10JUN

Onwurah: DSIT has no coherent strategy

3 min read

10:31UTC

Chi Onwurah, who chairs the Commons Science, Innovation and Technology Committee, published correspondence with DSIT minister Kanishka Narayan in which she said his sovereignty letter fails to set out a coherent strategy.

← Back to Sovereignty law adopted; $40bn US chip buy Jump to analysis ↓

TechnologyDeveloping

Key takeaway

The SIT Committee chair has told DSIT in print that its sovereignty strategy is not yet a strategy.

Chi Onwurah, the Member of Parliament for Newcastle upon Tyne Central and chair of the Commons Science, Innovation and Technology Committee, published correspondence with Kanishka Narayan, minister at the Department for Science, Innovation and Technology (DSIT), in which she writes that his sovereignty letter "fails to set out a coherent strategy for achieving technology sovereignty" ¹. Narayan pointed to the £500 million Sovereign AI Unit (SAIU), launched in April , and a proposed "advanced market commitment" for AI hardware as the government's strategy.

The SAIU's first cohort, named on 16 April, was exclusively infrastructure-layer ; no second cohort has been announced. Narayan's "advanced market commitment" is, on Computer Weekly's reading, a phrase rather than a published policy instrument. The Open Rights Group's national security report, which put the UK's annual cloud lock-in cost to US hyperscalers at £500 million , already framed the sovereignty problem as one of demand-side capture, not capital. The SAIU's £500 million sits on the supply side.

The immediate political pressure is now on DSIT to expand beyond the seven first-cohort firms. The committee chair has put the minister on record. Across the Channel, the comparable European pressure is the seven-CEO Handelsblatt op-ed urging Brussels to deregulate, and OpenAI's decision to pause Stargate UK on energy costs while it signed an 88,500 square foot lease at King's Cross. Onwurah's letter does not name those data points. The next SAIU cohort, if it lands, will determine whether the UK's sovereignty strategy includes anything above the chip and the data-centre.

Deep Analysis

In plain English

Chi Onwurah is a Labour MP who chairs the committee in the UK Parliament that scrutinises the government's science and technology policy. She wrote to Kanishka Narayan, the minister responsible, saying the government's plan for 'technology sovereignty' (being less dependent on US and Chinese tech companies) does not actually explain how it will be achieved. Narayan pointed to the £500m Sovereign AI Unit, a fund set up in April 2026, and a proposed financial mechanism called an 'advanced market commitment' for AI hardware. The exchange matters because the UK is outside the EU and cannot benefit from the EU's €20bn chip investment or its procurement preferences for European tech. The £500m fund is roughly 40 times smaller than the EU's equivalent spending over the same period, so critics argue it is a political announcement rather than a genuine strategy.

Deep Analysis

Root Causes

The UK's post-Brexit position means it operates outside the EU's Digital Single Market, the EU AI Act framework, and the Chips Act investment pool simultaneously.

This creates a three-gap problem: no access to the EU's procurement-preference market, no access to EU-funded chip production, and no coverage by the EU's GPAI (general-purpose AI) enforcement regime that at least provides regulatory clarity. DSIT is trying to fill all three gaps with a single £500m unit that is smaller than the EU's Digital Europe Programme quarterly allocation.

The 'advanced market commitment' mechanism for AI hardware that Narayan cited has no published design. Advanced market commitments work in pharmaceuticals (GAVI vaccine guarantee) because the product specification is fixed and the buyer is a government. AI hardware is neither: the compute stack changes every 18 months and UK government procurement is not a large enough market to anchor an AMC at meaningful scale without US or EU co-investment.

What could happen next?

Risk
If the Sovereign AI Unit cannot publish measurable output metrics before the next Select Committee hearing, Onwurah's framing that DSIT lacks a coherent strategy becomes the dominant parliamentary narrative, reducing ministerial authority to resist further scrutiny or to attract private co-investment.
Consequence
The published correspondence becomes a permanent part of the parliamentary record and will be cited by opposition MPs in any future AI procurement controversy, regardless of whether DSIT subsequently produces a more detailed strategy document.

Source Landscape

This story draws on neutral-leaning sources from Belgium

Belgium

Primary parallel: The UK's Industrial Strategy Challenge Fund (ISCF), launched in 2017 with £4.7bn and managed by UK Research and Innovation (UKRI), was subject to near-identical select committee scrutiny by 2019: ministers described it as transformative, committees found no published metrics for measuring industrial impact, and BEIS could not name a single domestic industry that reached commercial scale under its funding.

The ISCF was absorbed into the Innovate UK grant pipeline after the 2020 Spending Review.

Counter-parallel: The UK's Advanced Research and Invention Agency (ARIA), launched in 2022 with £800m, escaped the ISCF's fate by publishing explicit mission-oriented research goals before the first public scrutiny hearing. ARIA gave parliamentary committees a measurable target; DSIT has not yet done the equivalent for the Sovereign AI Unit.

Consensus view: The Institute for Government's Jill Rutter assesses the Onwurah-Narayan exchange as a standard parliamentary accountability mechanism producing an unusually precise policy gap: DSIT's strategy letter cites the £500m Sovereign AI Unit but provides no procurement pipeline, no performance metrics for the 'advanced market commitment' mechanism, and no published definition of what 'sovereign' means for AI hardware in the UK context.

Select committee chairs can embarrass ministers but cannot compel strategy revision; the substantive result depends on whether DSIT produces a written definition before the next committee hearing.

Counter-view: The Open Rights Group's Mariano delli Santi argues the £500m Sovereign AI Unit itself has the wrong governance structure for a sovereignty mandate: it is chaired by a venture capital investor (James Wise of Balderton Capital) whose fund holds positions in companies that could bid for the unit's contracts, creating a structural conflict of interest that Narayan's letter does not address and Onwurah's correspondence did not pursue.

Key tension: Whether the Sovereign AI Unit's output is measurable industrial policy or a reputational vehicle for the government to claim tech sovereignty without binding procurement commitments.

Sources:European Commission

Mentions:King's Cross →OpenAI →Parliament →Open Rights Group →Institute for Government →Handelsblatt →Innovate UK →James Wise →Balderton Capital →House Science Committee →Chi Onwurah →Kanishka Narayan →DSIT →Sovereign AI Unit →Stargate →UKRI →Advanced Research and Invention Agency →

First Reported In

Update #4 · CISPE moves first; Brussels misses again

European Commission· 7 May 2026

Read original →

Causes and effects

Caused by

Kendall names seven infrastructure bets for £500m Sovereign AI Unit

Onwurah's challenge directly references the SAIU's infrastructure-only first cohort as insufficient evidence of a coherent sovereignty strategy.

Occurred 16 Apr 2026

Read story →

This Event

Onwurah: DSIT has no coherent strategy

Three weeks after the £500m Sovereign AI Unit named its first cohort, the parliamentary committee that scrutinises it is on the public record saying DSIT cannot describe its own approach.

Different Perspectives

European cloud and open-source industry

European cloud providers gain a binding procurement mandate from CADA, confirmed by Gartner's $12.6bn sovereign-cloud figure for 2026. The $40bn Pax Silica commitment signals Brussels will not extend sovereignty discipline to the silicon layer, and the missing €350m Sovereign Tech Fund leaves open-source maintenance infrastructure unfunded beneath those same clouds.

United Kingdom

Science Secretary Kendall's £1.1bn Hardware Plan on 8 June chose demand-side instruments, advancing £150m to British chip startups via the British Business Bank, where Brussels chose supply-side alliance membership. Britain joined Pax Silica before the EU and has no collective EU procurement leverage; the Hardware Plan is the bilateral answer to the same silicon gap.

United States

Pax Silica, a State Department initiative launched in December 2025, secured EU membership the same afternoon Brussels adopted its cloud sovereignty law. Ambassador Puzder had named CADA a red line against the EU-US trade framework; the narrowed CADA scope and the $40bn chip commitment together represent the settlement Washington sought.

France

France was the only EU state to oppose Pax Silica accession at COREPER on 3 June, asking the Commission to clarify the Council's steering role inside the alliance. Paris backed CADA and hosts Mistral AI; a $40bn US-chip commitment contractually narrows the commercial space for the sovereign AI model that France is trying to scale.

European Commission

Von der Leyen framed CADA on 3 June as keeping 'most of our market open to like-minded partners', and the Commission's EVP Virkkunen simultaneously required majority-European ownership for the €4.12bn AI Gigafactories call. Brussels is managing rather than resolving the silicon dependency by asserting regulatory control at the cloud layer while formalising the chip relationship through Pax Silica.

European Central Bank

The ECB's digital euro pilot drew more than 50 PSP applications and is naming 10 to 30 participants in July, advancing on its own monetary mandate without requiring a Commission act. Its trajectory this week is the inverse of CAIDA's: the sovereignty instrument that restricts no US firm is the only one keeping its published calendar.