UK AI funding: what your business can actually get

There is a large amount of UK public money attached to AI, and most of it is not available to a business that is adopting AI. It is available to businesses that are building something, and the distinction decides which of the routes below you can use.

That single misunderstanding wastes more time than any other in this area. A company that has rolled out an AI tool across its customer service team has done something valuable and commercially sensible, and it has almost certainly not done anything that qualifies for a grant or for R&D tax relief.

This page covers the three routes that exist, what is open as of 15 August 2026, and how to tell in about two minutes which one applies to you.

The hard rules

Five things that determine everything else.

1. This is not tax or funding advice. It is published information with every claim linked to the funder’s own page, so you can check it. Deadlines move and competitions close early when oversubscribed. Verify before you act.
2. Buying and using AI is not R&D. Neither is configuring it, training staff on it, or integrating a vendor’s API in the documented way. R&D means seeking an advance in science or technology that a competent professional in the field would recognise.
3. Grants are competitive, slow, and match-funded. Expect months from application to money, and expect to pay a share yourself. They are not a cash-flow solution.
4. Not all support is money. The single most accessible AI resource in the UK right now is an allocation of supercomputer time, and it costs nothing.
5. Deadlines here are real dates and times, checked today. A closed competition quoted as open is the most damaging error a page like this can make, so if you are reading this well after the review date at the top, re-check the link before you plan around it.

What are the three routes?

They serve different things and are not alternatives to each other. A company can plausibly use all three in the same year for different purposes.

Answer: grants fund work you have not done yet, compute lets you do work you could not otherwise afford, and tax relief reimburses work you already did.

Which UK AI grants are open right now?

Verified against the funder’s own pages on 15 August 2026.

Farming Futures R&D fund: automation and robotics, round 2

Innovate UK is offering a share of up to £20m for automation and robotics in agriculture, horticulture and forestry.

• Who: UK registered businesses. Collaborations only, and the lead must be a UK registered business.
• Opened: 3 August 2026, 09:30
• Closes: 30 September 2026, 11:00

The collaboration requirement is the part that catches people. A single company cannot apply alone, so if you are interested you need a partner lined up before you start writing, not after.

Next generation AI: explainable AI

UKRI and EPSRC have opened a £12.05m call for research into making AI systems explainable.

• Who: UK research organisations eligible for UKRI funding. Applications are particularly welcomed from early and mid-career researchers.
• Maximum award: £602,500 at full economic cost, of which UKRI funds 80%, or £482,000
• Opened: 11 August 2026, 09:00
• Closes: 20 October 2026, 16:00
• Projects start 1 February 2027 and run up to 24 months

This one is not for companies. It is listed here because it is the largest open UK AI call and it is routinely mis-shared as a business opportunity. If you are a company with an explainability problem, the route in is a university partner, not an application.

Answer: if you are a business, Farming Futures is the live one, and only if you are in agriculture, horticulture or forestry and have a collaborator.

Is there anything open to any business, at any time?

Yes, and it is the least known thing on this page.

UKRI provides access to EuroHPC supercomputers for AI development, through the European High-Performance Computing Joint Undertaking.

• Status: open, with no closing date. Not a competition round you can miss.
• Aimed at startups, SMEs, industry and the public sector, as well as researchers.
• Focused on AI and data-intensive work, specifically foundation models and generative AI such as large language models.
• No funding is provided. Awards are allocations of compute time.

That last line is why it gets overlooked, and why it is undervalued. A company that cannot justify buying GPU capacity to test whether a model can be trained or fine-tuned for its problem can apply for the compute instead of the cash. For an early-stage team the compute is frequently the harder thing to obtain.

Answer: if your constraint is compute rather than money, this is open today and has no deadline to miss.

Does using AI qualify for R&D tax relief?

Usually not, and this is where most of the wasted effort in UK AI funding goes.

HMRC’s guidance is explicit: to qualify, a project must seek an advance in a field of science or technology. Only companies chargeable to UK Corporation Tax can claim.

The test is whether “a competent professional working in that field would recognise the advance as an appreciable improvement”. HMRC defines appreciable improvement as genuine and non-trivial, explicitly ruling out “a minor or routine upgrade”.

One change worth knowing: from 1 April 2023, mathematical advances can be treated as science for these purposes, whether or not they represent the nature and behaviour of the physical universe. That matters for AI work, where the advance is often algorithmic rather than physical.

What that means in practice

What you did	Likely position
Bought licences and rolled a tool out	Not R&D
Configured a vendor product, wrote prompts, trained staff	Not R&D
Integrated an API in the documented way	Not R&D
Attempted something a competent professional would say was not readily deducible, and could have failed	Possibly R&D

The word doing the work is uncertainty. If the outcome was known to be achievable by someone skilled in the field, adapting an existing solution is not an advance no matter how much effort it took.

Answer: if a competent professional in your field could have told you at the start that it would work, it is not R&D.

Which route fits my situation?

A small business adopting AI tools

None of the above, and that is not a failure. You are buying software, which is an ordinary business cost. The genuine support here is business-advice provision and skills funding rather than innovation money, and pursuing a grant for an adoption project usually costs more in time than it returns.

A startup building an AI product

Compute access first, because it is open now and has no deadline. Then watch Innovate UK competitions in your sector rather than AI competitions generally: the Farming Futures round is an AI competition that is not labelled as one, and sector rounds have smaller applicant pools than open AI calls.

A company doing genuinely novel technical work

R&D tax relief is the most reliable route, because it does not depend on winning anything. Keep contemporaneous records of the technological uncertainty: what you did not know at the start, what you tried, and what failed. That record is the claim.

A university or research organisation

The explainable AI call, and the standing UKRI opportunity feed rather than this page, since research calls turn over faster than any article can track.

Answer: most businesses reading this are in the first group, and the honest recommendation is to stop looking for grant money and spend the time on adoption instead.

What people get wrong

Assuming AI grants are for AI users. Almost every open call funds the development of new capability. Adoption is supported through advice and skills programmes, not innovation competitions.

Missing sector calls. The Farming Futures round is £20m for automation and robotics and does not have “AI” in its title. Searching for AI funding will not surface it. Watch your sector, not the technology.

Treating a grant as cash flow. Competition grants are match funded and take months. If the project only works with the grant, it is usually not fundable.

Claiming R&D relief for integration work. HMRC’s test is an advance in science or technology, and an adviser who tells you that any AI project qualifies is describing their fee, not the legislation.

Ignoring compute. The EuroHPC route provides the resource that is often the actual constraint, costs nothing, and has no closing date. It is the most available thing on this page and the least applied for.

What should I do now?

1. Decide which of the three routes you are in, using the two-minute test: are you building something novel, constrained by compute, or claiming for work already done?
2. If it is compute, apply now. No deadline, open to SMEs.
3. If it is a grant, find the sector call rather than the AI call, and line up a collaborator before writing anything.
4. If it is R&D relief, start the record today. The claim is only as good as the contemporaneous evidence of what was uncertain.
5. Subscribe to the UKRI funding opportunities feed rather than relying on any article, including this one. It updates the day a competition opens.

If none of the three applies, that is a real answer and a useful one. The time saved not chasing a grant you cannot win is worth more than the grant.