(#138) AI is a competition between 🇺🇸 and 🇨🇳
Thank you for being one of the +4,500 minds reading this newsletter
Here is what you’ll find in this edition:
Onto the update:
Who's winning the AI stack? USA vs China
The US and China are racing to dominate artificial intelligence. Which is funny, because what they’re really doing is racing to define reality. The US approach is basically: build giant closed models, sell them by the token, then sue anyone who dares remix Harry Potter in a prompt. China’s approach? Open-source everything, subsidize everyone, flood the zone with a thousand models trained on 98% censorship and 2% nationalism, and hope one of them becomes the global standard.
It’s capitalism versus techno-industrial statecraft. One side is raising $100 billion from VCs to buy Nvidia chips they can’t get fast enough. The other is building AI supernodes in the Gobi Desert while quietly asking: “What if the future runs on our software, even if it’s mid-tier?”. Sure, US models win on the leaderboards, but China’s playing a different game: they’re optimizing for adoption. They don’t need to win the exam if they get to write the syllabus. LINK
The Internet vs. S&P 493
What we’re looking at here is a three-way race: the S&P 500 (the market), the MAG 7 (Big Tech), and the S&P 493 (the rest of the economy pretending it’s still 2015). The MAG 7 is doing MAG 7 things, that is posting 50–60% net income growth like they are a startup. Meanwhile, the S&P 493 is limping along with 2–3% growth, which is just fast enough to get outrun by inflation and laughed at by Nvidia shareholders. If the S&P 500 were a group project, seven kids built a working rocket ship while the other 493 brought glue sticks and excuses.
Balaji’s point - “the Internet is not the same as America”- is showed here in bright red bars. The US economy is bifurcating: one part is writing AI models, dominating global cloud infrastructure, and optimizing ad auctions; the other part is asking the Fed if it can please have another quarter point. The stock market knows this, which is why it’s priced like the MAG 7 are the market, and everything else is beta ballast.
In conclusion, America, minus the internet, is kind of boring and probably overvalued. LINK
Bitter lessons for Europe 🇪🇺. A "robotaxy" story
Lyft (🇺🇸) is partnering with Baidu (🇨🇳) to deploy thousands of robotaxis across Europe. This follows its acquisition of Freenow, one of the continent’s largest taxi apps. Meanwhile, Hon Hai (aka Foxconn) is selling off an EV plant in Ohio to focus on building AI servers. Two deals. Different companies. Same underlying story. Let’s start with Lyft. They’re not bringing their own robotaxis. They’re bringing Baidu’s. Because apparently, the best way to disrupt transportation in Europe is by outsourcing autonomy to a Chinese AI firm. This is the future of Europe: a place where other people bring the tech, build the fleets, and write the software, while the local economy… observes. It is well known by now that Lyft acquired more than a taxi platform. They acquired geographic regulatory compliance. Because if you want to play in Europe, you don’t need a factory, but you need a legal wrapper. The continent has become a sandbox for innovation, not a source of it. Meanwhile, back in the USA, Hon Hai (ie. the same Taiwanese giant that built your iPhone) is shifting from electric vehicles to AI servers. Why? Because AI has margins and tailwinds. EVs, at least outside of Tesla, have neither. So Europe gets a fleet of Chinese robotaxis and America gets to manufacture the AI that powers them. Again, Europe gets the “user experience.” There was a time when "expanding into Europe" meant building things, employing people, or maybe writing some code in Berlin. Now it means launching a pilot program with regulatory footnotes and hoping the GDPR bot doesn’t eat your LIDAR feed. To be clear, this isn’t colonialism or not even soft power. It’s the economic equivalent of renting out your cities as product demos. The continent has the architecture and roads. What it doesn’t have is the chips, the models, or the ambition to scale them. Europe is, increasingly, an open-air museum where American and Chinese firms test their toys, then fly home with the data and build the next generation of innovation. Q.E.D. LINK
Europe externalized CO2 to China, but also the jobs & innovation
Europe has heroically reduced its CO2 emissions over the past two decades, presumably by sending all the factories to China. (ie. steel, chemicals, and solar panel production, just to name a few). Europe is becoming the environmental conscience of the world not by decarbonizing industry, but by decarbonizing itself out of having one. Meanwhile, China (bless their coal-fired ambition) decided they’d like to be rich first, and virtuous later.
This is what geopolitical arbitrage looks like. Europe emits less CO2 because China emits more, and China emits more because they’re building everything, for everyone, everywhere. The EU gets the clean air and ESG reports, while China gets the exports, the batteries, and the leverage.
This is about who gets to write the next chapter of industrial history. Spoiler: it’s not Brussels.
👩💻🧑🏻💻 Consulting brings net benefits to companies
At long last, we have it: the empirical vindication of consulting. The economists have stopped laughing, started measuring, and concluded that YES, consultants are actually doing something and not just generating invoices and airport lounge loyalty points 🙃.
This new NBER working paper by Bijnens, Jäger, and Schoefer is the consulting industry’s version of a glowing Yelp review backed by twenty years of Belgian VAT data. It turns out that when consultants show up, firms change things… and sometimes those things are even good.
The findings? Delightfully nerdy. High performers and low performers both hire consultants. The top ones to stay sharp, the bottom ones to survive. And once hired, consultants are like espresso shots for organizational change: episodic, intense, and often more psychological than operational. Firms spend 3% of payroll on them, which is basically the cost of a couple of middle managers, except these ones come in tailored suits and leave behind powerpoint decks with titles like "Transformation Roadmap v7 FINAL FINAL.pptx" 🫠
What’s most satisfying, though, is how the study finally treats consultants as an actual economic actor...not a meme. They’re not just "advice mercenaries" parachuting in with buzzwords; they’re intervention agents with tangible effects on firm productivity and hiring. Are they overpaid? Probably. Do they occasionally repackage the obvious as the profound? Of course. But also, they work 😅. So go ahead, strategy consultants of the world, collect your case fees. The data says you’ve earned them. LINK
Chat GPT for running a country? Sweden 🇸🇪 said “Why not?”
The Prime Minister of Sweden is using ChatGPT as a second opinion, which is somehow both extremely 2025 and oddly comforting. "What have others done?" and "Should we think exactly the opposite?" are basically the two pillars of modern policymaking anyway. It’s just that now, instead of asking aides, he’s asking a large language model trained by an American or (God forbid) by a Chinese company. It’s like replacing your think tank with autocomplete, but in a good way. There’s a subtle brilliance here: he’s not asking AI to decide, but he’s asking it to suggest. Like a political Clippy. "It looks like you’re drafting a tax reform. Would you like help avoiding public outrage?". And sure, there’s the usual disclaimer: no uploading confidential reports, no letting the algorithm write legislation (yet). AI as the diagnostic tool, the second set of eyes, the neutral nudge when politics gets blurry. It’s the same way a doctor might ask AI about rare side effects, except this time it’s a prime minister wondering whether banning Tiktok will trigger a trade war. And of course, this is how it starts. Not with robots seizing power, but with leaders asking them for advice, "just to check". First it’s "What would others do?". Then it’s "Draft me three policy options and rank them by popularity". Eventually, the assistant becomes the architect. Governance by autocomplete, democracy via prompt engineering. You wanted technocracy? This is the final form. LINK
The Brits are saying “No, thank you” to July & August holidays in hotter than ever Europe
Climate change has officially entered the holiday booking flow. What used to be a seasonal certainty (ie. Brits roasting in Mallorca mid-August) is now a liability. Tourists are voting with their feet, and their flight dates, ditching scorched summers in favor of cooler, quieter, and cheaper shoulder seasons. I know because I did too 🙋🏼♂️ It’s a rational response to instability, wildfire risk, heatstroke headlines, and the slow realization that "sunny" now comes with a risk disclaimer. Tour operators are responding not with innovation, but with a quiet calendar shuffle. Extend the season, reframe the offer, and hope no one notices the world is bending under extreme weather. Europe may still be the dream since only in 2024 were there 747 million tourists, but now it comes with a clause: only if it’s below 35C . LINK
on Manufacturing 🏭
Innovation without manufacturing is just powerpoint
The chart shows China casually walking away with global manufacturing while the USA and the EU cheer from the sidelines. "You go ahead, we’ll focus on the services economy", they said. Bold move.
And now? China owns the factories, the supply chains, the upstream permits, the downstream energy inputs, the rare earths, the robotics, the glass, the steel, the solar panels that power the factories that make the solar panels.
Meanwhile, the West? Well, we have the powerpoints. And consultants. And regulations so thick you need a lawyer, a therapist, and probably a shaman to build a semiconductor fab.
Here’s the main conclusion: if you don’t manufacture, you can’t innovate. Not really. You can imagine things, sure. But imagination without iteration is fiction. And iteration happens on the shop floor.
Want to lead in EVs? Build battery lines.
Want to lead in AI? Train the models on chips you actually make.
Want to lead in defense tech? Weld something.
Innovation is not an idea. It’s an industrial process. It requires permits to build, workers to tinker, and enough cheap energy to melt sand into something useful.
So when we marvel at China’s 30%+ share of global manufacturing value-added, we are actually looking at a strategic moat. A self-reinforcing flywheel of productivity, learning, and innovation… with export controls as a cherry on top.
America still dominates in chips design (for now).
Europe still has a few machine tool champions (for now).
But if you can’t make the thing, then eventually, you don’t own the thing.
Because you can’t debug hardware with a TED Talk. LINK
PRINCIPLE: Knowledge and productivity compound
Two people, same baseline. One works just 10% more. Not all-nighters, just a little extra every day.
One more book chapter, one more draft, one more question asked.
Fast forward five years:
–> The 10% one doesn’t just edge ahead, they’re on another level.
–> Because effort compounds. Knowledge compounds. Reputation compounds.
Most people underestimate how small advantages become massive over time.
They chase hacks. But the real hack is consistency. You don’t need to double your talent. Just work 10% smarter, longer, deeper.
That’s how you get 200% results. And no, it’s not magic. It’s math.
Head of Tax Technology & Innovation at Deloitte | I write about Antifragile leadership | MBA
1moI went into the article hoping there's a larger picture that also captures Politehnica Bucuresti :P