What actually happened on August 3
OLIX, a London-based AI chip startup founded in 2024 by James Dacombe, closed a $312M Series B at a $3.3B valuation. The UK government confirmed that its own Sovereign AI venture fund took part in the round, alongside Arm, Fundomo, Hudson River Trading, and Netflix co-founder Reed Hastings. The company described it as the largest chip funding round in European history.
The scale of the jump matters as much as the round itself. OLIX was valued at $1B in February 2026, just six months earlier. Tripling that to $3.3B in half a year says something about how quickly institutional and state capital is now willing to move toward AI hardware bets that sit outside the Nvidia ecosystem.
DataCenterDynamics, TechTimes, Yahoo Finance, thestack.technology, and BeBeez International all reported the round independently on the same day, which is itself a marker of how closely the AI hardware market is now being watched outside Silicon Valley.
The architecture: light instead of copper, SRAM instead of HBM
OLIX builds AI inference chips around a photonic interconnect. Rather than one general-purpose chip handling every stage of inference, different stages run on specialized processors connected by light rather than copper wiring. The company says this design gives better throughput per megawatt than conventional approaches.
The second design choice is the one that matters most for anyone thinking about supply chains: OLIX integrates SRAM directly rather than relying on high-bandwidth memory, the HBM that sits inside Nvidia's current chip designs. OLIX says this delivers a lower total cost of ownership than HBM-based designs.
That distinction is not a technical footnote. HBM production today runs through three suppliers - SK Hynix, Samsung, and Micron - and that concentration has already been a factor in rising AI hardware costs. A chip architecture that does not need HBM at all is a different kind of bet than a chip that simply uses HBM more efficiently.
Why every other headline is missing the real story
The coverage so far reads almost uniformly as a funding story: a European startup raised a record round to compete with Nvidia. That framing is not wrong, but it stops one layer too early.
The detail worth sitting with is who else is in the round. A state venture fund, the UK's own Sovereign AI fund, chose to put public money into a domestic chip architecture that is explicitly designed to avoid the HBM supply chain. That is not the familiar sovereign-cloud story, where independence means keeping data and workloads inside domestic data centers while still running on someone else's silicon. This is independence one layer further down, at the chip and memory-architecture level - a layer where the UK, and Europe more broadly, has had essentially no leverage until now.
That is the part that belongs in a business owner's notes, not the valuation number.
What this means for a multi-year AI infrastructure plan
OLIX says its first chip reaches customers in the second half of 2027. That is not a timeline that should change anything in an infrastructure plan being written this quarter. Nothing here is a reason to delay a purchase or renegotiate a contract today.
What it does change is the list of questions worth asking a hardware or cloud vendor during any multi-year AI procurement conversation. Ask directly what happens to pricing and availability if HBM supply gets tighter, and whether the vendor has any roadmap that does not depend on HBM at all. Before August 3, there was no credible, state-backed answer to point to. Now there is at least one, even if it is not yet shipping.
Treat this as a line item to revisit at the next infrastructure planning cycle, not as an action item for this week.
Read next: Arm's Growth Came From Rate, Not Just Volume | Nvidia Stopped Renting Its CPU Core From Arm



