Skip to content
You can now search across every topic, entity and event.What's new
UK Startups and Innovation
14JUN

Kendall names UK chip five at RUSI

4 min read
16:35UTC

Liz Kendall used a Tuesday 28 April address at RUSI to name Fractile, Olix, Lumai, Optalysys and Salience Labs as British supply-chain candidates, and pre-announced an AI Hardware Plan for London Tech Week in June.

← Back to London startup raises Britain's own AI modelJump to analysis ↓
TechnologyDeveloping
Key takeaway

Kendall named five British chip startups at RUSI on 28 April, pre-committing to a June plan.

Liz Kendall, Secretary of State for Science, Innovation and Technology, delivered an address at the Royal United Services Institute on Tuesday 28 April 2026 framing the Sovereign AI Unit and the Advanced Research and Invention Agency compute programmes as instruments of sovereign control 1. She named Fractile, Olix, Lumai, Optalysys and Salience Labs alongside Arm as the British supply-chain candidates an AI Hardware Plan is designed to underwrite, and pre-announced the plan's launch at London Tech Week in June 2026.

Kendall told the audience the SAIU was about "securing control and leverage over AI" and described the strategy as "our bet on Britain" 2. None of the five companies had been named in a ministerial speech before, despite all operating for at least three years. Public naming by a Cabinet minister moves a private company onto procurement radars overnight; for hardware companies dependent on long government purchase commitments, that visibility carries the weight of capital.

Each name works a different layer of the AI silicon stack. Fractile builds in-memory inference accelerators that compute alongside the weights rather than fetching them from separate memory; Olix designs edge AI silicon for inference at the network periphery; Lumai builds optical AI compute; Optalysys uses silicon photonics for matrix multiplication; Salience Labs runs hybrid photonic-electronic chips out of an Oxford spinout. Backing all five at meaningful scale would commit Britain to a portfolio of architectural directions rather than a single national champion.

Kendall left the AI Hardware Plan's instrument open. Procurement and equity pull in different directions: the SAIU writes equity cheques into model-training labs, while a hardware plan that reaches Fractile or Lumai at scale will need to underwrite a fab line or a multi-year purchase commitment, not a Series A. London's saturated AI datacentre grid is the structural pressure pulling the timetable forward; domestic compute that runs on domestic silicon is what the AI Growth Zones in Scotland and the north are meant to host.

The pre-announcement also raises expectations against a six-week deadline. A plan that arrives at London Tech Week without equity or procurement commitments will read as soft after the rhetoric. Kendall has named five companies and a date; the framework has to deliver an instrument that reaches them at industrial scale or be read as speech-craft.

Deep Analysis

In plain English

Most AI chips come from a small number of American companies, chiefly Nvidia. The UK currently builds almost no AI hardware of its own. Liz Kendall, the Secretary of State responsible for science and technology, used a speech at a defence think tank to name five British companies that are trying to change that. Each company is working on a different way to make computers that run AI systems more efficiently, particularly for uses where you cannot rely on a data centre in another country. Fractile puts memory and computing on the same chip. Lumai and Optalysys use light instead of electricity to perform calculations. Salience Labs combines optical and electronic approaches. Olix makes chips for devices at the edge of a network, away from large data centres. Naming them in a ministerial speech does not hand them any money directly. But it puts them on procurement radar screens at government departments and defence companies, which can be worth more than a grant if it leads to a multi-year purchase contract.

Deep Analysis
Root Causes

The five named hardware startups exist because two independent technical bets converged on a cost inflection point in 2023-24.

The optical compute bets (Lumai, Optalysys, Salience Labs) all rest on the same underlying constraint: GPU-based inference consumes power at a rate that makes large-scale deployment at the network edge economically unviable without grid-level infrastructure. Silicon photonics and hybrid photonic-electronic chips offer 10-100x improvements in operations-per-watt for matrix multiplication, which is the dominant computation in transformer inference.

The in-memory inference bet (Fractile) addresses a different bottleneck: modern large models spend most of their inference time fetching weights from separate memory banks rather than computing on them. Fractile's architecture collapses the fetch-compute boundary, cutting latency by placing computation inside the memory structure itself.

Both problems are most acute in sovereign defence and critical infrastructure applications, which cannot rely on cloud connectivity. Kendall speaking at RUSI (the Royal United Services Institute) was not coincidental: the customers for these hardware startups at scale are not consumers but government departments and defence primes.

What could happen next?
  • Opportunity

    Defence primes, including BAE Systems and Rolls-Royce, will now run competitive market assessments against the five named companies to comply with their own procurement rules. This accelerates due diligence that previously required a named government programme to trigger.

    Immediate · 0.8
  • Risk

    If the AI Hardware Plan arrives at London Tech Week in June as a strategy document without capital instruments, the five named companies face reputational exposure: they have been publicly positioned as national champions without being given the tools to succeed at that scale.

    Short term · 0.68
  • Precedent

    A government hardware plan that backs five different architectural approaches simultaneously represents a portfolio bet rather than a national champion pick. This diversification model, if it persists, changes how UK hardware startups pitch: a stronger case for complementary differentiation, not winner-take-all positioning.

    Medium term · 0.7
Sources:DSIT / GOV.UK·DSIT / GOV.UK
Mentions:Liz Kendall →RUSI →DSIT →Arm →Sovereign AI Unit →AI Research Resource →AI Growth Zones →Innovate UK →Nvidia →Advanced Research and Invention Agency →Fractile →Olix →Lumai →Optalysys →Salience Labs →AI Hardware Plan →London Tech Week →
First Reported In

Update #3 · SAIU rides $1.1bn Ineffable seed; hardware looms

DSIT / GOV.UK· 1 May 2026
Read original →
Causes and effects
Caused by
Slough saturates, AI datacentres head north
London datacentre grid saturation is the structural pressure making domestic AI hardware sovereignty politically urgent; Kendall's RUSI speech names hardware startups to address it.
Occurred 20 Apr 2026
Read story →
This Event
Kendall names UK chip five at RUSI
A Cabinet minister naming five private chip companies in a sovereignty frame moves them onto procurement radars overnight and creates a six-week political deadline.
Different Perspectives
European VC (Atomico, Plural, Highland Europe as PhysicsX / Lumen adjacents)
European VC (Atomico, Plural, Highland Europe as PhysicsX / Lumen adjacents)
European growth funds have backed three of the week's largest UK rounds via follow-on positions and co-investments; the PhysicsX cap table includes Atomico (European-domiciled, Skype-founded) and Siemens (German industrial), both returning investors who view UK physical-AI as a supply-chain multiplier across Continental manufacturing. European LP capital is filling the growth tier UK state vehicles have not yet reached.
UK regulated-industry coalition (Lloyds, BAE Systems, LSEG via Lumen Sovereign)
UK regulated-industry coalition (Lloyds, BAE Systems, LSEG via Lumen Sovereign)
Thirteen of Britain's most heavily regulated companies backed Cosine not as a philanthropic gesture but to acquire a data-compliant AI tool that replaces costly US API alternatives; each partner provides proprietary data in exchange for early access. Their participation signals that regulated incumbents, not venture funds, may be the structural customer base that sustains the UK's sovereign model tier.
US growth investors (General Catalyst, Intrepid Growth Partners)
US growth investors (General Catalyst, Intrepid Growth Partners)
US and allied growth investors followed Temasek into PhysicsX's Series C; General Catalyst also returned in the round after backing Geordie the previous week. The absence of any US-led domestic-capital equivalent is a structural reading: American funds enter at growth stage where returns are clearest, ceding seed and Series A economics to UK vehicles that are themselves contracting.
Temasek (Singapore sovereign fund)
Temasek (Singapore sovereign fund)
Temasek led PhysicsX's $300m Series C, its second major UK deep-tech cheque in six weeks after co-investing in Isomorphic's Series B with the SAIU; its thesis runs through Southeast Asian advanced-manufacturing adjacencies, not bilateral UK policy. Singapore's sovereign capital is now the default lead for British scale-ups above £200m that fall outside the BBB's priority sectors.
UK Government (DSIT / Liz Kendall)
UK Government (DSIT / Liz Kendall)
DSIT published its first sector scorecard on 10 June setting a £8.3bn 2025 baseline, and the Sovereign AI Unit's compute allocation enabled Cosine's Lumen Sovereign launch. The scorecard's own barbell figure, more capital in fewer rounds, exposes the policy gap DSIT has not yet addressed: no instrument currently leads venture rounds in industrial AI simulation sectors.
Spanish state finance (COFIDES, CDTI)
Spanish state finance (COFIDES, CDTI)
Spain's COFIDES and CDTI have co-invested alongside UK deep-tech rounds in prior cycles and track the British Business Bank's direct-investment activity as a benchmark for state-capital deployment in innovation. BBB's two direct co-investments in one week set a pace reference for Iberian equivalents.