$312 Million for Olix: UK AI Chip Startup Valued at $3.3 Billion

Two years old, two funding rounds, and a valuation that's already tripled since February. That's the pace London-based Olix is setting in the increasingly crowded fight over AI inference chips. The company just closed a $312 million Series B at a $3.3 billion valuation, barely six months after raising $220 million at just over $1 billion. James Dacombe, who founded the company in 2024 at age 23 after building neurotech outfit CoMind, is now sitting atop one of the fastest valuation climbs in UK tech history.

The investor list reads like a deliberate signal-sending exercise. Arm is in. So is Hudson River Trading, the high-frequency trading firm not exactly known for dabbling in unproven hardware bets. Netflix co-founder Reed Hastings wrote a personal check. New York's Fundomo, already an investor from the February round, is reportedly leading again. And quietly, without appearing in Olix's own press release, the UK government's Sovereign AI fund confirmed its own participation on the same day — making Olix the fifth company that fund has backed with direct equity. AI Minister Kanishka Narayan framed it plainly: nations that control chip supply control leverage. Someone at the fund even pointed out that Dacombe's holding company is named Bletchley Industries, a nod to Britain's wartime codebreaking legacy that practically writes its own headline.

What Olix is actually selling is a bet that general-purpose AI chips are hitting a wall. Their pitch: a data centre is a factory that makes tokens, yet everyone runs wildly different computing steps through the same generic silicon, like building every car part on one universal machine instead of a dedicated assembly line. Olix's answer, the X-1 platform, spreads a model's computation across many chips, each one handling a slice of the workload, linked by an optical interconnect that moves data with light rather than copper. The first product, DX-1, targets the decoding stage of inference and claims over 10,000 tokens per second per user for 100-billion-parameter models — figures nobody outside Olix has verified yet. More interesting than the speed claim is the memory story: DX-1 reportedly runs without high-bandwidth memory or advanced packaging, the exact components currently in shortest supply across the industry.

To back the technical pitch, Olix pulled in serious credibility. Nick McKeown — Stanford professor emeritus, co-inventor of OpenFlow and P4, and a Marconi Prize winner this year — joins the board. Matt Briers, who spent nine years as Wise's CFO and steered that company through London's first tech direct listing in 2021, becomes Olix's new finance chief. Both hires suggest a company preparing for scale and scrutiny, not just another lab chasing a demo.

The timing matters as much as the technology. AI chip startups pulled in roughly $8.3 billion globally this year alone, per Dealroom, as the industry's spending shifts from training massive models to the much costlier business of running them constantly for agents and reasoning tasks. Nvidia bought Groq's assets for $20 billion in December and then dropped another $4 billion into photonics. Cerebras just raised $1 billion at $23 billion. Against that kind of American capital, Britain's bet on Olix looks less like a startup story and more like an attempt to buy a seat at a table the US and increasingly China already dominate. Whether DX-1 ships to real customers by late 2027, as promised, will tell us if the bet pays off — or if this was just very well-timed fundraising.