James Dacombe’s Olix raises at $3.3bn valuation

Three hundred and twelve million dollars. That's what James Dacombe's Olix pulled in this week, led by New York's Fundomo, with Arm, Hudson River Trading and Reed Hastings all writing checks. The number that matters more, though, is the valuation: $3.3bn, up from just over $1bn back in February. Nine months, tripled. That's the kind of jump that usually needs a war or a bubble to explain, and right now it's probably a bit of both.

Dacombe is 25. He's done this before, sort of — his first company, CoMind, worked on brain monitoring, which is about as far from chip design as you can get. Now he's chasing Nvidia directly, promising chips that are faster and cheaper, which is the pitch every Nvidia challenger makes and almost none of them deliver. What makes Olix's version slightly more credible is the design choice: its chips skip the components everyone else is fighting over, leading-edge silicon, high-bandwidth memory, advanced packaging. If the shortage story holds, that's not a workaround, that's a moat.

Olix isn't alone in this scramble. Axelera in the Netherlands raised over $250m in February. Fractile, another UK outfit building inference chips, closed $220m in May. Money is flooding into AI infrastructure broadly, chips, data centres, the wiring in between, and Europe wants a piece of a market America and increasingly China have owned. There's a specific twist here too: reports that a Chinese firm has started mass-producing chipmaking equipment rattled public chip stocks hard, and some investors in European startups actually see that as good news for the continent, since it could loosen the supply bottleneck everyone's stuck fighting over.

And that split, private enthusiasm versus public jitters, is the real story underneath the headline number. Public chip stocks have been shaky for months as investors start pricing in just how much debt is funding this entire buildout. Private investors keep writing nine-figure checks anyway. Olix hired Matt Briers, Wise's former CFO, presumably to bring some financial discipline to a company about to tape out its first chips later this year, with customers expected to get hardware sometime next year. That's the real test. Valuations are cheap. Working silicon isn't.