Google DeepMind and Google Cloud will make their state-of-the-art models available to as many as 200 European scientific researchers directly. This also comes with $3 million in compute credits. It was announced by the companies on Wednesday.
Numbers are of small size and specific. The impact is not possible.
What Google DeepMind and Google Cloud Are Offering
There are two components of the program. The first is direct access to frontier models and scientific tools that Google DeepMind has created. The second is a $3 million compute credit, provided via Google Cloud.
Both parts matter. The real power behind advanced AI models lies in the ability to run them, which requires a team to afford them. In smaller labs, particularly universities, compute is one of the factors that typically prevents a good project. Credits help to clear some of that obstacle.
The AI Tools Researchers Can Use
The researchers will have access to five scientific tools:
- AlphaFold 3: Structure Prediction in Molecular Biology
- WeatherNext: Forecasting Weather.
- The human genome studied with AlphaGenome.
- The fastest algorithm discovery and improvement tool, AlphaEvolve.
- The analysis of the planet with earth observation data is done with the help of the AlphaEarth Foundations.
They are both focused on biology and climate, genetics, computing, and earth science. One program has very different activities in it.
Why Molecular Structures and Extreme Weather Matter
The firms cited two examples of what these tools can do: predicting extreme weather patterns and mapping molecular structures.
The middle is drug discovery and disease research, which is why the mapping of molecular structures is at the heart of it. Understanding the structure of a molecule aids in understanding its behaviour. Weather forecasts for extreme events impact the farmer, the city planner, emergency services and families. Early warning messages provide people more time to prepare.
The companies claim that the tools can enable researchers to address complex problems at a more rapid pace. A reasonable goal for speed is available. The limits of science are imposed not by the range of man's ideas, but by the narrow confines of his resources.
What This Means for European Science
Frontier models have mostly been in the hands of large technology companies and well-funded institutions. The program is designed to increase the opportunities for a targeted community of European scientists. The number of up to 200 researchers is not a large amount, so not all laboratories will be able to benefit. Nonetheless, each researcher that joins can be helping an entire team and the results of their research can extend beyond the scope of one project.
Questions that are still unanswered should be left unanswered and a neutral reading should state that. The announcement doesn't provide details on how researchers will be chosen or what the program will cost the companies, except for credits. That information will influence the extent of the benefits' distribution.
The Role of Compute Credits in Scientific Research
Computing credits is easy to miss. They don't get publicity as a new model does. Large AI systems require a lot of computing power, though. If not, having access to a model is equivalent to having a car without petrol.
Until recently, a $3 million pool shared among as many as 200 researchers provided a real shot in the arm for a project that has been in need of one. It will not finance anything. It provides practical ideas for teams to get going.
Frontier Models and the Future of Research
The program represents a general trend. AI is becoming more than a research tool, it's a research tool. It is a tool that scientists use to test their ideas, model systems, and even to sift through huge amounts of data that would take humans years to analyze.
Such a change carries with it real obligations. Human checks, reproductions and reviews are still required for results generated by AI tools. These tools are useful aids to that human judgment and form the core of good science.
The Takeaway
European researchers get access to AlphaFold 3, WeatherNext, AlphaGenome, AlphaEvolve, and AlphaEarth Foundations, as well as $3 million in compute credits, from Google DeepMind and Google Cloud. The scale is quite small and the purpose is evident. Provide able scientists with the tools and let them work on hard problems.
The result they uncover will determine whether or not the program will be considered a success. It's an opportunity for the 200 scientists who do get in to make more rapid progress on issues that impact health, climate and everyday life.
