How to combine grants and R&D tax relief for AI and deeptech companies

For AI and deeptech companies, the strongest non-dilutive funding model is rarely a straight choice between grants and R&D tax relief. It is a sequence. Grants are usually best suited to a defined work package at the start of a technical programme. R&D tax relief is more useful once qualifying spend is already flowing through payroll, testing, subcontractors and related development costs. In the UK, that distinction matters more now because public support for AI is becoming more targeted, more strategic, and more closely tied to national capability.

The point is not that grants replace equity, or that tax relief replaces grants. Each instrument solves a different financing problem. Grants help fund a specific technical leap that may still look too early, too narrow, or too uncertain for private capital. R&D tax relief then helps recover part of the cost of the wider experimental work once the company is building, testing and iterating at scale. Used together, they can extend runway without forcing founders to give away more equity than they need to.

Having support from an R&D tax and grants consultancy can help businesses navigate the complexity of non-dilutive funding at each stage of the funding journey. For example, FI Group by EPSA brings together in-year R&D tax support, pre-submission reviews, grants roadmapping, and international R&D strategy across more than 20 countries, with a single point of contact backed by local experts. This gives finance leaders greater visibility, peace of mind, and stronger outcomes at group level.

What the UK is actually trying to fund now

The UK’s AI policy is no longer just a growth story. It is becoming part of industrial policy. The AI Opportunities Action Plan, published in January 2025, set out 50 recommendations, and by January 2026 the government said 38 had already been delivered. That sits alongside five designated AI Growth Zones, a Sovereign AI Unit with the next phase launching in April 2026 backed by up to £500 million, and UKRI’s first dedicated AI strategy, backed by £1.6 billion of direct AI investment through the end of the decade.

That matters for founders because it changes the funding environment in two ways. First, it creates more thematic grant opportunities around frontier AI, secure software, connectivity, quantum and strategic infrastructure. Second, it makes it easier to explain why a company should design a funding stack around milestone-led grants plus retrospective tax support, rather than expecting one large equity round to do all the work.

Grants first, R&D tax later

In practice, grants and R&D tax relief support different stages of the same innovation cycle.

Grants are usually most useful when a company needs to prove one thing: feasibility, integration, demonstration, a safety case, a collaboration, a market-specific pilot, or a regulated proof point.

R&D tax relief tends to become more valuable once the business moves into broader technical execution: paying engineers and data scientists, testing new models, refining systems, handling failed iterations, and carrying the cost of experimental development across a wider programme.

That is why the best finance teams do not ask, “Which one should we use?” They ask, “Which one comes first, which costs sit where, and how do we evidence the boundary properly?” That is the more useful question for AI, robotics, quantum, advanced connectivity and regulated software businesses because their development cycles are rarely linear.

What changed after the old SME scheme

This is the rule change that makes the combined model much more workable than it used to be.

Under the old SME R&D scheme, subsidised expenditure could not qualify for SME relief. HMRC’s manual was explicit: if expenditure was subsidised, SME relief was not available for that spend, and if a project received notified state aid, no expenditure on that project could qualify under the SME scheme. That was the source of years of confusion for innovative SMEs that won grants and then found the interaction with tax relief was less generous than expected.

The old large company scheme worked differently. HMRC states that subsidised expenditure could still qualify there, and SMEs sometimes had to fall back into RDEC treatment for spend disallowed under the SME rules because it was subsidised.

For accounting periods beginning on or after 1 April 2024, the old SME and RDEC schemes were replaced by the merged scheme and ERIS. HMRC says the merged scheme and ERIS now replace the previous regimes, that the qualifying expenditure rules are aligned across the two new schemes, that the merged credit rate is 20%, and that ERIS is reserved for loss-making R&D-intensive SMEs with the extra 86% deduction and up to a 14.5% payable credit.

The practical consequence is significant. Founders are no longer dealing with the same old SME-era subsidy trap that made grant-backed work feel separate from a broader R&D tax strategy. The question now is less “grant or tax?” and more “which costs qualify, in which entity, under which post-April 2024 scheme, and how strong is the evidence?” That is a better framework for AI and deeptech companies, where grant-funded work often leads straight into a wider technical roadmap rather than sitting as a one-off experiment.

A better non-dilutive model for AI and deeptech founders

A more complete funding model usually works like this:

1. Use grants to de-risk the first hard milestone

This is where a company needs outside validation as much as money. A grant award tells investors, partners and customers that an external assessor believes the technical step is credible and strategically relevant.

2. Use R&D tax relief to support the broader execution layer

Once the team is spending on developers, researchers, model training, hardware iteration, testing and failed attempts, R&D tax relief becomes more valuable because it supports the wider experimental effort, not just the narrow work package that matched a grant call.

3. Re-forecast after every funding event

Every award changes the funding mix. A grant changes the cost map. A financing round changes the runway calculation. A new collaboration can change which costs sit where and who should claim what.

4. Treat documentation as part of the financing model

For deeptech companies, evidence is not admin. It is part of capital efficiency. Weak technical narratives, vague baselines and sloppy cost allocation do not just create HMRC problems. They make the whole non-dilutive strategy harder to repeat.

That last point is often missed. A robust claim is not only about getting one year’s benefit. It is about creating a repeatable operating rhythm for grants, tax, and later private capital.

Current and recent UK grant routes worth tracking

The live pipeline in April 2026 shows how broad the market has become.

• Artificial intelligence champions: frontier AI phase one is open to UK-registered SMEs, offers up to £3 million, and closes on 29 April 2026. It is aimed at feasibility-stage frontier AI and machine learning technologies.
• Secure software for resilient growth is open with £5 million available and also closes on 29 April 2026. That makes it relevant not only to cybersecurity vendors but also to AI and fintech firms building resilient, trusted platforms.
• UK-Germany Collaborative Innovation for Quantum Technologies 2026 is open with £3 million and closes on 15 April 2026, which is directly relevant to deeptech teams operating in quantum-adjacent or cross-border innovation models.
• Solution Development for Advanced Connectivity Technologies opened on 1 April 2026, closes on 13 May 2026, and carries £25 million of support. That is a reminder that some of the most valuable deeptech funding does not arrive under a label that says “AI”.
• On the recent side, Innovate UK Growth Catalyst, Investor Partnerships Round 2 closed in February 2026 with at least £100 million allocated through investor partners, explicitly targeting growth-oriented start-ups from seed to Series A. It is a good example of where non-dilutive and private capital are now being designed to work together rather than as separate lanes.

The deeper lesson is that AI and deeptech founders should stop screening opportunities only for labels like “AI call” or “deeptech fund”. In 2026, some of the most relevant routes are built around resilience, connectivity, quantum, productivity or strategic sector adoption.

What previous winners tell us about funding progression

The most useful way to read past winners is not as a vanity list. It is as evidence of how public funding can sit early in the capital stack before institutional money arrives.

Monumo won nearly £1 million from Innovate UK in August 2024 for generative AI motor design. Earlier that year it had already raised a £10.5 million seed round. That is a good example of grant funding and early equity reinforcing each other rather than competing.

Quantum Motion was named among recipients in the UK’s NQCC quantum testbeds programme, then moved into a Series B financing supported by Future Fund: Breakthrough and Bosch Ventures. That is a classic deeptech trajectory: public funding around capability and infrastructure, followed by growth capital once the technical platform matures.

Yaspa received an Innovate UK-backed award in April 2024 for safer gambling payments infrastructure, then announced a $12 million investment round in July 2025 that Crunchbase classifies as Series A. That is a strong fintech example because it shows grant support attaching to a narrow, high-value payments problem before larger private capital follows.

These examples do not prove that grants cause venture rounds. They do show something more useful: grants can fund the technical milestone that sharpens the investment story, while tax relief supports the underlying R&D engine as the company scales. That is the stack.

Why fintech belongs in this conversation

Fintech is not a side case. It is one of the strongest use cases for a combined grants-plus-tax model, especially in London.

The UK remained Europe’s fintech leader in 2025. KPMG reports that the UK attracted more fintech investment than France, Germany, Belgium, the Nordics, Ireland, China and Brazil combined, and accounted for more than a third of all EMEA fintech funding.

That matters because many fintech businesses are now doing work that looks much more like deeptech than conventional software. Fraud detection, secure architecture, transaction intelligence, digital identity, privacy-preserving analytics, embedded finance infrastructure and regulated AI all create the sort of technical uncertainty that can fit both a grant narrative and an R&D tax narrative. The best opportunities are not always in calls labelled “fintech”. They often sit in adjacent themes such as secure software, trust infrastructure, AI adoption, data resilience or sector productivity.

Revolut is a useful signal here, but not for the reason usually given. Its November 2025 secondary share sale valued the business at $75 billion, which shows the scale London-founded fintech can reach. But management has also indicated that any eventual IPO is more likely to be years away than months away, and that a US listing may be more attractive than London. In other words, the real lesson is not “IPO soon”. It is that the UK can still build fintech businesses of enormous scale, even if public-market timing remains uncertain.

Where FI Group by EPSA fits

For companies trying to combine grants and tax relief across multiple projects, FI Group by EPSA is most useful when the challenge is not finding one scheme, but building a coherent non-dilutive funding strategy. In practice, that means helping businesses decide when a project is grant-ready, how to separate grant-funded and self-funded costs, and how to build the evidence base needed for an R&D tax claim that can withstand scrutiny.

The mistakes that still cost companies money

The first mistake is treating grants and R&D tax relief as if they sit in different departments and never meet. They do meet, in the cost base, in the technical narrative, and in the board’s capital plan.

The second mistake is assuming that because one claim was paid, the process must be strong. That logic was always weak, and it is weaker now.

The third mistake is using AI to speed up documentation without a human checking the technical framing. AI can help organise notes. It is a poor substitute for explaining the baseline, the advance, the technical uncertainties and the evidence trail in a way HMRC will respect.

The fourth mistake is waiting too long to map the next layer of the stack. By the time a founder starts asking about grants, tax relief, investor timing and runway in separate conversations, the financing model is already fragmented.

A better approach is simple: design the stack early, track it as the roadmap changes, and treat evidence quality as part of finance, not just compliance.

FAQs

1. Can an AI or deeptech company still claim R&D tax relief if it wins a grant?

Yes, but the treatment depends on the period, the scheme, and how costs are allocated. The old SME scheme created much more friction for subsidised expenditure. For accounting periods beginning on or after 1 April 2024, the merged scheme and ERIS changed that landscape materially. It is useful to ask the help of a qualified R&D tax consultancy that also advises on grants, this is where FI Group by EPSA are best placed to advise.

2. What is the practical difference between a grant and R&D tax relief?

A grant is usually awarded for a defined future or early-stage work package and is competitive. R&D tax relief is usually claimed after qualifying R&D spend has been incurred and documented. Grants are selective and milestone-led. R&D tax relief is broader and retrospective.

3. What changed after 1 April 2024?

The old SME and RDEC schemes were replaced for new accounting periods by the merged scheme and ERIS. HMRC says the qualifying expenditure rules are aligned across the two post-reform schemes, with the merged credit rate at 20% and ERIS reserved for loss-making R&D-intensive SMEs.

4. Which live UK competitions are most relevant right now?

As of 9 April 2026, notable open routes include AI Champions: Frontier AI Phase One, Secure Software for Resilient Growth, UK-Germany Collaborative Innovation for Quantum Technologies 2026, and Solution Development for Advanced Connectivity Technologies.

5. Why is fintech especially well suited to a combined model?

Because many fintech products now involve real technical uncertainty, regulated infrastructure, secure software, AI-led decisioning and data resilience. Those features make them relevant to both grant competitions and R&D tax relief, especially in the UK market.