Frontier AI: SME Champions Phase 1: Is It Worth Your Time?

Frontier AI: SME Champions, Phase 1 is an Innovate UK competition for UK-registered SMEs with a genuinely new AI or machine learning technique at TRL 1 to 3. It closes on 11 November 2026 at 11:00am, and it is only worth your time if the AI itself is the invention, you can do the work without subcontractors, and you can accept very long odds. Details below are from the Innovation Funding Service (IFS) listing and Innovate UK Business Connect, checked on 5 October 2026.

What the competition funds

Innovate UK is offering a share of up to £3.1 million for feasibility studies in frontier AI and machine learning, according to the IFS competition page. This is Phase 1 of a three-phase pipeline. Only successful Phase 1 applicants are invited to Phase 2, which the listing describes as up to 12 months and up to £1 million for building demonstrators. Phase 2 is a separate competition, and the later stage is subject to business case approval, so Phase 1 funding does not guarantee anything beyond itself.

The listing defines frontier AI narrowly. The advance has to come from innovation in model or system architecture, training methodology, core learning algorithms or core control algorithms. AI must be "the core technical contribution and main source of operational advantage" of the project.

Applicants choose one of four themes:

• AI-enabled health and life sciences
• Advanced materials with AI
• Secure AI for national security and defence
• AI safety and assurance technologies

Within the chosen theme you then pick a specific priority. Business Connect lists examples such as predictive healthcare and medicines discovery, material property prediction and defect detection, resilient edge intelligence, and agentic runtime control or formal verification for safety work.

Key facts at a glance

One date discrepancy is worth knowing about. Some aggregator listings show a later close, such as 2 December. The IFS page and Business Connect both give 11 November at 11:00am, so plan to that. IFS is the authority, and it is worth rechecking it before you commit effort. Note also that an earlier competition with a similar name, "AI Champions: Frontier AI Phase 1", has appeared on IFS with different terms, so make sure you are reading the current listing.

The £150,000 to £250,000 range needs a second look on IFS. If it is eligible project cost, the grant at a 70% rate tops out around £175,000. Our own arithmetic on £3.1 million against that range suggests roughly a dozen to twenty awards in total, which is an inference, not a published number.

Eligibility rules that rule companies out

• UK-registered micro, small or medium-sized enterprise only. Large companies and research organisations cannot lead. To check your size status, see our guide to Innovate UK eligibility.
• No subcontractors. Both sources state this plainly. If your plan relies on a university lab, a cloud specialist or an outside data science firm to do part of the work, the plan has to change before you can apply.
• One application per organisation. You cannot hedge across themes.
• Work in the UK. The project must be carried out in the UK, with results exploited in the UK and most funding spent here.
• Starting TRL 1 to 3. The project must end in a proof of concept and a technical whitepaper. If you already have a working prototype in customers' hands, you are probably above the band. Our technology readiness levels guide explains how assessors read those numbers.

The listing also names things it will not fund: routine integration or deployment of existing tools without novel AI or ML development, projects that are mainly literature reviews, hardware projects where AI is a secondary feature, and fully autonomous targeting systems. It also flags that applications lacking a quantified serviceable addressable market, clear IP ownership or a specific route to defensibility are weak.

How it is assessed

According to the IFS listing, three independent assessors review each application, with scoring on ten questions (questions 8 to 17) and seven further questions unscored. Innovate UK applies a portfolio approach, aiming for variety across themes and markets, so a strong application in a crowded theme can lose to a weaker one in a thinner theme. The listing also gives an estimated success rate of about 2%, drawn from comparable competitions. That is an estimate, not a prediction for this round, but it tells you how selective the field is likely to be.

Because the project is short, assessors have little room to forgive vagueness. Expect them to look for named baselines with numeric targets, a validation method you can finish in one to three months, evidence that the technique is new, and a believable case for defensibility and scale-up. A whitepaper on technical progress, validation evidence and Phase 2 readiness is a required output, so budget time for it inside the three months.

Who should apply, and who should not

Apply if you are a UK SME with a novel model architecture, training method or core algorithm, you can explain what is new in a paragraph, you have a specific benchmark to beat, and your team can do the work in-house in one to three months. A company with an unpublished technique and a clear route to Phase 2 fits what the competition describes.

Think carefully, and probably do not apply, if

• Your product wraps third-party models. Integration without novel AI development is excluded.
• You need a university partner or contractor. Subcontractors are not allowed.
• Your AI is a feature of a hardware product. The listing excludes hardware where AI is secondary.
• You are above TRL 3. A working product is the wrong fit for a proof-of-concept competition.
• You cannot give a quantified market size or a clear IP position. The listing names both as gaps.

With odds around 2% on the listing's own estimate, the cost of applying matters. A well-prepared Phase 1 application is real work for a small team. If the technical case is marginal, a different funding route may repay the same effort better. Our guide on when not to apply for Innovate UK funding sets out how to make that call, and how Innovate UK funding rates work will help you test the budget.

Preparing before the deadline

1. Register for the 7 October briefing, or check the IFS page for any recording or Q&A, and note anything that clarifies the scope.
2. Check your SME status and that no part of the work is subcontracted.
3. Pick one theme and one priority. Write the single sentence that says what is technically new.
4. Define baselines and numeric targets for the one-to-three-month project.
5. Build the budget against the 50% to 70% rate for your organisation size and category, and check the subsidy position, covered in our subsidy control explainer.
6. Work back from 11 November at 11:00am. Leave several days for review.

How GrantHero can help

GrantHero's job here is first to tell you whether this competition is worth your time. Discovery, which is free on our pricing page as of 5 October 2026, covers grant matching and eligibility fit checks, so you can test your company against criteria like these before committing. If you decide to go ahead, the paid tiers combine AI drafting with experienced grant experts: Expert Support is listed at £1,300 plus a 3% success fee, and Full Service at £2,500 plus a 3% success fee. Be realistic about that against a grant of this size and the odds above: for some companies the free fit check, and an honest decision not to apply, is the right outcome. Our technology page describes how the platform and the experts work together.

Before you rely on any date or rule above, confirm it on the IFS page, as competition details can be amended.