Official UK procurement notice
Request for Information: Sovereign AI R&D Procurement
Authority: Department for Science, Innovation & Technology (Central government)
What is being bought
Published requirement descriptionWe want to build government procurement into our offer to support AI startups. By acting as an early customer, government can help to validate novel AI capabilities and de-risk investment for the wider market. We know the difference government contracts can make for startups by providing revenue, credibility, and leveraging private investment. The benefits for applicants could include:
Government contracts to develop prototype AI capabilities, providing revenue and validation.
The opportunity to establish close relationships with government departments and understanding of market demand (such as MOD, DHSC, DSIT, NCSC and others).
We want to design our procurement in a way that supports high-potential AI startups that are developing sovereign AI capabilities. If you are developing something which aligns with our Sovereign AI Focus Areas below, then we want to hear from you. The information gathered will help to design the scheme and ensure it works well for companies.
We are considering a competitive scheme where companies can apply to win government contracts to develop sovereign AI capabilities and address public-interest challenges.
We are considering awarding contracts up to £5 million per project, with a duration of 12 to 24 months, flexible depending on the scope of the project.
To be eligible, projects would have to align with one of the five Sovereign AI focus areas. There will also be further challenges set by government departments to develop specific capabilities which address government priorities in areas such as (but not limited to): advancing scientific discovery, health and social care, national security and defence, cybersecurity, transport, energy & net zero, public service delivery.
The output of the contracts will be a demonstrator stage AI capability which proves proof-of-concept (not a final commercial solution), with a clear route to commercialisation with public or private sector customers.
Companies would retain ownership of all background IP and foreground IP created during the project. Companies are free to exploit the IP commercially or sell it to other customers. HMG would retain usage rights to the foreground IP developed during the project. This is to ensure adherence to procurement law and knowledge retention. Government is not intended to capture any economic value from the IP.
This is a time limited market engagement exercise: please submit your response by 16th May 2026.
CPV: IT services: consulting, software development, Internet and support
The lotProcurement structure
single lot · no lot value publishedPipeline intelligence · planning stage: how this market runs before the tender is published
How Department for Science, Innovation & Technology buys IT services: consulting, software, internet (CPV 72), and who usually turns up
Department for Science, Innovation & Technology’s published record next to the market for the same work. Frameworks and DPS are left out: they draw different bid counts. Whole-market view: IT services: consulting, software, internet (CPV 72) →
How the contest usually goes
Winner shares use published awards; bid and price charts use contested lots.
i
How often small and medium firms win.
Share of awards where the winner’s size is known.
The counts under this question show how many awards have a known supplier size.
Small firms have won here. Neither number says whether you qualify; it describes who has been winning. Small firms win more of this buyer’s work than the market’s: 83% against 37%, n=12.
i
How often local firms win similar work.
Winners with a registered address in the delivery region, out of awards with a known address.
Local means the winner is registered in the region where that award was delivered.
A local branch does not count if the registered address is elsewhere.
No winner locations published for this authority. Local means a registered address in the region where that award was delivered.
i
How long buyers here take to decide after bids close.
Middle value in days from the bid deadline to the award date, on competitions that published both.
Only decided tenders count, so slow ones still waiting are not in.
Half of these awards have an award date within 21 days of the bid deadline.
i
How often work is awarded with no recorded rival.
Direct awards and tenders where every published lot count shows one bid.
The market share is a minimum: some awards have no bid count.
Thin markets produce this too; it is not proof of a tailored specification.
Direct awards or one recorded bid are common here. Check the route before spending time on a bid.
How many bids did past lots attract?
i
How many bids past lots received.
Each colour is a bid-count range. Only competed lots with a published count are included.
Past bids do not predict this tender.
1 bid0% / 12%
2–317% / 28%
4–625% / 28%
7+58% / 32%
How did awards compare with estimates?
i
Where winning prices landed against estimates.
Each colour shows awards below, close to or above the published estimate, one lot at a time.
Past prices are a guide, not a bid price.
Below estimate64% / 46%
Within ±2%36% / 37%
Above estimate0% / 16%
Assess each lot i
What past awards tell you about each lot.
See the forecast, this buyer’s earlier suppliers and firms with relevant wins.
No supplier is confirmed as a bidder.
i
What past awards tell you about each lot.
See the forecast, this buyer’s earlier suppliers and firms with relevant wins.
No supplier is confirmed as a bidder.
Open a lot for its forecast, this buyer’s previous supplier and possible competitors, from published award history, not a bid list
Lot #1CPV 72Published estimate£80m
This lot lists more than one class in CPV 72. The comparison uses that group.
i
How many bids this lot may attract.
The forecast uses similar past competitions. The bar shows either that forecast or past bid counts; the line below says which.
Each lot shows its own forecast based on its own category; forecasts for different lots of the same notice are not comparative. The bar shows 791 past lots.
A forecast, not a list of bidders.
1 bid12%
2–328%
4–628%
7+32%
i
Where the award price may land.
The forecast uses similar awards and the buyer’s estimate. The band covers the middle half of likely prices.
No comparable awards for this lot yet.
A statistical guide from past awards, not a price to bid.
Who has supplied this buyer on similar work? i
Who has done similar work for this buyer.
Past contracts are matched by buyer, title and type of work.
A past supplier may not bid again.
i
Who has done similar work for this buyer.
Past contracts are matched by buyer, title and type of work.
A past supplier may not bid again.
No earlier contract of this buyer matches this lot on the published record.
Who could compete for this lot? i
Who could compete for this lot?
Firms with relevant past wins. On a framework, we show firms on that agreement.
None is confirmed as a bidder.
i
Who could compete for this lot?
Firms with relevant past wins. On a framework, we show firms on that agreement.
None is confirmed as a bidder.
Candidate suppliers for this notice have not been worked out yet, so this is not a verdict that nobody fits.
About this data
- Source. Official UK public procurement notices (Find a Tender), reused under the Open Government Licence v3.0. New notices arrive through the day; the figures are rebuilt nightly.
- Two views, kept apart. The authority’s own track record and the wider market for the same kind of work sit side by side and are never mixed. Where the local picture is too thin, a figure widens to a broader market and says which one it used.
- Forecasts are labelled. The price forecast and its range, possible competitors and the earlier-work matches are our estimates from past awards: not published facts, not win chances and not price advice. Official notices do not say who is bidding.
- Every figure shows what it rests on. Not every notice publishes a bid count, a value or a delivery region, so each figure carries the number of records behind it, and one built on few records is marked as such. The on a figure explains it; the counting rules are in the methodology. Spotted something wrong? Tell us.
Procedure terms & legal framework
From the official record- Legal basis
- UKPGA · 2023/54 Procurement Act 2023
- Procurement category
- Services
- Notice status
- Planned procurement
- Procurement threshold
- Above threshold (FTS publication)
- Regulatory regime
- Standard
- Commercial structure
- Stand-alone contract (no framework)
- Competition type
- Competitive procedure
- GPA / WTO covered
- Unspecified in notice
- Recurring procurement
- Unspecified in notice
- First published
- 17 Apr 2026, 11:10 BST
- Last source update
- 17 Apr 2026, 11:10 BST
- FTS notice number
2026/S 000-035079- Latest notice
- Preliminary market engagement notice (UK2)
- Classification (CPV)
- Delivery area
- Official registry OCID
ocds-h6vhtk-068509Find a Tender
How it was run
Official notice- Submission deadline
- —
Timeline
17 Apr 2026 → 3 Jul 2026 · 4 dates- Planning notice published190 days ago17 Apr 2026, 11:10 BST
- Market engagement160 days ago16 May 2026, 23:59 BST
Full details of the engagement process can be found here:
forms.office.com/pages/responsepage.aspx ↗
This is a time limited market engagement exercise: please submit your response by 16th May 2026.
- Tender notice expected114 days ago1 Jul 2026, 23:59 BST
- Contract starts113 days ago3 Jul 2026, 00:00 BST
Notice history
1 event · 1 release on Find a Tender- Preliminary market engagement notice published17 Apr 2026, 11:10 BSTUK2
Commercial outcome & contract awards
Pipeline notice- Awards (0)
- No award published
- Contracts (0)
- No contract published
- Bid statistics
- No aggregate bid statistics published
Contracting authority & participating public bodies
Official documents & specifications (1)
Official links; attachments are not copiedNotices published for this procedure
- Preliminary market engagement notice Preliminary market engagement notice on Find a Tender