Read Index Ventures' own description of its mandate and you find the tell hiding in plain sight. The firm just raised $2bn in fresh capital, lifting its deployable war chest to $3.5bn, and it describes what that money is for as "the ten-hour time zone that runs from Tel Aviv to San Francisco."
Notice what that sentence leaves out. It names Tel Aviv. It names San Francisco. Europe—the continent where Index is headquartered, the one whose ecosystem this raise is being celebrated as a triumph for—is the implied middle, the connective tissue, the place the money passes through on its way somewhere else.
That is the paradox of this week. In a single stretch, Europe minted both its largest-ever VC AI raise and its largest-ever chip round. And both are structurally engineered to send the value westward.
In a single stretch, Europe minted both its largest-ever VC AI raise and its largest-ever chip round. Both are structurally engineered to send the value westward.
Last week this column tracked the silicon-side Nvidia challengers—Fractile, D-Matrix—as a compute-economics story. This week the same thread reappears on a different axis entirely: not whether the chips beat Nvidia, but whether the continent that funds them ever gets to keep the winner.
Start with Index, because it says the quiet part cleanly. The $3.5bn splits into a $400m seed fund, a $900m venture fund, and $700m added to its 2024 growth vehicle, now $2.2bn. The stated purpose is backing founders "from first check to the public markets and beyond."
The record it is deploying against is real and mostly American in its payoffs. Wiz sold to Google for $32bn—reportedly worth about $3.8bn to the fund, per TNW. Figma went public. A Revolut secondary marked the company at $115bn. Index booked close to $9bn, per TNW.
And the AI portfolio it wants to double into reads like a map of global, not European, concentration: Anthropic, Mistral, Fireworks AI and Physical Intelligence. These are bets on companies designed to win a winner-take-most market—a market whose late-stage capital and regulatory clarity sit in the United States.
That is the crucial distinction. Index is a European-domiciled firm making champion-agnostic bets. The money is European-raised. The conviction is about whoever scales fastest, wherever that happens.
There is even an internal dissent worth noting. Neil Rimer, who co-founded Index and stepped back from daily investing in 2021, told TechCrunch he has "a strong sense that there will be some sort of a redistribution" of the wealth AI is minting. His firm's response was to deploy more, faster, into precisely the boom he is uneasy about.
Now Olix, the more revealing case because it is a European company that could, in theory, be a European champion. James Dacombe, 25, London-based, two years old, just raised $312m at a $3.3bn valuation—more than triple February's $1bn mark. It is Europe's largest-ever semiconductor round.
The pitch is genuinely differentiated: specialized chips for each stage of AI inference, wired by an optical network moving data with light instead of copper. The first chip, DX-1, dodges the components in shortest supply—no high-bandwidth memory, no advanced packaging.
But watch where the company actually lives. The round was reportedly led by New York's Fundomo. The board added Stanford's Nick McKeown. And the hiring map spans London, Bristol, Austin, Toronto and San Francisco.
A UK company, yes. A UK-anchored one, no. Olix is being built across the same westward gradient Index described—British invention, American capital, transatlantic talent, and a company structured from the outset to sell into a global market.
The eu-startups line about founders "structuring their companies with momentum and cross-border growth in mind from the outset" is not a warning here. It is a description of what Olix already is.
And then there is scale, which is where European conviction runs into the hard ceiling. Etched is valued at $10.3bn. SambaNova at around $11bn. Both dwarf Olix's $3.3bn. Cerebras raised a $1bn round at roughly $23bn, per PitchBook.
European chip funding hit $2.1bn this year, almost double last year's total—and, as the report noted, the continent's sector "remains a fraction of the size of America's." The continent is growing fast off a small base while the base it is chasing grows faster off a larger one.
That is the scale-up gap in one comparison. Early money is available in Europe. The hundreds of millions—then billions—that decide winners in a capital-intensive game like frontier silicon are disproportionately American.
Which is why the capital raised this week is a necessary but wildly insufficient condition. Conviction that cannot buy scale is just well-funded feeder-league: you develop the talent, and someone else fields the winning team.
So what actually settles this? Two things, and only two.
The first is a single date. Olix plans to tape out this year and put DX-1 in customers' hands in H2 2027. As one account put it bluntly, whether the silicon delivers "is the only number that will settle it." Everything before that customer ship is a pitch deck with a Marconi Prize winner on the board.
Until then, the headline claims—over 10,000 tokens per second per user on 100B-parameter models, scaling to 10-trillion-parameter architectures—are unfalsifiable. A European champion is not a valuation. It is a working product tested against real workloads.
The second is regulatory. This same week, the EU's AI Act moved "from theoretical to operational," in the words of A&O Shearman's Peter Van Dyck, with transparency obligations, deepfake labeling, and fines up to 3% of global turnover.
The stated worry is about US labs whose outputs reach EU users. But the load-bearing question for European champions is the one AWS's adoption data surfaces: SMEs facing up to €500,000 in compliance costs, startups dedicating one or two full-time employees to compliance, and companies prioritizing "markets where requirements are clearer and more predictable."
That is the mechanism by which regulation becomes a scale-up tax. Not because the rules are wrong, but because fragmentation across 27 interpretations makes Europe the place great companies scale from rather than scale within.
The eu-startups prescription is honest about the fix: make the Single Market real for AI, so compliance in one country unlocks all 27; treat scale-up capital as seriously as company creation. Both are structural. Neither was solved this week.
So here is the uncomfortable read on the two biggest European AI bets of the cycle. They are not evidence that Europe is building champions. They are evidence that Europe can raise champion-agnostic capital—and that the capital, when tested, will follow whoever scales fastest.
Index has told you that in its own words: the value runs a ten-hour arc, and Europe is not its terminus. Olix will tell you in H2 2027, when DX-1 either ships or does not.
Until then, the money is real, the conviction is real, and the scale problem is unsolved. That combination has a name, and it isn't a champion. It's a feeder league with an excellent balance sheet.