Google is not launching a new model this time; it is handing out access to a set of existing frontier tools. On October 14, 2026, Google announced on its official blog that Google DeepMind and Google Cloud will give up to 200 top European scientific researchers direct access to its frontier models, together with $3 million in cloud compute credits. Reuters reported the same day that the pledge was unveiled at an AI summit in Dublin.
Places run through RAISE, with the Commission involved in selection
The resources will not be sold commercially. They are channelled through RAISE, the Resource for AI Science in Europe, an initiative meant to bridge the compute gap European scientists often face, with the European Commission involved in choosing participants according to same-day reports. The arithmetic is sobering: split evenly across 200 people, $3 million works out to about $15,000 each — a ticket in, not a budget that funds an entire compute-hungry project. Google's post does not publish allocation rules, a timetable, or selection criteria; those details are still to come.
Five tools aimed at five kinds of problem
Selected researchers get direct access to five scientific tools: AlphaFold 3 for predicting biomolecular structures, WeatherNext for high-resolution weather forecasting, AlphaGenome for interpreting genome sequences, AlphaEvolve for evolving better algorithms, and AlphaEarth Foundations for satellite and Earth-observation data. What they share is a high bar to entry: each is proven in its field, but each demands either gated access or heavy compute. Google's stated scenarios run from forecasting extreme weather to mapping molecular structures, which puts climate, biomedical and Earth-science groups — not just computer scientists — inside the intended audience.
Why a tech giant lends its tools to academia
European academia's compute anxiety is no secret: frontier models sit with a handful of companies, and university teams with strong ideas often cannot afford or cannot queue for the capacity they need. Packaging model access with credits through a public initiative costs Google relatively little, while embedding its tools in the daily workflow of European research — once papers, pipelines and follow-on projects build on AlphaFold or WeatherNext, the ecosystem stickiness lasts far longer than a one-off grant. For Europe, it is also the template RAISE wants: public bodies handle selection and fairness, companies supply models and compute. One reservation stands: access is not results. Data preparation, validation and reproducibility still fall on the research teams, and how much science $3 million actually accelerates depends on how places and credits are divided, not on launch-day headlines.
