How European Researchers Access DeepMind Science Models

Google DeepMind and Google Cloud announced on 14 October 2026 that they will give up to 200 European scientific researchers direct access to frontier science models. The package includes $3 million in compute credits. The offer runs through the European Union’s RAISE initiative.

This gives selected teams tools such as AlphaFold 3 for molecular structures, WeatherNext for forecasting, and others without building their own large GPU clusters.

What the access includes

Researchers receive direct access to these models:

• AlphaFold 3: Predicts structures and interactions of proteins, DNA, RNA, and other molecules.
• WeatherNext: Supports advanced weather and extreme pattern forecasting.
• AlphaGenome: Helps analyse genomic data and DNA sequences.
• AlphaEvolve: An AI agent focused on designing algorithms.
• AlphaEarth Foundations: Maps planetary surfaces from satellite data.

The $3 million in Google Cloud credits is shared across the selected researchers. Google stated the goal is to help close the compute gap that European teams sometimes face.

The announcement appeared on the Google Blog by Maureen Costello, Vice President for UK, Ireland and Sub-Saharan Africa at Google Cloud. It ties into the EU’s AI for Science work.

How RAISE connects the pieces

RAISE stands for Resource for AI Science in Europe. It is a virtual research institute launched by the European Commission. It pools compute, data, talent, and funding for European researchers working on AI and with AI.

The DeepMind and Cloud pledge is one of the private-sector contributions under the RAISE Pledging Initiative. That initiative launched around the same time at the AI in Science Summit in Dublin. Other organisations such as Nvidia and Elsevier also made pledges.

Selection of the up to 200 researchers is expected to go through the European Commission and RAISE processes. As of the announcement date, Google did not publish a public application form, deadline, or exact eligibility criteria. Researchers at EU institutions and associated countries working on public-interest topics such as health, climate, and food security are the intended audience.

Check the official RAISE pages on the European Commission Research and Innovation site for updates. The RAISE Hub, Forum, and Observatory also launched in mid-October 2026 as central places for the community.

Practical steps for interested researchers

1. Review the official Google announcement and the RAISE site to confirm current status.
2. Monitor the European Commission Research and Innovation portal for calls or selection notices related to AI in Science and the pledging initiative.
3. Join the RAISE Forum or Hub if available to stay connected with the community and hear about opportunities.
4. Prepare a clear description of your research problem, the data you hold, and why frontier model access plus cloud credits would accelerate it. Focus on public-interest outcomes.
5. Contact your institution’s research office or European project advisors. They often track Horizon Europe and related AI science calls.
6. If selected, expect API or platform access plus a share of the compute credits. Exact technical onboarding details will come from Google Cloud and DeepMind once the programme starts.

No public application window was open at the time of the 14 October announcement. Watch for further details from the Commission or Google.

Tips for making the most of model access

• Start with a well-defined scientific question rather than broad exploration. AlphaFold 3 and similar models perform best on specific molecular or genomic tasks.
• Combine the new access with existing public resources. AlphaFold already has a large public database of predicted structures. The additional access and credits let you run custom jobs at higher volume.
• Document compute usage carefully. Shared credits mean efficient experimental design matters.
• Pair model outputs with wet-lab or observational validation. AI predictions are tools, not final results.
• If your work involves genomic variants, related guides such as How Researchers Query AlphaGenome Atlas Variant Scores can help structure analyses.

Current limitations

Google has not yet published the exact selection process, start date, or how credits will be allocated per researcher. The programme is limited to up to 200 participants. Commercial use of some models may follow different paths, such as Gemini Enterprise for AlphaFold 3.

Access is for scientific research. Terms of use for each model still apply. Researchers should review any licensing or data-handling requirements once details appear.

Why this matters for European science

Many research groups lack the large-scale GPU infrastructure needed to run the newest scientific models at volume. Direct access plus dedicated credits lowers that barrier for a selected set of teams. The models target concrete problems: molecular structures, weather extremes, genomic interpretation, algorithm design, and Earth observation.

The pledge is part of a broader set of commitments under RAISE. It sits alongside public funding already allocated under Horizon Europe for AI in science.

Researchers who want to stay informed should bookmark the Google Blog post and the European Commission RAISE pages. Further operational details are expected as the pledging initiative moves forward.

Sources

• We’re powering European scientific discovery with advanced AI compute and frontier models — Google Blog, 14 October 2026
• RAISE: Resource for AI Science in Europe — European Commission
• Google DeepMind, Google Cloud to allow European researchers access to frontier models — Reuters, 14 October 2026
• AlphaFold: The making of a scientific breakthrough — YouTube, Google DeepMind