A London semiconductor start-up called Olix has raised $312 million in Series B funding, valuing the company at $3.3 billion.
The Times reports that the round triples Olix's valuation. UK outlet digit.fyi puts the raise at £232 million, describing it as a Series B for the British AI chip start-up.
What makes Olix unusual is the architecture. According to a report carried on MSN, the company is betting on photonic interconnects — moving data using light rather than electrical signals — combined with on-chip SRAM memory, in order to bypass high-bandwidth memory, or HBM, in AI inference workloads. That same report frames the deal as Britain's biggest semiconductor bet.
To understand why that matters, it helps to know what HBM does. High-bandwidth memory is the expensive, stacked memory that sits alongside today's leading AI accelerators and feeds them data fast enough to keep their processing cores busy. It is also one of the tightest supply bottlenecks and biggest cost drivers in the AI hardware market. A design that keeps memory on the chip itself and uses light to shuttle data between components is aiming squarely at that bottleneck.
Inference — running trained AI models to answer queries, rather than training them in the first place — is the workload Olix is targeting, per the MSN report. That is the part of the AI stack that scales with usage, so efficiency gains there translate directly into lower running costs for anyone serving models at volume.
The sources available describe the funding, the valuation and the technical approach, but not customers, shipping timelines or independent performance results. Those remain claims by the company rather than demonstrated products.
Why it matters: if photonics and on-chip memory can genuinely deliver AI inference without HBM, it would loosen one of the industry's hardest supply constraints — and it would do so from a British start-up rather than the incumbents that dominate AI silicon today.