A practical guide to the £100 million Sovereign AI R&D Procurement Scheme, its four live challenges, application process, and what UK SMEs need to know before applying.
Sovereign AI has become one of the most talked about phrases in UK business news this month, and for good reason. On 31 August 2026, Chancellor John Healey announced the launch of the first competitions under the £100 million Sovereign AI R&D Procurement Scheme, unveiled at the G20 Finance Ministers meeting in North Carolina. The announcement was published jointly by HM Treasury and the Cabinet Office, alongside AI Minister Kanishka Narayan. For smaller British AI companies who have spent years watching public sector contracts go to the same handful of large incumbents, this is one of the more interesting openings to come out of government in a while.
What the Sovereign AI R&D Procurement Scheme actually is
The scheme sits under the government's wider Sovereign AI programme, which is backed by up to £500 million overall. Its stated purpose is to help promising British AI companies start in the UK, grow here, and eventually compete internationally, while directing that innovation toward genuine public sector problems rather than leaving it to sit in a lab. What makes this particular scheme worth paying attention to if you run a smaller business is how directly it addresses a complaint that founders have been making for years. Public procurement has historically favoured companies with a long track record, strong cash reserves and high turnover, which quietly shuts out newer firms with genuinely good technology. This scheme removes those barriers outright, with no minimum turnover, net assets or cash reserve requirement to apply, and upfront payments available where appropriate so smaller teams are not left funding delivery out of their own pocket before they get paid.
Just as importantly, companies that win a contract keep the intellectual property they create, both the background IP they already held and any new foreground IP developed during the project. Government only takes a licence to use the outputs, data and results for its own public sector purposes, rather than any claim of ownership over the technology itself. That means a solution built for one NHS trust or one government department is not locked away afterwards. Suppliers are free to take it further, and are actively encouraged to commercialise it, selling the same technology elsewhere in the UK or abroad once the initial pilot has proven itself.
The timing of all this is not really an accident. On the same day, the Chancellor also confirmed that the UK plans to open its AI Economics Institute up to international cooperation with G7 countries, focused on sharing evidence on AI's economic impact. Taken together, the two announcements suggest government sees AI policy and AI procurement as two sides of the same coin, with smaller domestic suppliers meant to be part of that picture rather than an afterthought.
The four challenges open right now
Four competitions have launched as the first phase of the scheme:
- NHS productivity, with the Department of Health and Social Care: AI systems that automate workflows, coordinate patient care and support clinical decision making, in line with the ambitions of the NHS 10 Year Health Plan.
- Compute efficiency, led by the Department for Business, Innovation, Science and Trade alongside ARIA's Scaling Inference Lab: technology that makes AI computing infrastructure cheaper and more efficient to run at scale.
- Defence integration, with the Ministry of Defence: securely connecting data and frontier AI capabilities across defence systems.
- Agent security and resilience testing, with the National Cyber Security Centre: helping organisations understand and manage the risks that come with increasingly capable AI agents.
Contract values across these competitions range from £250,000 up to £10 million, with most expected to land somewhere between £1 million and £3 million. The scheme itself runs until March 2030, so this is not a one off funding round but the start of a longer pipeline, with further challenges expected to follow as the programme develops.
How the assessment process actually works
The process starts with a short Expression of Interest form rather than a full tender submission, and it is assessed on a simple pass or fail basis, alongside a due diligence check, to confirm eligibility and fit with the scheme's aims. Government aims to turn these around within two weeks, and it is worth submitting an EOI at least two weeks before whichever competition deadline you are targeting, since anything submitted later may not be reviewed in time to be considered for that batch. Pass that stage and you are invited to join the scheme's Approved Supplier List, before moving into a more detailed assessment of your full proposal, where it is scored out of five against a set of published criteria, and you need to score at least four in every single criterion to reach the final decision stage. If a proposal does not make it through, applicants are welcome to try again in a future batch, though the guidance is clear that any resubmission needs to either be materially different from the original application or clearly address whatever reasons were given for not progressing the first time around.
What suppliers need to know before applying
Applications go through an Expression of Interest form on the Sovereign AI website, and the process has been deliberately kept lighter than a typical government tender, with no exclusivity requirements and prompt payment built in. A supplier briefing session took place on 7 September 2026 to walk applicants through the process, and the deadline for proposals in this first competition batch is 1 October 2026, with two further batches due to close on 1 December 2026 and 1 February 2027, so anyone considering applying should aim for the earliest slot they can realistically hit. Applicants need to be UK registered, the project itself needs to be based in the UK, and proposals must fall within the stated funding range. It is worth reading the full competition guidance before starting an application, since eligibility and assessment criteria differ slightly across the four challenges, and the original notice on Find a Tender is also worth a look for background on how the scheme was shaped.
For smaller AI companies who have felt locked out of government work in the past, this scheme was built with exactly that frustration in mind. Whether or not AI is your core product, it is worth watching how these first competitions play out, since the model being tested here could well shape how government buys innovative technology from smaller suppliers for years to come.
More than a contract: the wider Sovereign AI toolkit
The R&D Procurement Scheme is only one part of what government is currently offering AI companies. Sitting alongside it is the AI Research Resource, which gives a small number of strategically significant companies access to serious compute power, up to a million GPU hours each, on national supercomputers including Isambard and DAWN, though that application round is currently closed. There is also a Strategic Assets Programme aimed at funding things individual startups usually cannot build alone, such as high quality AI datasets and autonomous lab infrastructure, and its initial round has closed as well. The one still open right now is the Visa Reimbursement Scheme, which covers costs like Certificate of Sponsorship fees and the Immigration Skills Charge when a Sovereign AI backed company needs to bring in specialist talent from abroad. None of these require a company to have already won a procurement contract, so it is worth keeping an eye on the Sovereign AI website in case the compute and assets rounds reopen.
If you are actively bidding for public sector work, it is also worth keeping an eye on Supplierverse for related opportunities as more challenges under the scheme are expected to open in future phases.