Olix Raises $312 Million at a $3.3 Billion Valuation to Build Photonic AI Chips in the UK

A UK chip startup founded only two years ago has just tripled its valuation in six months. Olix, a company building AI chips around light-based connections between processors, has raised $312 million in a Series B round that values it at $3.3 billion. The deal is one of the largest recent bets on British chip design, and it arrives with a signal that goes beyond one company: the UK government, through its Sovereign AI venture fund, has joined the round alongside Arm, the Cambridge chip designer whose technology sits inside most smartphones on the planet.

The February 2026 round had valued Olix at $1 billion after a $220 million raise. The new round, reported by Data Center Dynamics, lifts that figure to $3.3 billion. Investors named in the report include Arm, Fundomo, Hudson River Trading and Netflix co-founder Reed Hastings. The Sovereign AI fund also announced an investment, putting public money behind a private startup in a sector where the UK has long argued it should punch above its weight. That mix of backers matters, because chip companies need patient capital, specialist customers and a supply chain that can actually build what they design.

What Olix is actually building

Olix says it is building systems around a novel photonic interconnect. In plain terms, that means moving data between chips as light rather than as electrical signals through copper wires. As AI models get larger, the bottleneck is often not a single chip working hard. It is getting data in and out fast enough, and linking many chips together so they behave like one much larger machine. Copper connections use more power as speeds rise, and they generate heat. Light-based links, in theory, can carry more data with lower energy use, which is why several chip startups are chasing photonics at the same time.

The company describes its approach as a suite of accelerators, starting with a first chip called DX-1. According to Olix, DX-1 is designed to deliver strong decode performance, which is the stage where an AI model generates its answer token by token after it has read a prompt. Decode speed shapes how responsive a chatbot or assistant feels, and it shapes how many users one system can serve at once. Olix says each stage of inference can run on a specialised chip in its system, rather than forcing one general chip to do every job in the same way.

A second idea sits at the centre of the design. Olix integrates SRAM, a type of fast on-chip memory, with photonics, and it deliberately avoids HBM, the high-bandwidth memory stacks used in many leading AI accelerators. The company says that choice should help it beat HBM-based designs on throughput per megawatt and on total cost of ownership. It also argues that avoiding HBM insulates it from supply constraints that have affected the wider chip industry, where leading memory and packaging capacity is fought over by much larger buyers.

The slow and wide interconnect

Olix talks about a slow and wide optical interconnect that moves data directly from die to die at very low latency and low energy cost. The phrase sounds odd at first. Slow, in this context, does not mean the system is sluggish. It refers to sending many parallel light paths at modest speeds per path, rather than pushing a small number of paths to extreme speeds. The combined width can still move a huge amount of data, while each individual path stays easier to build and less power hungry.

Why the UK government is involved

AI Minister Kanishka Narayan was quoted in the report saying the future of AI will be built on chips that power models, and that countries that build chips will build leverage. He described Olix as exactly the kind of ambitious company the government wants to back through Sovereign AI, noting that in just two years it has become one of the UK’s most exciting AI startups, with chip technology that could help shape the future of AI. That language frames the investment as industrial policy as much as venture capital: backing a domestic designer that could reduce dependence on overseas chip suppliers.

What to watch next

Three questions will decide whether this valuation looks wise in hindsight. The first is delivery: does DX-1 reach customers in the second half of 2027, and does it perform on real inference workloads, not just on the company’s own benchmarks? The second is power and cost: does the SRAM plus photonics approach really beat established HBM-based systems on throughput per megawatt and total cost once cooling, networking and software are counted? The third is customers: will cloud providers, AI labs or enterprises commit to a new architecture, with the software work that switching implies, or will they wait for a second generation?

Frequently Asked Questions

What is Olix?

Olix is a UK AI chip startup founded in 2024 that is building accelerators around a photonic interconnect, starting with a chip called DX-1. It aims to link many chips together with light-based connections so AI inference can run faster and with lower power use than on traditional systems.

How much has Olix raised, and what is it now worth?

Olix has raised $312 million in a Series B round, according to Data Center Dynamics, valuing the company at $3.3 billion. That is up from a $1 billion valuation after its $220 million raise in February 2026. Investors named include Arm, Fundomo, Hudson River Trading, Reed Hastings and the UK government’s Sovereign AI venture fund.

When will Olix deliver its first chip?

Olix said the financing will help fund delivery of its first chip to customers in the second half of 2027. That date is a company plan, not a confirmed shipment, and the chip still has to prove itself on real customer workloads, power use and cost before it can be judged a success.