The UK government’s £1.1bn AI hardware plan anchors a new national compute strategy, including a £750m supercomputer in Edinburgh and targeted support for UK chip startups, alongside major private commitments from AMD and Nebius worth nearly £4bn, and a growing stable of British AI firms spanning photonics, frontier models and industrial digital twins.
In sum – what to know
Funding and computers – the UK government’s £1.1bn AI hardware plan anchors a new national compute strategy, including a £750m supercomputer in Edinburgh, and support for UK chip startups.
Capital and infrastructure – AMD and Nebius have committed £2 billion and £1.75 billion, respectively, to UK chip supplies and neop-cloud capacity; AMD, in particular, is all across UK initiatives.
Networks and models – UK startups Oriole, Cosine, and PhysicsX are variously making waves with new GPU photonics, sovereign frontier models, and specialist AI digital twins for Industry 4.0.
Lots going on, lately, on the UK AI scene – as trailed in the newsletter yesterday, and meshed into discussion about France’s telco carve-up and parallel sovereignty campaign, also in the context of Google’s mega deal to outsource data centre capacity to SpaceX (xAI) for $920 million per month, and what it all means in terms of power (gross IPO valuations for private US tech firms) and control (desperate catch-up measures for public EU authorities). Here’s a closer look at the UK story, and there is some room – in the choreographed messaging – for positivity, it seems.
We will take each item, in turn; but the headline release is about the UK’s new £1.1 billion AI ‘hardware plan’, which includes a £750 million bundle for a supercomputer in Edinburgh, plus support items for UK start-ups building AI hardware. There’s more below, and it might be read in conjunction with France’s parallel; moves at its Choose France summit last week – where Softbank unveiled €75 billion of inward investment in AI data centre capacity. But both funding rounds, in France and the UK, should be viewed in the context of the Europe’s sovereignty pitch.
The UK government writes: “The global AI chips market is expected to reach $1 trillion in the early 2030s. If Britain could secure just five percent of this market it would bring $50 billion in revenue to the UK with tens of thousands of highly paid jobs in tech. British companies – from Arm, whose chip designs are used in everything from smartphones to AI data centres, to startups like Fractile and Olix, which have raised more than £320 ($440) million between them – are already leading the next generation of AI hardware. This plan backs them, and the startups coming behind them.”
Fractile and Olix should be discussed again; instead, there is detail below about the various efforts with AI innovation from Oriole Networks, Cosine, and PhysicsX – in the fields of scale-up/out interconnect photonics, sovereign frontier models, and industrial AI digital-twins. There is scattered detail below about UK initiatives by US chip firm AMD (£2 billion, for starters), as well, timed to coincide with the UK government’s new AI fund – all announced at London Tech Week. Also, Dutch neo-cloud data centre provider Nebius, has committed to invest £1.7 billion in capacity buildout.
Nebius deployed its first (Nvidia) GPU infrastructure in the UK in November last year; it will add three new sites, it said, including the latest Nvidia full-stack AI factory tech, with a view to reach 65 MW in 2027. British firms are “already” running production workloads on Nebius at scale, the company said – citing financial firm Revolut, serving using its “token factory” platform to deploy, tune, and scale crime-fighting agents to tackle bank fraud, and biology firm Prima Mente, using AI to tackle brain diseases. Nebius is hiring at some pace in the UK, it said.
It presented quotes from all the parties. UK AI minister Kanishka Narayan said: “We’re determined to make the UK the best place in the world to build and deploy AI.” Arkady Volozh, founder at Nebius, said: “The UK is one of the places where AI is being built, deployed, and adopted at the same time — by startups, by enterprises, and by the public sector.” Paolo Guglielmini, vice president for EMEA at Nvidia, said: “The UK is one of Europe’s most ambitious AI markets, with a clear public policy framework and a strong base of innovators.”
Meanwhile, new data from Tech Nation valued the UK tech sector at £1.2 trillion in 2026 and suggested UK AI startups have raised more than £8.2 billion in venture capital in the first half of 2026. “This supports the Prime Minister’s statement that UK technology startups have raised close to half of all European technology investment so far this year,” it stated. New data from Omdia also forecasts that IT spending across Europe will increase 8.2 percent in 2026, its fastest pace since 2021, reaching a total of $1.3 trillion.
Which is all a little incidental, perhaps; but works as a way into an overview of what’s happening in UK, with regards to AI economy building and sovereignty control. Here are some of the headlines…
Funding and computers
The UK government has announced a new £1.1 billion AI ‘hardware plan’ including certain measures to support British firms developing AI chips and computing power. At the top of the bill, the government will direct £750 million of the sum towards a new national AI supercomputer, a “heterogeneous mixed chip system” to be opened at the University of Edinburgh in Scotland by 2030. This will operate as part of the UK’s AI Research Resource (AIRR), which gives researchers and enterprises shared access to a network of national supercomputers – which also includes the Isambard-AI supercomputer in Bristol and the upgraded Zenith / DAWN supercomputer in Cambridge.
At London Tech Week yesterday (June 8), US chip-maker AMD said it is working at the University of Cambridge with Dell on the Zenith supercomputer and also the Sunrise fusion supercomputer system. Sunrise is another UK AI supercomputer, developed with the UK Atomic Energy Authority (UKAEA), and geared for healthcare research, climate modelling, materials science, engineering simulation, fusion research and scientific AI model development.
Of the new Edinburgh supercomputer portion, £400 million will go just to purchase “next generation” AI chips, including £150 million for inference chips and a further £250 million for “more specialised chips”. Somewhere in these commitments is a promise to direct purchases towards UK firms. The press statement says £150 million will go on “novel chips from innovative startups and British firms”, without making clear it is the same £150 million to go on inference chips, nor whether “innovative startups” and “British firms” are distinct or overlapping categories.
It suggests the UK has tilted the tables, at least, even if it is not pushing full-force protectionism. The move creates “an immediate opportunity for British firms” who are “already at the cutting edge” and “well placed to compete” and in more “novel” and “specialised” AI chip design. It adds: “The government is acting as an early customer to help bring new technologies to market.” Another £120 million is for a new ‘hardware innovation programme’ to give start-ups funding-flex to innovate – so “the next generation of world-leading chip companies are grown here in Britain”.
The tender process for the new supercomputer is about to kick off. The pitch is to give researchers, start-ups, and public services the computing power they need to develop and run AI securely in Britain. There is a raft of broader support measures, too. The British Business Bank is backing a new UK fund led by Silicon Valley investors Playground Global to invest in UK AI hardware companies – to the tune of £150 million, which represents the largest fund investment it has ever made.
Technology Secretary Liz Kendall said: “The UK is already a global leader in chip design, and I believe this is a race Britain can win. To do that, we must back more British AI – and that means investing in the chips, computing power and skilled people behind it. That is exactly what this plan does, backing the British firms developing the next generation of AI hardware, so we get more jobs, more growth, and more control over the technologies our future depends on. We are backing Britain because we believe in Britain.”
The message is that money is now available for startups to “grow and stay in the UK” – to help them “crowd in more private investment” to help the UK “compete with the biggest players on the world stage”. Playground Global will open its first office outside the US in the UK.
As well, there is £20 million to expand a ‘scaling inference lab’, like an ecosystem incubation and funding unit, hosted by the Advanced Research and Invention Agency (ARIA), the UK government’s deep-tech research funding agency (like DARPA in the US), and non-profit community AI infrastructure platform CommonAI. There is another £45 million for AI training, taking the UK government’s AI skills fund to £80 million, and a plan for a new £12 million university centre for doctoral training.
AMD has said it has a new project with Imperial College London to support computational science and research that relies on large-scale computing resources, including healthcare innovation and climate modeling. The government says it is already funding 300 undergraduate bursaries in semiconductor design this year, rising to 400 next year, and 500 after that; it has set aside £20 million to fund 500 more UK PhD students to research the same. The government has a new industry deal with UK-based (Softbank-owned) chip design giant Arm to support its skills initiatives.
Networks and models
There is genuine hope about UK-based AI startup growth and innovation, too. Oriole Networks is working with AMD at the inference lab, and is ready to deploy a new scale-up/out system that uses light rather than electrical signals to move data between chips inside data centers. It uses AMD’s GPUs and CPUs, and has been developed and tested at the lab as “the world’s first pure photonic AI network”, with wider industry rollout is expected in 2027 – to “supercharge AI output”, and push “inference throughput and interactivity out by an order of magnitude”.
The firm’s photonic network solution, PRISM, replaces electronic switches in the network core with optical circuit switching. “Electrical switches are inefficient, power-hungry, and generate enormous heat,” it writes. “Data center networks have been pushed to the breaking point.” Its switched optical circuits reduce GPU power consumption by 81 percent; GPU idle time drops from 60 percent, typically, to less than one percent. So it claims; cooling and water usage is minimized, and more tokens per second are served from the same stack.
PRISM works across any GPU accelerator platform. AMD, which has just committed to invest £2 billion in the UK over five years, is also providing technical collaboration to run large‑scale network models relevant to frontier‑scale AI systems. The proposition has gone from R&D to production in three years. James Regan, the company’s chief executive, said: “A year ago, we were proving the physics; today, we’re proving the business.” Madhu Rangarajan, corporate vice president at AMD, said the solution is a “fundamentally different way to connect accelerators at scale”.
There are pockets of AI innovation elsewhere in the UK, too. Britain’s first fully sovereign frontier AI model, Lumen Sovereign, is being prepped. The project, backed by the government’s Sovereign AI programme; local AI startup Cosine has been picked to lead it, and has rallied key UK or UK-based firms to help with the design work – including BAE, Babcock, BT, Era4, HSBC, Leonardo, Lloyds, LSEG, NatWest, PwC, Telefónica, Thales, and The Alan Turing Institute. The open-weight model Lumen Sovereign model will use Isambard-AI. It is set for release in late 2026.
A statement explains: “Rather than pre-training from scratch, Cosine’s focus is on rapid, high-impact capability scaling: continual pre-training and expert expansion to the 256k-context sparse mixture-of-experts (MoE) architecture, targeting a long-horizon agentic coding model within a 500,000 H100 GPU-hour compute budget; enhancing the MoE architecture for highly specialised, domain-specific reasoning; lightweight instruction recovery so the model maintains state-of-the-art proficiency in coding, tool use, and complex agentic workflows.”
The coalition will establish a framework for the “co-definition” of the model’s use-case parameters, security protocols, and governance standards. About a third of the total compute time in training has been to optimise performance on complex software engineering problems; two thirds have focused on mission-critical use cases. Lumen Sovereign will be capable of deployment entirely within a customer’s own infrastructure – “including fully air-gapped environments, with absolutely no external data transfer required”. The sovereignty play is explicit. “We are taking a massive step toward technological independence, ensuring that Britain remains an AI maker, not just an AI taker.”
Cosine says: “Until now, the UK’s defence, financial, and public sector institutions have faced a difficult choice: either fall behind in AI adoption or rely on models trained and operated overseas, which can ultimately create unacceptable operational and sovereignty risks. A sovereign LLM provides strategic advantages beyond avoiding dependency, including the ability to operate in high-security, air-gapped environments for complete domestic data control. This assurance-by-design is often non-negotiable in regulated sectors… Furthermore, it offers a competitive advantage by providing crucial cost protection against vendor lock-in and price escalations from expensive foreign-built APIs.”
There is more, too – in the related-critical field of Industry 4.0. A UK startup called PhysicsX – developing an AI digital-twin model for industrial design; backed by Nvidia and Siemens – has just announced an oversubscribed $300 million Series C funding round – with a final valuation of $2.4 billion. The round was led by state-owned Singapore investment firm Temasek, with participation also from M&G Investments, Intrepid Growth Partners, Applied Materials, Atomico, General Catalyst, July Fund, NGP, and Radius – plus from Nvidia and Siemens again.
PhysicsX’s AI digital-twin solution is used by enterprises in the manufacturing and utilities sectors, in the aerospace, defense, energy, semiconductors, automotive, and data centre markets. It says its AI models “predict physical behaviour in seconds rather than hours or days, enabling engineering teams to evaluate orders of magnitude more design variants and carry physics insight across the full product lifecycle”. It says GPU economics is mature enough to support “physics AI” at production scale.
Jacomo Corbo, co-founder and chief executive at the firm, said: “Almost every hard problem in the physical economy – better aircraft, better chips, better engines, better energy systems – comes down to how fast and well engineers and machine operators work through the underlying physics. For decades, that has been the binding constraint. Physics AI removes it. We give engineers the ability to explore thousands of designs where they once managed a handful, in seconds rather than weeks, across the most demanding industries in the world. We are also enabling more reliable, more efficient, and altogether new ways of doing engineering, manufacturing, and production.”
The company has doubled year-over-year revenue, tripled booked revenue, and doubled its customer count over the past year – it says. It employs 300 people, double the headcount of a year ago; it has offices in London and New York, and an expanding presence in the Bay Area and Singapore.
