AI

Google’s science AI offer runs through Europe’s research network

· Geeknewz Author

Original editorial illustration of a research network surrounding RAISE, with text about Google's European science pledge.
Original Geeknewz editorial illustration; not a scientific model.

Google’s latest scientific AI offer comes with a route through an existing European research programme. Up to 200 researchers will receive access to its frontier models and a shared $3 million pool of computing credits, but this isn’t a public giveaway that any lab can claim.

The practical detail is who gets selected. AFP reports that the European Commission will choose researchers from beneficiaries of RAISE, the EU’s AI science initiative, citing clarification from Google. For a research team, that connection matters more than dividing the headline budget by the maximum number of recipients.

Google’s October 14 announcement names AlphaFold 3, WeatherNext, AlphaGenome, AlphaEvolve and AlphaEarth Foundations as the tools involved. Google DeepMind and Google Cloud are behind the pledge, which includes making some frontier models available through RAISE.

The company points to extreme-weather prediction and molecular-structure mapping as examples of the research it wants to support. It doesn’t publish a per-team credit allocation, a credit expiry date or a detailed access timetable in the announcement. Those are consequential omissions if you’re deciding whether to build a project around the offer.

The programme already has a funding structure

The European Commission describes RAISE as a virtual research institute that pools computing resources, data, talent and funding. Its pilot launched in November 2025. Dedicated Horizon Europe funding includes €33 million in the 2025 work programme and €107 million in the 2026–27 programme.

Google’s credits therefore sit alongside an existing public research effort. They shouldn’t be confused with the programme’s total funding, and credits aren’t cash that a lab can redirect toward staff or equipment.

Access also needs a more careful reading than the headline might suggest. The announcement doesn’t say Google is releasing model weights or removing existing licensing conditions. A team planning to share its methods with collaborators needs the actual terms, including what others would require to reproduce the work.

For everyone outside the selected group, the useful thing to watch is what gets produced: documented experiments, comparisons with established methods and results other researchers can examine. A model-access pledge establishes an opportunity; it doesn’t establish a scientific breakthrough.

If your lab participates in RAISE, ask its programme contact about selection and the model-specific terms before budgeting around these credits. If it doesn’t, don’t treat the announcement as an open application window.