AI Champions Frontier AI Phase 1: Advanced Materials with AI is the part of the competition aimed at SMEs building frontier AI methods for materials prediction, generative models and multi-objective optimisation, physics-ML discovery and simulation acceleration, or multimodal knowledge discovery platforms. As with the wider competition, Phase 1 is a single-applicant feasibility round for UK-registered SMEs, with £150,000 to £250,000 project cost, 3 to 6 months duration, and no collaborators or subcontractors.
This theme has strong policy backing. The government’s AI for Science Strategy identifies advanced materials as one of five priority areas and points to an initial £50 million commitment for the National Materials Innovation Programme to support data-driven tools for discovery, design and validation. UKRI also says new materials can radically transform society and are central to sectors including healthcare, aerospace, automotive, defence and energy.
This theme is not simply about using AI somewhere inside a materials business. It is about using frontier AI to change how materials are discovered, designed, simulated, screened, or translated into use.
The three official priority areas are:
That makes this page relevant to AI-native materials startups, computational chemistry firms, simulation companies, deep tech software businesses, and advanced manufacturing SMEs where the real novelty sits in the AI method.
This theme is not simply about using AI somewhere inside a materials business. It is about using frontier AI to change how materials are discovered, designed, simulated, screened, or translated into use.
The three official priority areas are:
That makes this page relevant to AI-native materials startups, computational chemistry firms, simulation companies, deep tech software businesses, and advanced manufacturing SMEs where the real novelty sits in the AI method.
Strong bids in this theme usually do one very specific thing well.
Examples might include:
This is already happening in the market. UKRI has highlighted work by Materials Nexus and the University of Sheffield using machine learning and experiments to find lower rare-earth magnet alternatives for electric machines. That is the kind of technically credible direction assessors will understand. (
The same core Phase 1 logic still applies. Your proposal needs to show:
For this theme, the benchmark question matters a lot. If you are claiming better optimisation, better candidate generation, or faster simulation, you need to show what you are better than.
The first problem is that the AI is too generic. If the real story is “we use AI to help scientists work faster”, that is too soft. The competition wants novelty in the model, training method, or learning/control approach.
The second problem is over-claiming laboratory delivery. In a 3 to 6 month feasibility project, you are not going to industrialise a new material. You need to prove the core AI feasibility clearly enough that a later demonstrator project becomes credible.
The third problem is trying to cover every industrial use case at once. A battery materials project, a defence materials project, and a semiconductor materials project are not one bid. Pick one commercial beachhead.
The fourth problem is weak defensibility. In materials AI, defensibility can come from proprietary data, a unique workflow, domain-specific know-how, better coupling between physics and ML, or a superior validation loop. If your approach is easy to copy from public models and public data, you need to address that head-on.
A good Theme 2 work plan is usually built around:
That discipline matters because Innovate UK expects a technical white paper at the end of Phase 1 and a clear translation from Phase 1 evidence into Phase 2 objectives.
FAQs
