Public AI computing and publicly funded research are where AI meets UK science. These are results that DSIT and UK Research and Innovation have reported, quoted word for word.
This page quotes 7 findings from five official records by DSIT, UKRI and HM Treasury, published between July 2025 and September 2026. The most recent comes from UKRI in September 2026 and covers health research. One of them carries the source's own note on how the figures were gathered, shown under the quote.
Chemistry
68 million
chemical combinations screened by the EIMCRYSTAL AI model (DSIT) · see the record
Supercomputing
350+
research projects supported by the DAWN AI supercomputer (DSIT, January 2026) · see the record
AI safety science
30
frontier AI models tested by the AI Security Institute (DSIT, January 2026) · see the record
Health research
13
bacterial species in the AI-derived gum disease panel (UKRI) · see the record
Supercomputing
DSIT reports that the Isambard supercomputer was formally launched in Bristol on 17 July 2025, meaning the AI Research Resource (AIRR) is fully up and running.
“The Technology Secretary flicked the switch on the Isambard supercomputer at its formal launch in Bristol today, meaning the AI Research Resource (AIRR) is now fully up and running”
DSIT reports that University of Liverpool researchers have used Isambard to develop EIMCRYSTAL, an AI model that speeds up the discovery of new chemical reactions for industry by sifting through 68 million chemical combinations.
“Researchers at the University of Liverpool meanwhile have been using Isambard to develop their EIMCRYSTAL system. Their model harnesses AI to speed up the discovery of new chemical reactions for use in industry, sifting through 68 million chemical combinations”
DSIT reports that in its first year the Sovereign AI Unit allocated compute to AI researchers and start-ups, including the University of Cambridge's MACE materials foundation model, and provided £8 million seed funding for the OpenBind consortium's dataset for AI-driven drug discovery.
“allocated sovereign compute to support AI researchers and high-growth start-ups to conduct groundbreaking work in the UK, including the University of Cambridge’s MACE (a materials foundation model for atomistic simulations and discovery) invested in creating data assets to support UK companies, including providing £8 million seed funding for the OpenBind consortium’s structural dataset to unlock AI-driven drug discovery”
UKRI reports that researchers at STFC's Hartree Centre, IBM Research and Salient Bio used AI to narrow hundreds of variables down to 13 key bacterial species, and that this compact panel was just as accurate at diagnosing gum disease as models relying on far more data.
“From hundreds of variables, AI distilled the complexity of the entire microbiome down to just 13 key bacterial species. This compact panel proved just as accurate at diagnosing gum disease as models relying on vastly more data”
UKRI reports that in an MRC-funded study led by the University of Oxford, people with higher scores on the AI-derived 'HyperScore' were more likely to experience future cardiovascular problems, even when blood pressure readings alone did not fully distinguish risk.
“The researchers found that people with higher HyperScores were more likely to experience future cardiovascular problems, even when blood pressure readings alone did not fully distinguish risk.”
The source notes: The researchers caution that the approach is still at an early stage and is not yet ready for routine clinical use.
DSIT reports that the Isambard supercomputer was formally launched in Bristol on 17 July 2025, meaning the AI Research Resource (AIRR) is fully up and running. See the record.
How many projects has the DAWN supercomputer supported?
DSIT reports that Cambridge's DAWN AI supercomputer has supported over 350 projects. See the record.
What we left out
We only publish results that an official source reports as having happened. We leave out forecasts, targets, estimates of what AI could or may do, announced investment that has not yet happened, and findings from bodies that are not on our official sources list. How we choose what to include.