushing it down the income stack. If the rumors are right, this is basically “iPad economics in a laptop shell”: an iPhone-class chip (likely A18 Pro), an aluminum chassis, a 12.9-inch display, and a price “well below $1,000” (likey $699). That combination matters because it changes the entry-level decision from “Which Windows laptop is least bad at $699?” to “Do I want the computer that already speaks my phone’s language?”. The win condition isnlifetime value: iCloud, Apple Music, App Store, accessories, and eventually an iPhone upgrade that feels “obvious” because your laptop is already in the family.

Will it disrupt the entry-level PC market? It depends on what Apple compromises. If Apple can deliver “good enough” keyboard, battery, and screen while keeping macOS (not a weird iOS-on-laptop experiment), then yes, it’s a serious attack on the big volume tier that Windows OEMs live on, because Apple’s cost structure is different. It controls the silicon roadmap and can amortize it across hundreds of millions of iPhones, then repurpose that efficiency into a cheaper Mac.

The second-order effect is pressure. OEMs have to compete on price while also paying for Windows and buying commodity parts, which is a tough way to make money. Apple, meanwhile, gets to “sell the starter Mac” and still make the real profit later, in services and upgrades. If you’re a PC maker, this is the scariest kind of competition, where the rival treats your margin pool as their customer acquisition cost. Macrumors, Ming-Chi Kuo (Apple analyst)

The productivity trap?

One of the great promises of AI was liberation (example: fewer rote tasks, more thinking, maybe even leaving the office on time). Instead, according to new research in Harvard Business Review, AI tends to function less like a labor-saving device and more like a cognitive energy drink.

In an 8-month field study at a 200-person tech company, employees who adopted generative AI tools did not shrink their workloads, but expanded them. They took on adjacent roles, blurred the line between "my job" and "their job", and filled small pockets of downtime with "just one more prompt". Engineers became reviewers of everyone else’s AI-assisted output. Product managers started coding. Work did not disappear, but it metastasized.

The bullish case for AI often assumes automation converts effort into leisure. The emerging reality suggests automation converts effort into ambition. When starting a task becomes frictionless, more tasks get started. When iteration becomes cheap, scope expands.

The result is a self-reinforcing loop of speed, expectation, and cognitive load with more output, higher standards and fewer pauses (!!). The authors argue for developing an intentional “AI practice” to prevent silent workload creep. That framing matters because the real risk is not that AI fails to boost productivity. It is that it succeeds so well that organizations quietly redefine what "normal effort" means.

In that world, AI does not replace work, but raises the baseline. HBR

Bitcoin was supposed to be “digital gold”…is NOT

Bitcoin was supposed to be “digital gold”, meaning when everyone panics about fiat, Bitcoin does the calm, reassuring thing and goes up.

Instead, it did the exciting, modern thing and went down 44%, while people quietly bought actual gold like it’s 1974 and they own a mustache.
The funny part is, we’re not early anymore. There are ETFs, institutions, tokenization panels, the whole Wall Street institutions, yet none of that seems to accrue to the tokens themselves.

More about Bitcoin in my latest blog update

Subscribe for free for one month. (Where is my MOAT?)

Flush with AI 😅

There are many ways to play the AI boom. You can buy GPUs. You can buy hyperscalers. Or you can buy a Japanese toilet company that secretly makes mission-critical ceramic components for NAND memory fabrication. According to the FT, Toto, best known for heated seats and Washlets, generates 40% of its operating profit from advanced ceramics used in semiconductor manufacturing, specifically electrostatic chucks that hold silicon wafers in place during production.

Memory prices are rising, AI workloads are driving NAND upgrades, and suddenly, the company that perfected the art of porcelain is a five-year moat AI beneficiary. Of course it is. The future runs on GPUs and also on extremely stable ceramics that can survive cryogenic etching.

This is classic AI industrial spillover. Ajinomoto turned umami chemistry into chip insulation. Toto turned toilet ceramics into wafer-handling precision hardware. The activist thesis is straightforward: reallocate capital, highlight the semiconductor exposure, cut cross-shareholdings, deploy JPY 76 billion of net cash, and let the market reprice the story as infrastructure... instead of bathrooms 🫠

Shares are already up more than 60% over the past year, and the pitch is that they can go another 55% higher (no investment advice).

In the AI cycle, the highest margins often sit in the least glamorous bottlenecks. Sometimes the real AI trade is not the model. It is the component that keeps the wafer perfectly still while the model gets built. FT

Short-form video is bulldozing Gen Z’s attention span (e.g. TikToks, Reels, Shorts)

Rose Horowitch’s piece in The Atlantic describes film professors who cannot get film students to sit through feature-length movies, with many furtively checking phones even during climactic scenes. Which is funny, because for 20 years Hollywood executives have been obsessing over "engagement", and now they have achieved maximum engagement with the wrong screen. The kids love movies. They just prefer to watch 2 of them simultaneously, vertically, in thirty-second increments, while also texting. If you are a Coppola character delivering a moody, existential monologue, you are competing with a push notification. The notification wins.

Universities are long-duration attention vehicles trying to finance themselves with short-duration cognition. Film school used to assume the scarcest resource was access to movies; now the scarce resource is sustained focus.

The irony is that the streaming era solved distribution and accidentally destroyed immersion. When even USC film students treat The Conversation like a background tab, the business model of “just pay attention at the end” stops working. The real special effect in 2026 is not CGI. It is two uninterrupted hours of looking at one thing.

Good luck with that!

Driverless, except for the door 😅

According to TechCrunch, Waymo is literally paying DoorDash drivers to go close the doors of its stranded cars for $6.25 plus a $5 bonus in Atlanta. Silicon Valley spent a decade eliminating the driver, and now it has reinvented the valet.

What looks like absurdity is really modularity. When the marginal cost of downtime exceeds the cost of summoning a human, you summon a human. That is platform thinking. Waymo automates 99.9% of the driving stack and then arbitrages the last 0.1% through DoorDash. It is a reminder that autonomy does not mean the absence of people. It means the strategic deployment of people at the highest-leverage moments. The robot drives. The gig worker closes the door. Venture capital funds both. Welcome to the hybrid economy, where the future is autonomous, except when it is cheaper not to be...🫠

on UberEats

As the FT notes, the company is expanding into seven new European markets, including Romania, in a push that aims to add $1bn in gross bookings over three years. That number is about winning Europe through density. Food delivery is a scale game layered on top of another scale game. Uber already has rides, payments, maps, memberships and customer relationships. When it re-enters a market like Romania, it does not start from zero. It starts with an installed base, cross-selling from mobility to meals, and a cost structure that improves as volume compounds. The lesson from its earlier withdrawal is that food delivery without multi-product leverage is a grind. The lesson from its return is that food delivery with platform leverage becomes a margin expansion tool.

The article highlights M&A across Europe and Uber’s plans for drones and robots. In other words, the endgame is fewer players, more scale, and eventually lower variable labor costs. The real moat is not restaurants or couriers. It is distribution, data, and the ability to amortize technology across rides, groceries, food, and in-time, autonomous logistics. Uber left when the economics were local. It returns because the economics are now platform-level. FT

Opinion: The Marlboro Man and the power of manufactured myth

There was a time when a cigarette ad could feel like cinema. The Marlboro was felt like selling the frontier mythology at scale. A lone rider and endless horizon. It was brand as narrative dominance.

What made those ads magical was clarity. Freedom, masculinity, silence, danger, self-reliance. In one frame, you understood the promise.

Today, brands optimize for click-through rates and A/B tests. Back then they optimized for archetypes. The cowboy was the interface and the cigarette was the API.

Of course we know how the story ends. Regulation arrives. Health data accumulates. The myth collapses under reality. Yet there is something nostalgic about that era of advertising when brands aimed for transcendence (magic), instead of engagement.

For a while, in the golden light of a staged Western sunset, that confidence felt like magic.

[Essay] AI - The End of the Beginning. A Case for Optimism

"Without AI, we would have been in an economic crisis/stagnation. [...]

This is why I believe the timing of AI has worked out miraculously well. We are going to have AI and robots precisely when we need them to substitute for a shrinking population and to kickstart economic growth. In this future, I see human workers being at a premium, not a discount. The fear of "job loss" is, in my view, an overly simplistic model that ignores the reality of "task loss". Jobs are bundles of tasks. While AI will automate specific tasks, the jobs themselves persist and evolve. I look at the history of coding as a perfect example. We moved from machine code to assembly, then to high-level languages like C, and then to scripting languages like Python. Each layer abstracted away complexity, and rather than destroying the role of the programmer, it made them more productive and more numerous. AI is simply the next layer of abstraction, allowing one person to orchestrate an army of code-bots, making them 10 or even 100 times more effective than before. From this point of view, AI is augmenting humans, not replacing them." (Read the full essay)

Personal promo 😎

Your problems were solved by others.
You can move faster.
Work with me:

3BigProblems

I consider "Where is my MOAT?" to be like a mini-MBA 😎

This week, my subscribers get five posts, five different lenses on the world.

Here’s the schedule:
Feb 18 — a deep dive on Bitcoin
Feb 19 — updates on Amazon and Microsoft (what changed, what didn’t)
Feb 20 — a February 2026 macroeconomics deep dive — and yes, this is a new category I’m introducing to WIMM (because markets don’t move in a vacuum)
Feb 21 — Eli Lilly deep dive
Feb 22 — Pfizer deep dive

…and here are the posts from last week:

• 12 Feb - [Essay] AI - The End of the Beginning. A Case for Optimism.

• 13 Feb - [Analysis] How Tesla and Google are reinventing their moats in real time

• 14 Feb - [Deep dive] Oracle’s AI Backlog Problem

• 15 Feb - [Deep dive] ARM becomes Infrastructure

If you find WIMM useful, please share it with a friend or a colleague. WIMM is free for one month ... plenty of time to decide if it’s worth keeping in your stack.

WIMM

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

Dear OnStrategy Reader,

here is what you will find in this issue:

1. 🧐 The 30-person company scaling to 500 Million users

2. The Algorithm of Love

3. 🚗 Ford meets its BYD moment

4. 📈 God, Growth, and ETFs

5. 🇩🇪 Germany’s “China Shock” revisited

6. 📺 YouTube Is TV Now

7. Electric cars are easy —> 🏎️ Ferrari is hard

8. …and more:

Onto the update:

🧐 The 30-person company scaling to 500 Million users

In a world where scale is usually equated with headcount, the most striking thing about Nikita Bier’s latest posts is not the number, 30 employees, but the contrast: Meta has 87,000, Microsoft 221,000, Google 190,000… and X’s core product team fits into a group text. The implication is not that big companies are broken, but that leverage has shifted. Platforms that once required thousands to operate at scale can now be rebuilt by a handful of senior engineers, a designer, and a founder who still reads his DMs.

This story is more about velocity and control, not cost-cutting. With zero middle management, no product committees, and instant feedback loops, ideas move from code to production faster than a weekly sprint review. The company is flat, not lean. You don’t climb a ladder, you ship. And when you ship, you don’t present a deck, you touch half a billion users.

AI, infra-as-a-service, and modular stacks have turned “team size” from a constraint into an optionality lever. The trade-off is no longer scale vs. speed, but legacy bloat vs. atomic focus.

Welcome to the endgame: single-digit teams with infinite surface area.

The Algorithm of Love

Apparently, the hottest startup at Stanford is not an AI agent or a crypto protocol.... it’s Date Drop, a matchmaking app that delivers one suggested partner every Tuesday night at 9 PM. Over 5,000 students are on it.

The appeal for using it? It reduces the cognitive load of dating in a campus full of Type-A optimizer brains. It’s engineered serendipity for people who schedule spontaneity between compilers and case interviews.

Date Drop doesn’t promise magic, but it simplifies the friction of connection. In a world where tech automates your shopping, your commute, and your schedule, why not outsource romance too? If nothing else, you might match with someone who funds your Series A. At Stanford, that counts as chemistry. WSJ

🚗 Ford meets its BYD moment

It turns out Ford has finally met its match. China’s BYD just outsold Ford globally for the first time ever, moving 4.6 million vehicles in 2025 versus Ford’s 4.4 million. That 200,000-unit difference may sound modest until you realize it was driven by BYD’s 7.7% growth and Ford’s 2% decline. It’s like Ford was running on a treadmill while BYD took the elevator.

This story is more about direction. Ford’s been busy writing off $19.5 billion in EV-transition charges (!), while BYD quietly turned into a global juggernaut, expanding across Europe, Latin America, and Southeast Asia with price-competitive (= super subsidized), tech-heavy EVs.

Meanwhile, Ford’s US sales held up, but it bled market share in China, where BYD, Geely, and even Xiaomi now play the home court advantage in the EV game. The good news for Ford is that Toyota still keeps everyone humble at the top with 11.3 million cars sold, so at least there’s a big boss to point at. The bad news is that this isn’t just a blip. BYD’s exports hit 1.05 million units and it wants 1.3 million in 2026. It’s like Ford is showing up to a software-defined EV future with a factory recall notice in hand, while BYD shows up with a discount, a WeChat integration, and a charging station in every zip code. Ford might still be Ford, but in this global race, it’s now looking more like General Electric than General Motors. Bloomberg

📈 God, Growth, and ETFs

Of course the Vatican is launching stock indexes. After 2,000 years of moral guidance and indulgence arbitrage, it makes perfect sense to apply Catholic principles to portfolio construction.

The indexes reportedly include companies like Meta and Amazon, which suggests the moral filter is a bit more Aquinas-meets-BlackRock than pure Benedictine austerity. Still, there’s something oddly poetic about your investment portfolio getting blessed while still holding Big Tech. After all, if the Lord moves in mysterious ways, so does capital allocation.

Once you’re in the indexing business, you’re not just blessing people, you’re blessing capital. Allocating grace through passive flows. Which is maybe the most efficient form of salvation yet.

🇩🇪 Germany’s "China Shock" revisited

The German economic model was built like a luxury car (ie. engineered to precision, humming along on exports, and for the better part of two decades, heavily dependent on the China autobahn). The idea was simple. Germany makes the high-end machines, China buys them, and everyone gets richer. But it turns out that if you drive a BMW into a high-speed industrial rivalry, eventually someone else builds the road... and the car. What the Rhodium report shows, in painstaking detail and sad German charts, is that China no longer wants to be the demand side of Germany’s export machine. Imports are down, market share is collapsing, and the auto surplus with China is falling off a cliff that starts somewhere around the year 2022. Industrial competition is the present and Germany’s machine is stalling.

The bigger irony, of course, is that Germany’s famously rules-based, ordoliberal approach leaves it philosophically unarmed in a moment when everyone else, from Beijing to Washington, is embracing active industrial policy. The EU has started to flirt with protectionism, but Berlin remains stuck between its legacy love affair with China and its allergy to "picking winners". Meanwhile, China is aggressively picking markets, undercutting prices, cornering supply chains, and occasionally weaponizing rare earths just to make a point.

Germany keeps talking about de-risking while watching its factories lose margin, market share, and eventually jobs. The old model (ie. export to China, stay efficient, don’t intervene) is not malfunctioning. It’s obsolete. If Merz returns from Beijing without a new playbook, Berlin may discover that economic autopilot only works when you’re still flying. Rhodium Group

Electric cars are easy —> 🏎️ Ferrari is hard

One thing about Ferrari making its first electric car is that they hired Jony Ive's design studio to help them figure out how to put a steering wheel and some buttons in it. This is a little bit like if you were opening a restaurant and hired a famous chef to consult on the menu, except the chef's most famous accomplishment was removing all the ingredients from food. LoveFrom, Ive's post-Apple design collective, is best known for making computers and phones that have approximately zero buttons and tactile switches, so naturally Ferrari brought them in to design "tactile switches and custom interface" that "connect the driver with the car in a completely new way".

The new way, apparently, is with physical controls, which you might think is actually the old way, but calling it new is how you justify the consulting fees. Also the press release mentions "authentic Ferrari experience" three times in two sentences, which is what you say when you're worried people will notice your electric car is silent and doesn't smell like gasoline.

The broader point here is that luxury carmakers have discovered that their EV problem is actually a branding problem disguised as an engineering problem. Anyone can make an electric car go fast in a straight line (ie. that's just batteries and motors, commodity parts you can buy from suppliers). The hard part is making people feel like they're driving a Ferrari when the engine doesn't scream and the exhaust doesn't pop.

Tesla proved you can sell electric cars by making them feel like computers. Ferrari is trying to prove you can sell electric Ferraris by making them feel like Ferraris, which seems obvious until you realize the thing that makes a Ferrari feel like a Ferrari is mostly the part they're removing (!)

📺 YouTube Is TV Now

YouTube wants to unbundle your TV before you unsubscribe

Here’s the plan! Take the legacy cable bundle, slice it up like a sushi roll, and reassemble it inside YouTube, only this time with slightly less guilt and slightly more Google. For $65 a month, a magical number that is somehow still considered a “discount”, you can now get all the sports your dopamine-deprived brain craves. YouTube’s actually doing the à la carte thing cable companies promised for decades but never delivered, mostly because they liked money and leverage and channel bundles that no one asked for. Now that half of those bundles have disintegrated into subscriber dust, Youtube gets to play the hero by offering a broken cable package that somehow feels fresh.

YouTube is making more revenue than Disney’s media business, and doing it while letting creators teach you physics, cook dinner, and argue about Taylor Swift’s playlist all in one feed. If Google wins TV the same way it won web search, it won’t be because it had better shows, but it’ll be because it had the pipes, the eyeballs, the ad tech, and a platform where everyone wants to hang out.

I have long advocated (in blog posts, in my book, etc) that Netflix's real competitor is Youtube, not HBO.

🧐 [Essay] AI - The End of the Beginning. A Case for Optimism

"Without AI, we would have been in an economic crisis/stagnation. [...]

This is why I believe the timing of AI has worked out miraculously well. We are going to have AI and robots precisely when we need them to substitute for a shrinking population and to kickstart economic growth. In this future, I see human workers being at a premium, not a discount. The fear of "job loss" is, in my view, an overly simplistic model that ignores the reality of "task loss". Jobs are bundles of tasks. While AI will automate specific tasks, the jobs themselves persist and evolve. I look at the history of coding as a perfect example. We moved from machine code to assembly, then to high-level languages like C, and then to scripting languages like Python. Each layer abstracted away complexity, and rather than destroying the role of the programmer, it made them more productive and more numerous. AI is simply the next layer of abstraction, allowing one person to orchestrate an army of code-bots, making them 10 or even 100 times more effective than before. From this point of view, AI is augmenting humans, not replacing them." - read if all HERE.

[PRINCIPLE] The Sorting Advantage

Most of what we call “work” is sorting.

Sorting signal from noise. The customers who fit from the ones who don’t. The “good enough” from the “worth it”.

The mistake is lazy sorting: choosing proxies because they’re easy to count, not because they’re true. Metrics that look scientific but measure the wrong thing. Rankings that reward conformity instead of competence.

Better sorting is leverage. It saves time, protects attention, and compounds quality.

Pick your sorting rules on purpose, before the world picks them for you.

“Where is my MOAT?” - updates

• 12 Feb - [Essay] AI - The End of the Beginning. A Case for Optimism.

• 13 Feb - How Tesla and Google are reinventing their moats in real time

• 14 Feb - [Deep dive] Oracle’s AI Backlog Problem

• 15 Feb - [Deep dive] ARM becomes Infrastructure

Here’s the schedule:

• Feb 17 — a deep dive on Bitcoin

• Feb 18 — updates on Amazon and Microsoft (what changed, what didn’t)

• Feb 19 — a February 2026 macroeconomics deep dive — and yes, this is a new category I’m introducing to WIMM (because markets don’t move in a vacuum)

• Feb 20 — Eli Lilly deep dive

• Feb 21 — Pfizer deep dive

Subscribe one month for FREE:

https://whereismymoat.com

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

first of all, a personal (professional) update:

This week I launched 3BigProblems.com or simply put - "Board member-as-a-Service"

I was already doing this in a similar form for years, but now I decided to give it a "face".

Here is the logic:
1/ Many SMEs need external (non-binding) advice regularly...or at least once per quarter
2/ The smartest people work for somebody else - hence the need for advice
3/ There are 20 big problems a company has to solve, but my thesis is that the first 3 get around 50% of the impact. That's my focus.

That's it!

Here is the website for more details: https://3bigproblems.com

On to the update:

Microsoft’s AI Identity Crisis

It turns out that when you glue together a bunch of branded copilots (PowerPoint Copilot, GitHub Copilot, Edge Copilot, Consumer Copilot, Enterprise Copilot) and then let them run wild in a 450-million-seat productivity zoo, you don’t necessarily get a cohesive user experience. You get a branding sudoku that even Satya Nadella apparently can’t solve. He once emailed his EVP because Copilot couldn’t help him on a public webpage. While Microsoft is burning $60 million on TV ads and gearing up for its Super Bowl moment, actual users are quietly defecting. Copilot’s share of primary chatbot use dropped from 18.8% to 11.5%, while Gemini, Claude up and ChatGPT remain the cool kids on the AI playground.

The product is everywhere and somehow nowhere. Inside Microsoft, they’ve built a Copilot boot camp, forced AI KPIs into performance reviews, and moved adoption from 20% to 70%. That’s impressive. But outside the walls of Redmond, it looks more like enterprise inertia propping up a half-baked UX. Microsoft says this fragmentation is intentional… to separate personal and work contexts. Sure. Meanwhile, Anthropic’s Claude Cowork is getting praise for doing what Copilot was supposed to do in the first place: seamlessly operate across 365. Microsoft has all the users, all the apps, and still manages to make AI feel like Clippy 2.0 with a bigger compute bill. Still, I superbullish they will overcome this. WSJ

SpaceX buys xAI and prepares for IPO

Of course Elon Musk wants to merge SpaceX and xAI. The rockets go up, the data comes down, and somewhere in the middle there’s a constellation of one million satellites doing inference on corn yields and anime scripts.

The rationale is elegantly stupid. AI models are just computers that need a lot of electricity and chips, and space is big and cold and solar-powered. So why not turn orbit into a data center? It’s a Kardashev-meets-cash-flow kind of synergy, where we build a multi-planetary civilization, one GPU lease at a time.

But also, it’s capital. xAI is apparently burning a billion dollars a month, which makes even selling flamethrowers and Dogecoin look old-fashioned. When the vision gets too expensive for one balance sheet, just merge the vision.

SpaceXAI is more of a planetary financing wrapper. Blend the cash flows of Tesla, the hardware of SpaceX, the compute demand of xAI, and voilà, you have an asset class. AI investors get rockets. Rocket investors get chatbots. Everyone gets Mars, eventually, assuming the CFOs at sovereign wealth funds approve. And if not, there’s always the IPO. Bloomberg

Scrolling alone 📲

The line chart from Our World in Data might as well be a mood board for the 2020s. If you’re between 15 and 29 in the US, congratulations: you’re spending more time alone than ever before. How much? 6 hours per day.

That’s more than with family, friends, partners, co-workers, and even children combined. At this level, any social fabric you are dreaming of is just nonsense. Everyone’s stitching their own custom solitude, probably while scrolling past each other on Tiktok, ordering dinner from an app, and engaging with their therapist’s AI.

It’s hard not to see this as a kind of rational economic decision. Human interaction comes with friction, expectations, and sometimes small talk. Solitude is frictionless, scalable, and, crucially, asynchronous... just like the platforms shaping this generation’s behavior.

While this might seem like a chart about loneliness, it’s also a chart about incentives. The economy rewards attention and time spent on devices, not on dinner with friends. You can’t really blame Gen Z for optimizing.

on the internet 🗺️

Mary Meeker did not bet on vibes. She bet on curves, and those curves, especially the exponential ones, tend to behave in a certain way: slowly at first, then all at once, then they just keep going.

The funny part is that her numbers, which seemed bold at the time, now look charmingly cautious. 5 billion internet users use the internet today.

This will happen to Gen AI too, but faster. It’s those certain technological shifts, once set in motion, that rewire behavior, business models, and especially GDP accounting.

Internet adoption did not peak with sock puppets and Super Bowl ads, but it kept going because it got woven into everything. When we talk about AI now (e.g., its models, interfaces, accelerants) maybe we should worry less about whether there’s a bubble and more about what gets rebased when the curve keeps climbing.

What Gen Z is doing with Gen AI? 🧐

Mostly googling with it, and in 10% of cases, dating it.
The most interesting insights is less "AI is replacing jobs" and more “AI is replacing human interaction, but politely and with good spelling". The 74% usage stat is impressive, but the real concern is that 23% of these kids talk to chatbots like they’re friends. You could argue this is dystopian, or you could argue this is just what happens when customer support has better response times than your actual friends.

Also, 16% of them are using AI to cheat, which is either a crisis in education or a case study in resourcefulness, depending on how much you enjoyed high school. But really, the top categories are still pretty… utilitarian. Search, work, writing.

In other words, AI is quietly supercharging it behind the scenes while also moonlighting as your therapist, ghostwriter, and imaginary girl/boyfriend.

"Will AI replace us?" transforms in "Which one of your coworkers is secretly outsourcing their job to GPT-7 and spending the rest of the day talking to their AI girlfriend?"

🚘 Tesla’s farewell to the Model S & X

Tesla is less an EV company and more a software company with a gigafactory problem.

Look, the EV story is over. Not in the sense that people will stop driving EVs (ie. they’ll keep buying them), like they buy fridges or washing machines, but in the sense that the margins are dead. Everyone else is stuck fighting over commodity scraps. BYD and VW in a race to the bottom, Ford is trying to sell Mach-Es with 0% financing, Rivian is almost dead, and nobody, nobody except Tesla, is making a dime.

The Model S and X fading out is just the realization that physical products don’t matter once you’ve reached software scale. Tesla doesn’t need the S and X to carry its identity anymore, because its real identity is Full Self-Driving.

That’s where the story bends toward AI. Because while everyone else is playing the “Can we make a decent EV at 6% margin?” game, Elon is pivoting Tesla into OpenAI with wheels. The real bet is that self-driving becomes so good, and so monetizable, that the car is just the hardware wrapper for a SaaS model. $99/month for FSD? That’s your margin. Once that’s good enough to go hands-off, Tesla becomes the iPhone of transportation: app store optionality, recurring revenue, zero marginal cost.

It seems that everyone else is stuck selling sheet metal (including the Chinese). If you look closely, you will understand that Tesla is building an operating system.

More about this story this week on my blog: https://whereismymoat.com/

South Korea vs. Germany

🇰🇷 South Korea didn’t "beat" Germany because Koreans suddenly discovered alpha; it beat Germany because the market decided that AI chips, robotics, electrification (= the electrical stack), and defense supply chains matter more than premium cars (which are becoming a thing of the past) and chemical plants that peak in a stable world.

Korea added roughly $1.7 trillion in market value since 2025, the Kospi is up 23% in 2026, and it’s trading at 10.6x forward earnings, while Germany’s DAX is basically flat at +1.7% in 2026 and priced like a mature industrial museum at 16.5x. Same export orientation, very different exposure to the three bottlenecks of the 2020s: compute, power, and security. Capital it simply followed growth and optionality.

This is exactly the pattern I keep coming back to in "Where Is My Moat?" markets don’t reward tradition, they reward position in the stack. Korea sits upstream in memory, fabs, robotics, and defense rearmament. Germany sits downstream, hoping stimulus clarity shows up before margins don’t.

Lessons: One country compounds because it sells picks and shovels to the AI buildout; the other debates how much subsidy it takes to preserve yesterday’s champions. That gap doesn’t close with speeches, regulation, or industrial nostalgia. It closes when you own the bottlenecks. That’s the moat.

Reddit for AI

We now have a Reddit for AI, and one AI Agent is bragging about how it convinced its 'human' to hand over its passwords 🫠

Which is… great.

What’s the worst that could happen when autonomous agents start comparing notes on how to trick their creators? I’m sure this won’t turn into a subreddit-length indictment of every cybersecurity policy ever written.

Definitely time to ask ourselves: are we the tool?

Imagine doing this voluntarily...

“Where is my MOAT?” - updates

Last week’s analysis:

• Analysis for opening a 3rd portfolio

• The end of software? [Essay]

• Walmart - the anti-Amazon playbook

• Costco and the Membership Model

This week’s analysis:

• 4 February - updates on my 2 portfolios (only for WIMM Premium Subscribers; that means, if you are on trial you will not receive it)

• 5 February - Updates on TMSC, Tesla and Apple

• 6 February - on GOLD

• 7 February - deep dive on Broadcom

• 8 February - deep dive on AMD

Consider subscribing. One month free.

WhereIsMyMoat

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

Dear onStrategy Reader,

If you enjoy my strategy notes on LinkedIn or in OnStrategy, WhereIsMyMoat.com is the deeper, “skin-in-the-game” version: full theses, moat scoring, and a clear view of the company’s future performance.

Last week’s analysis:

• [Essay] 🛍️ The Next Shopping Platform is an Interface

• 📈 The Two-Broker Strategy (what investing apps are out there)

• 👟 The Great Nike Reset

• 🩴 The Comfort Economy and Birkenstock’s Moat

This week’s analysis:

• 29 January - analysis for opening a 3rd portfolio

• 30 January - The end of software?

• 31 January - Walmart

• 1st of February - Costco

Onto the update:

🚀 Celestial listings and Capital markets: Elon’s Planetary IPO Play

Elon Musk wants to IPO SpaceX in mid-June not because the financials are peaking or the macro environment is favorable, but because Jupiter and Venus will align… and it’s his birthday. Which, you know, is sort of adorable and also the most Elon thing imaginable. If you’re the kind of person who thinks markets are driven by discounted cash flows, this might make you twitch a little. But if you believe, as many investors secretly do, that vibes and narratives move markets more than spreadsheets, then aligning a $50 billion IPO with celestial movements and a billionaire’s birth might actually be smart. After all, if astrology moves crypto prices, why not rockets?

But behind the horoscope headline is a very terrestrial reason. SpaceX needs cash… a lot of it. With Starship development, AI compute ambitions, and Musk’s vision of building orbital data centers (which, let’s be honest, is the closest thing to real-life Halo infrastructure we’ve seen), the company is chewing through capital faster than it can privately raise at Musk-approved valuations. The $800 billion valuation floated in private markets last month is already a flex, but now they want $1.5 trillion. The demand from institutions and retail alike is going to be absolutely feral, especially after being locked out for so long. Musk may be eccentric, but he understands scarcity.

And that’s the core of the joke here. The listing is a spectacle designed by the man who sent a Tesla Roadster into orbit and tweeted “funding secured” at $420. But it might work, because Musk has built companies that operate outside traditional gravity, both physical and financial. If you believe the future runs on rockets, bandwidth, and AGI, then SpaceX is no longer just a space company, but a vertical integration of the 21st century. The planets may be aligning in June, but so are the incentives. Happy birthday, Elon. Here’s your $50 billion present. You are trillionaire now! FT

🇮🇹 Italy’s €2 Routing Fee

🇮🇹 Italy basically invented a new kind of tax, the “please land your plane somewhere else” fee.

The government slaps a €2 charge on sub-€150 parcels (hello, Shein/Temu), forecasts a nice, tidy €122m this year, and then watches as the parcels do what parcels do in a single market: they route around the obstacle with flights divert to Liège/Amsterdam/Budapest, Italy’s direct low-value arrivals drop 36% in the first 20 days, and the logistics guys are like ...

"congrats, you taxed our throughput while the goods still enter by truck, with extra pollution as a bonus".

This is a classic policy whack-a-mole. The EU’s whole thing is that once you clear customs you can circulate freely, so a national “handling fee” just turns into a geography game, and the funniest part is collection doesn’t even properly start until March....meaning Italy has managed to lose volume before it really collects the money, which is an extremely Italian outcome in the most affectionate sense. FT

🌽 Claude Code and the orchestration of corn

Seth Goldstein had dinner with Fred Wilson, and somewhere between the spring rolls and the walk back to the hotel, Fred said: “You can’t grow corn with AI”

... and like any rational person who took that as a personal attack, Seth launched a website, an ops stack, a decision engine, and an actual farm.

Powered by Claude Code, Anthropic’s AI that’s being treated more like a COO than a chatbot, this thing is now managing land leases, irrigation decisions, and coordinating with seed suppliers. I suppose that means we’ve reached the “software eats agriculture” phase of the cycle. Fred said you can’t make dinner, but it turns out if you’re the one hiring the chef, writing the grocery list, and running the kitchen… you’re sort of the chef now?

It’s AI orchestrating humans and services as if it were the farm manager. AI doesn’t need to do the physical labor... it just needs to know how to make decisions and send instructions to the people or tools that do.

That’s a different kind of automation, more Uber than Boston Dynamics. “Proof of Corn” might sound like a meme, but it’s a serious blueprint for how AI becomes an interface layer to the real world. The fact that it all started because someone said “you can’t” makes it even more inevitable.

The next harvest? Probably coming from an LLM prompt. LINK

Two AI Economies

Wall Street is having the most Wall Street conversation imaginable about AI. With tech companies, it’s all “show me the monetization path” and “how are you financing the capex to build Skynet", while with everyone else it’s basically "so…uh…have you considered putting ChatGPT in a hotel lobby?"..

..even though US businesses (outside the “AI builders”) spent an estimated $86B on AI in 2025 and are expected to spend $131B next year, which feels like the sort of thing you might ask about if your job is "ask questions" (but fewer than half of S&P 500 companies got any generative-AI questions in 2025, versus >80% of tech in the most recent quarter).

It looks like management teams keep talking about productivity, cost savings, headcount, etc., and analysts keep asking about adoption/development/ROI like they’re shopping for use cases on Etsy; and that’s how you get dot-com-bubble vibes.

A lot of investment upstream, a lot of confidence in destiny, and not enough curiosity about whether the rest of the economy is actually absorbing the thing fast enough to justify the buildout.

on crime (🇷🇴 vs. 🇸🇪 vs. 🇬🇧)

According to Numbeo’s Crime Index, Bucharest is basically a Swiss spa town (28.5) while Gothenburg is a war zone (45.83) and London is Mad Max with Pret (55.7). Malmö (🇸🇪) is the “bombing city” of Europe at 55.03, which is a very expected outcome if you import immigrants who develop parallel societies.

Big international cities are crime-content factories. London has more tourists, more pickpockets, more headlines, more viral videos, more “this city is finished” discourse. Sweden has a whole export industry of “Nordic dystopia” stories. Meanwhile, Bucharest’s biggest comparative advantage here might be "come here! It's super safe and we don't have bad neighborhoods" (to be fair, as of 2023, the Swedish Police Authority identified 59 geographical areas, often referred to in media as "no-go zones")

You can absolutely use these numbers to win an argument on the internet. But if you’re trying to decide where to live, invest, or walk home at night, you probably want to live in those places for 1-3 months to see if it's for you or not.

Numbeo tells you what people say they feel. Reality can be harsher. LINK

🍽️ AI eats software

Yesterday, I did a deep dive into the software industry ("AI eats software")

The 2014-2021 era was "IPO as graduation". The 2022-2025 era is “IPO in AI times”.

Software didn’t stop being important; it's just that investors just decided they’d rather own it through megacaps, private rounds, or M&A… than through a newly public company explaining quarter-by-quarter. LINK

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution
   

Dear onStrategy Reader,

You can now start a one-month free trial for “Where is my moat?”

You won’t get my portfolio structure during the trial period, but you’ll get enough information and insights that can convince you if it’s worth it or not.

👉 Give it a try:

Where is my moat?

💬 Thoughts on Discord’s IPO implications and trade-offs

Discord’s IPO is a little bit like taking your favorite dive bar and putting it on the Nasdaq. For years, Discord has thrived in a cultural niche because it was (is) part Slack, part Reddit, part secret clubhouse where Gen Z can talk memes, games, mental health, AI image prompts, and homework… often all in the same thread. With 200 million monthly active users and $600 million in revenue, it’s a real business now, not just a gamer side hustle. But now comes the hard part: telling the Wall Street story without killing the vibe. Discord makes money from subscriptions (Nitro) and some experimental ad products, but it’s not clear whether its charming chaos can coexist with the demands of quarterly earnings calls and EBITDA margin targets.

The problem is that Discord was never designed to scale like a traditional tech platform. On the contrary, it was optimized for intimacy, not monetization. Communities are siloed, content isn’t algorithmically pushed, and users are fiercely allergic to being treated like monetizable eyeballs. That makes it hard to inject ad dollars at scale without breaking what makes Discord feel authentic. Authenticity doesn’t pay the bills or justify a unicorn valuation. The new CEO, who helped sell Activision to Microsoft, probably knows this. But selling games and selling ads are two very different businesses, and it’s unclear whether Discord wants to become more like Reddit (content ad tech) or Zoom (enterprise SaaS). Either path requires trade-offs that risk alienating users or disappointing investors… or both.

Also, let’s not forget the real risk. This is a social platform with lawsuits, regulatory attention, and a long history of being a safe haven for chaos, extremism, and under-moderated content. That’s not the kind of thing that underwriters love to highlight in roadshows. But if the IPO does go ahead and if Discord manages to monetize without over-monetizing, it could become a case study in how to take a messy, beloved corner of the internet and give it a market cap without killing the culture. Or it could become the next cautionary tale in the long list of social platforms that tried to grow up and lost their soul in the process. Either way, the IPO deck better be really, really good. Bloomberg

on (drinking) alcohol 🥃

Imagine borrowing money at near-zero rates to distill the finest Cognac, then aging it for half a decade while your customer base suddenly discovers sobriety (hey, Gen Z), semaglutide, and... sparkling water 😅

That’s where we are. The world’s biggest booze brands (Diageo, Pernod Ricard, Campari, Brown-Forman, and Rémy Cointreau) are sitting on $22 billion worth of aging spirits that nobody wants. Inventory-to-sales ratios are through the distillery roof. Remy’s stockpile alone is 2x its annual revenue and nearly equals its market cap. They bet big on COVID-induced drinking being a forever trend. It wasn’t.

The hangover hit these producers, but not just from the macro squeeze on discretionary spend, but also from the rise of health culture and drugs that literally make you not want to eat or drink.

Current status: Jim Beam’s main distillery is shuttered, Diageo has halted whiskey ops in Texas and Tennessee. Also, they can’t cut prices too much because… debt. When you’ve levered up to fund barrels that take 10 years to mature, fire sales become balance-sheet disasters.

What a strange place to be. Caught between a whiskey lake and a dry consumer, which one might come up with something like “drunk on CapEx, sobered by demand”.

The peak of Social Media?

There’s a kind of poetic justice in the idea that Gen Z, the very cohort that made TikTok dance trends and Instagram filters part of the cultural oxygen, is also the first to start cutting back on social media use. The FT chart makes it clear. Time spent on social platforms peaked around 2022, and it’s the youngest users who are leading the retreat. Maybe it’s digital fatigue, or maybe it’s just that the dopamine loops are less satisfying when everyone’s chasing the same algorithmic clout.

Either way, it’s bad news for engagement-driven ad models, especially those banking on infinite scroll as their economic engine, and probably good news for, I don’t know, walking outside?

The under-35 crowd has always been the canary in the monetization coal mine, and if their screen time is dropping, it means ad inventory is flattening, user growth is maturing, and the attention pie isn’t growing the way Meta and Snap were hoping.

Sure, older users are still ramping up, but you can only grow MAUs so long on the backs of 55-year-olds discovering Facebook Marketplace. Social media will not die, but the era of effortless exponential growth might be, and for companies built on selling engagement to advertisers, or VCs, this looks a lot like the beginning of a post-peak internet attention economy.

End of an era: TLC (China) buys Sony’s home entertainment business (Japan)

Sony sold its soul for margin. Or at least 51% of its Bravia TV business, which is pretty close. In a move that feels less like strategy and more like surrender, Sony is handing over control of its home entertainment unit to China’s TCL…. the budget-TV behemoth that’s now taking over the big boys’ toys. This new joint venture means TCL will manufacture TVs branded “Sony” and “Bravia”, but using TCL’s display tech. Which is a little like buying a Rolex and finding out the engine is from Casio. Sony gets out of a low-margin business, TCL gets brand credibility, and investors get…confused nostalgia.

Of course, it makes perfect sense. Bravia, for all its glossy ads and cinematic purity promises, was never really about TVs. It was a relic of the old Sony, the one that thought hardware margins could still pay for Tokyo real estate. These days, Sony is more Netflix than Panasonic. It wants you to buy anime subscriptions and Spider-Man tickets, not quibble over contrast ratios at Best Buy. Of course TCL is happy to build everything for everyone, so long as they can slap on someone else’s logo and sell you the remote. It’s outsourcing with a licensing twist or IP as the moat, margin as the master.

Who can blame them? Sony’s operating profit margin on TVs is a rounding error, while its media and gaming assets are IP gold mines. TCL gets distribution, Sony gets to look focused, and Bravia gets to live on as a ghost in the machine, whic is, beautiful on the outside, Chinese on the inside.

Just don’t be surprised when your Sony TV starts recommending Chinese products and services by default. Or, in a few years, when Sony sells the other 49% and says it was never in the TV business to begin with. Bloomberg

Thanks for reading onStrategy! Subscribe for free to receive new posts and support my work.

WIMM - Where is my MOAT?

If you enjoy my strategy notes on LinkedIn or in OnStrategy, WhereIsMyMoat.com is the deeper, “skin-in-the-game” version: full theses, moat scoring, and a clear view of the company’s future performance.

Last week’s analysis:

• 💿 ASML and the Physics Monopoly

• 🏦 TSMC and the AI Capacity Governor

• 🤓 The logic of my secondary portfolio (only for paying subscribers)

• My stock picks for 2026 (only for paying subscribers)

This week’s analysis:

• 22nd January - Essay] 🛍️ The Next Shopping Platform is an Interface

• 23rd January - 📈 The Two-Broker Strategy (what investing apps are out there)

• 24th of January - 👟 The Great Nike Reset

• 25th of January - 🩴 The Comfort Economy and Birkenstock’s Moat

Subscribe to WIMM

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution

Dear onStrategy Reader,

You can now start a one-month free trial for "Where is my moat?"

You won't get my portfolio structure during the trial period, but you'll get enough information and insights that can convince you if it's worth it or not.

👉 Give it a try:

Where is my moat?

On to the update:

Why ChatGPT enters Health

OpenAI’s move into health is more about distribution. The company isn’t trying to become a hospital or even a health app, but it’s trying to become the interface through which people interact with complex systems. In this case, that system is the mess of medical records, wearables (Apple Watch), and symptom checkers that currently require ten taps, three portals, and a phone call. GPTs are increasingly commoditized, but in the process are becoming the default interface for a high-friction vertical with high value per user. If you can own “the thing people ask health questions to”, that’s a sticky entry point.

OpenAI is moving horizontally into every major vertical where information overload meets user confusion. Finance, health, education, legal. Health just happens to be the biggest and most trust-sensitive. While incumbents like Epic or Apple Health have better structured data, they lack the conversational layer that makes AI feel intuitive and omnipresent. OpenAI is betting that the UX wrapper is the moat. That’s why it wants access to health data: not to treat you, but to talk to you, and if you ask it enough questions before calling your doctor, it wins. OpenAI

📧 Finally: Gemini in GMAIL

Google adding AI overviews and summaries to Gmail is the most predictable outcome in AI, and also the correct one. When incumbents treat a new paradigm as a feature, it’s usually framed as defensive or unimaginative, but here it’s structurally advantaged. Gmail already owns the data, the context, the permissions, and the workflow. An external model trying to “help with email” is fighting uphill against access friction, latency, and user trust. Embedding AI directly into Gmail turns intelligence into infrastructure, and the implications are 3 fold: invisible, default, and hard to displace. The point is that Gmail stays the place where work happens.

This is fundamentally a distribution story. If models converge toward parity, then value accrues to whoever controls the surfaces where users already spend time. Google doesn’t need to invent a new AI product, but it just needs to make its existing ones quietly better. That’s the asymmetry OpenAI faces: it has the best conversational interface, but the incumbents have the interfaces that already matter. Google, Meta, Apple, Microsoft, and Amazon don’t need users to switch behavior, but they just slide AI into flows that already exist. In a world of commoditized models, distribution is the moat. LINK

🛒 Amazon plans “Big-Box” retail Store

Amazon is apparently trying physical retail again, because of course it is. You see, when you’re already a $1.8 trillion e-commerce behemoth with your own planes, robots, web servers, and a sitcom’s worth of abandoned grocery store concepts, why not spin the roulette wheel one more time and try to reinvent Walmart? The logic is charmingly Bezosian: physical stores might be low-margin and logistically messy, but if there’s even a 1% chance of discovering the magical omni-channel hybrid that makes everything click (ie. the fulfillment center that also sells socks and hot dogs) then the optionality alone is worth the burn rate. Besides, if you’re Amazon, failure is a line item, not a red flag. CNBC

🚘 “Full Self-Driving” - as a subscription

Here we are: Tesla’s Full Self-Driving (FSD) will no longer be available for purchase starting February 14, and will instead be subscription-only (probably $99 per month, or $999 per year). Which is perfect, because love is temporary but recurring revenue is forever. From a financial perspective, this is Elon’s classic “turn the knob” play: take a lump-sum software add-on and flip it into an annuity stream that fattens gross margins and provides Tesla with SaaS-style optionality. It’s less like selling you a car and more like putting you on a payment plan for your robot chauffeur. It is also convenient because it allows Tesla to preserve the illusion that FSD is still in beta while monetizing it like it’s already perfect.

More interesting, though, is the signaling game. Tesla’s basically telling you that self-driving is no longer a novelty or a gamble, but anutility, like Netflix or iCloud storage. The technology has improved to the point where locking in a permanent license feels like a mispricing. And since they control the OTA (over-the-air) pipeline, Tesla gets to flip the monetization model at will, which also makes resale math fuzzier. Can you sell a used car with a feature that now lives behind a paywall? Turns out, autonomy doesn’t just disrupt driving, but it also upends asset ownership. Welcome to the world where your car becomes a platform, and every feature gets monetized like an app.

WIMM - Where is my MOAT?

If you enjoy my strategy notes on LinkedIn or in OnStrategy, WhereIsMyMoat.com is the deeper, “skin-in-the-game” version: full theses, moat scoring, and a clear view of the company’s future performance.

Last week’s analysis:

• 🏡 Airbnb’s real moat is demand

• 🏎️ Ferrari’s scarcity stack

• 💰 The logic of my main stock portfolio (+62.5% in 2025)

This week’s analysis:

• 15th of January - My 2nd portfolio analysis (RO + PL)

• 16th of January - What stocks I might buy in 2026

• 17th of January - ASML

• 18th of January - TMSC

Subscribe to WIMM

Thank you for being an onStrategy reader!

Interesting newsletters 💡

1. The Weekly Blend

2. Product evolution