The UK government has taken an equity position in AI semiconductor startup OLIX as part of a $312 million Series B financing that values the London-based company at $3.3 billion, extending Britain’s effort to build domestic control over the hardware required to run advanced artificial intelligence systems.

The investment came through Sovereign AI, a government-backed venture program designed to support strategically important British AI companies. Officials did not disclose how much public money was invested in OLIX or the size of the government’s resulting stake. The company is the fifth startup to receive an equity investment from the program, while 11 companies have received some form of Sovereign AI support, including access to computing resources.

OLIX announced the financing alongside investments from Fundomo, Arm and Hudson River Trading. Netflix co-founder Reed Hastings also participated as an individual investor. Existing backers including Hummingbird Ventures, Crane, Plural, Creandum, Phoenix Court and Transition increased their commitments, according to the company.

The new valuation is more than three times the roughly $1 billion valuation OLIX received when it raised a $220 million Series A earlier in 2026. Founded in 2024 by entrepreneur James Dacombe, the company has quickly assembled one of Europe’s best-funded attempts to challenge the established architecture of AI data centers.

OLIX is focusing on inference, the computing process through which a trained AI model generates answers, images, code or other outputs. Inference has become an increasingly important part of AI infrastructure spending as companies move from training large models to operating them continuously for consumers, businesses and autonomous software agents.

The startup argues that using a general-purpose accelerator for every stage of inference creates avoidable costs and efficiency losses. Its X-1 platform is intended to distribute different stages of a model’s workload among specialized processors, resembling a production line in which each machine performs a particular task.

Those processors would be connected through what OLIX describes as a “slow and wide” optical interconnect. Instead of relying entirely on conventional copper links, the system would use light to move data between chips with low latency and reduced energy consumption. A deterministic compiler would coordinate workloads across complete racks of hardware.

The first processor planned for the platform is DX-1, a decode accelerator designed for the stage at which a language model reasons and produces output tokens. OLIX says that, for models with approximately 100 billion parameters, DX-1 can deliver more than 10,000 tokens per second per user while achieving higher output throughput per watt than general-purpose chips running large batches.

Those figures are company projections and have not yet been demonstrated in broadly available commercial systems. OLIX remains in the development and manufacturing preparation phase, and the financing is intended to support delivery of DX-1 to its first customers during the second half of 2027.

The design is also meant to reduce exposure to some of the semiconductor industry’s most constrained components. OLIX says DX-1 will hold model data in fast on-chip static random-access memory and will not require high-bandwidth memory or advanced packaging. Those technologies have become important bottlenecks as Nvidia, hyperscale cloud operators and other chip developers compete for limited production capacity.

Avoiding those components could give OLIX more flexibility in manufacturing, but it will also require the company to prove that its alternative architecture can deliver competitive performance, software compatibility and reliability at data-center scale. Building an accelerator is only one part of entering the AI infrastructure market; customers also expect mature developer tools, networking, technical support and predictable supplies of complete systems.

The company strengthened its links to Silicon Valley by appointing Nick McKeown to its board. McKeown, a professor emeritus of computer science and electrical engineering at Stanford University, helped develop software-defined networking, OpenFlow and the P4 programming language. He also co-founded networking companies Nicira, acquired by VMware, and Barefoot Networks, acquired by Intel.

OLIX separately appointed Matt Briers as chief financial officer. Briers spent nine years as CFO of financial technology company Wise and helped guide it through a 2021 direct listing on the London Stock Exchange. The appointment suggests OLIX is building financial and operational systems suited to a larger manufacturing business rather than functioning solely as a research venture.

Although OLIX is anchored in Britain, its ambitions and workforce are increasingly international. The company has engineering operations in London and Bristol and is recruiting across silicon, photonics, compiler and systems engineering in Austin, Toronto, San Francisco and its British locations. That footprint puts it in direct competition with U.S. startups and established semiconductor companies for scarce chip-design talent.

For American technology companies, OLIX could eventually become either a supplier, acquisition target or competitor. U.S. cloud providers and AI developers are actively searching for alternatives that can lower inference costs and reduce dependence on a single accelerator architecture. Specialized-chip companies such as Cerebras, Groq, Etched and SambaNova are already pursuing different portions of that opportunity, while major cloud platforms continue developing custom silicon internally.

The British government’s participation adds a geopolitical dimension. Governments increasingly treat access to AI processors as an economic and national-security issue, particularly as export controls, fabrication capacity and energy availability influence where advanced systems can be deployed. Britain’s strategy is not to replicate the entire semiconductor supply chain, but to use its strengths in chip design, photonics, research and venture formation to retain a larger share of the industry’s value.

Sovereign AI was introduced as a £500 million initiative offering selected companies equity investment, computing access, research support and assistance recruiting international talent. It complements a broader £1.1 billion UK AI hardware plan that includes funding for a national supercomputer, government purchases of emerging processors and investment in semiconductor skills.

OLIX now has enough capital to move deeper into engineering validation, manufacturing agreements and early customer deployments. Its central test will come when prospective buyers can compare the X-1 platform with incumbent systems using real models, production software and data-center power constraints. Until then, the $3.3 billion valuation represents confidence in an architecture and team that have yet to deliver their first commercial rack.

For Britain, the investment is a wager that government capital can help prevent promising hardware companies from relocating or being sold before reaching scale. For Silicon Valley, it is another sign that the next phase of AI competition will not be confined to model developers. The companies that reduce the cost of producing each AI-generated token may gain influence over the economics of the entire industry.