Whither Europe?
After Davos
Whither Europe? After Davos
If we follow the money and the plumbing this week, Europe's problem stops looking like "too little AI" and starts looking like "too little power." Not compute power - political and financial power.
The conversation with Andrew Keen and Gené Teare begs a hard question that Carl Benedikt Frey only half answers: is it game over for Europe? Frey's "not yet - but close" sounds academic until we set it against this week's deals.
Look at who owns the AI foundations.
Apple, the archetype of vertical integration, quietly concedes that "Apple Foundation Models didn't actually have a foundation" and signs up to run Siri and future "Apple Intelligence" on Google's Gemini. Creator Studio spreads AI features across Final Cut, Logic and Pixelmator - but the cognition is rented.
The same week, Skild AI raises $1.4B at a $14B valuation to build an "omni‑bodied" robotics brain, while Nvidia executes the Groq non‑acquisition: license the LPU IP, hire "substantially all" the staff, leave GroqCloud as a legal fig leaf so antitrust can pretend there's still competition.
In my framing in the discussion I make the point that this isn't just product news; it's a new stack of power: models, chips, and robotics brains at the core; everyone else - including Apple - on the application edge. Europe has almost no one in that inner ring.
Gené's data makes it worse. U.S. venture is concentrating around the same four "venture majors" - a16z, Sequoia, GC, Lightspeed - while a16z alone pulls in another $15B, openly claiming that "the fate of new technology in the United States rests partly on our shoulders." SignalRank's Series B analysis shows these majors sitting in 60-70% of "qualifying" deals. Carta rewires fund admin with 24/7 AI agents, and $120B of AI data‑center debt quietly migrates into special‑purpose entities held by banks and pensions.
We're not just centralizing technology; we're centralizing the plumbing of capital and risk around a tiny set of U.S. actors.
Now contrast that with China's model, which Gené and I call out: MiniMax, a DeepSeek rival, pops 87% on Shanghai debut; Hygon doubles post‑IPO. Beijing routes its AI and chip bets through domestic exchanges with explicit state blessing. It's chaotic, but it is unmistakably a power project.
Europe? Frey's warning is brutal: without a true single market for services, it really could be the end of the European dream of continent‑wide progress. The DLD panel made the subtext explicit: Europe regulates platforms it doesn't own, splinters its market, and wonders why founders and capital head for Delaware or Shanghai.
This week's stories suggest the question isn't "Can Europe catch up in AI models?" It's: can Europe stomach the power moves everyone else is already running?Off‑balance‑sheet data‑center vehicles. "License + talent" deals that would make Brussels blanch. Evergreen "firm, not fund" capital that commits for decades, not election cycles.
Looking ahead, three tensions will tell us whether "not yet - but close" becomes "too late":
Foundations: Does any European actor build or anchor a true foundation layer - model, chip, or robotics brain - or does the continent resign itself to being an OEM for U.S. and Chinese cognition?
Capital: Can Europe create its own "venture majors" and permanent‑capital vehicles instead of outsourcing growth equity to a16z and Sequoia?
Market: Will the single market for services ever be completed, so a Skild‑equivalent can scale from Lisbon to Tallinn without 27 frictions?
Frey is right: the endgame is political, not technical. The new stack is power. The only real open question is whether Europe chooses to wield it - or just regulate whoever does.
Essay
Nvidia and Groq, A Stinkily Brilliant Deal, Why This Deal Makes Sense
Stratechery • Ben Thompson • January 6, 2026
Essay•AI•Semiconductors•Mergers•Competition