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

Set Trending Topics as a preferred source on Google.

London-based deep tech company Olix has raised a $312 million Series B. Arm, high-frequency trading firm Hudson River Trading, Netflix co-founder Reed Hastings — and the British state — are on board. The round lands in the hottest phase yet of the race for AI inference chips.

Two years after its founding, London chip startup Olix has announced a funding round of $312 million at a $3.3 billion valuation. The company, which develops its own chips, lasers and networking for AI inference and ships them as complete racks, only reached unicorn status in February of this year: back then it raised $220 million at a valuation of just over $1 billion — one of the largest Series A rounds ever raised by a UK technology company. Within roughly six months, the valuation has more than tripled.

Olix was founded in 2024 by James Dacombe, who had previously built the neurotech company CoMind. He was 23 at the time of founding.

Who invested

According to the company, the round includes New York deep tech investment firm Fundomo, British chip designer Arm and US high-frequency trading firm Hudson River Trading. Angel investors also participated, among them Netflix co-founder Reed Hastings. Industry media report Fundomo as the lead investor; the firm was already involved in the February round.

All existing investors increased their commitments: Hummingbird Ventures, Crane, Plural, Creandum, Phoenix Court and Transition.

Not mentioned in Olix’s own announcement, but communicated by the British government on the same day: the state venture fund “Sovereign AI” is also participating in the round. That makes Olix the fifth company to receive an equity investment since the fund launched; according to government figures, the programme supports eleven startups in total, some of them with access to compute. AI Minister Kanishka Narayan justified the investment by arguing that countries which build chips build leverage. Sovereign AI Head of Ventures Joséphine Kant pointed to the name of Dacombe’s holding company: Bletchley Industries — a nod to Bletchley Park, home of Britain’s Second World War codebreakers.

What Olix is building

Olix argues that the industry’s standard approach to building inference hardware is running into efficiency limits. The image the company uses: a data centre is a factory whose product is the token. Hundreds of computing operations with very different requirements all run across one and the same general-purpose chip — while any other factory would have a dedicated machine for each stage of production.

The company’s answer is called X-1: models are “unrolled” across a large number of chips, with each chip handling only one part of the model. Because model architectures are changing quickly, each chip retains a flexible compute fabric and does not hard-code any particular architecture, according to Olix. The chips are linked by a “slow and wide” optical interconnect that moves data using light instead of copper; distributing workloads across racks is handled by a deterministic compiler.

The first chip in the platform is DX-1, a decode accelerator for the stage in which a model generates its output. For models in the 100 billion parameter class, Olix promises more than 10,000 tokens per second per user at higher output throughput per watt than general-purpose chips running large batch sizes; the architecture is meant to scale to models of 10 trillion parameters and beyond. There are no independent measurements of these figures so far.

A second point is strategically relevant: DX-1 holds the model in fast on-chip memory (SRAM) and, according to the company, works without advanced packaging and without high-bandwidth memory (HBM) — precisely the components the industry is currently shortest of.

McKeown and former Wise CFO Briers join

With the round, Olix brings in two prominent names. Nick McKeown — co-inventor of software-defined networking, OpenFlow and P4, Professor Emeritus at Stanford and winner of the 2025 Marconi Prize — joins the board of directors. McKeown co-founded Nicira (sold to VMware) and Barefoot Networks (sold to Intel), among others.

The new CFO is Matt Briers, who spent nine years as CFO of fintech Wise and took the company through its 2021 direct listing on the London Stock Exchange — at the time the first direct listing by a technology company in London.

Context: the race for inference

The round comes at a moment when capital is pouring into specialised AI chips. According to Dealroom data, AI chip startups raised around $8.3 billion worldwide in 2026 alone. The reason for the shift in focus is inference: it is no longer training models but running them in everyday use — agents, reasoning models, deep research — that drives compute and energy costs.

The competition is well capitalised too. Nvidia acquired assets of inference startup Groq in a $20 billion deal in December and subsequently put $4 billion into photonics companies. Cerebras most recently raised $1 billion at a $23 billion valuation, d-Matrix took $275 million in a Series C, and Etched is valued at roughly $5 billion. The US lead in capital and scale therefore remains substantial — and is a key reason why London frames the Olix case as an industrial policy opportunity.

What happens next

The fresh capital is meant to get DX-1 to the first customers by the second half of 2027, and to fund the build-out of the wider custom silicon platform along with the manufacturing and supply chain commitments required. Olix says it is hiring across silicon, photonics, compiler and systems engineering — in London, Bristol, Austin, Toronto and San Francisco.