Finnish biology research hits computing shortage as AI demand surges

Finnish biology research is increasingly dependent on major technology companies for computing capacity, Yle reports. But the tools researchers need most are in short supply, with one University of Helsinki research group unable to obtain the secure computing capacity it sought from Google.

Associate Professor Andrea Ganna’s biomedical research group is building an artificial intelligence model using Finnish health records. The aim is to use the model to identify early signs of more than 200 diseases and predict the likelihood of adverse drug reactions.

The model, known as a foundation model, is being trained to detect patterns in registry data and material from FinnGen, a project that collects information on the genetic heritage of Finns.

Ganna said the research group mainly develops its own tools rather than using consumer language models such as ChatGPT or Claude. He would like to use those models for information gathering and programming, but said Anthropic’s flagship model refuses to help with biological matters.

Researchers must apply to the manufacturer for permission to remove the safety restrictions on the most powerful AI models. Ganna said he had not applied because he considered the threshold too high.

The immediate problem arose when Google could not offer the group the computing capacity it needed. The researchers use Google’s cloud services to train AI models and sought GPU computing in a secure environment.

“We were told that there was no available capacity for us,” Ganna said.

Pawel Luty, a representative of Google’s cloud services, told Yle that the solution requested by the Finnish research group was unusual and in very high demand. Google could not provide the requested equipment, but said it was looking for alternative solutions together with the group.

Prices for computer components have risen significantly in recent years, and Google’s cloud services are therefore offered to research groups according to demand.

Growing uncertainty over access

Finland has one of the world’s most powerful supercomputers, but the LUMI supercomputer in Kajaani is connected to the internet and therefore does not meet the security requirements of the research groups. Universities consequently depend on Google, Amazon and other technology giants.

Samuel Kaski, an artificial intelligence professor at Aalto University and director of the ELLIS Institute, said European universities could not necessarily rely on familiar services.

Uncertainty over access to AI tools increased in June, when the United States administration ordered AI company Anthropic to remove access to its most powerful model from all users outside the United States. Media reports have suggested that the order was linked to poor relations between the company and President Donald Trump.

“The uncertainty about which tools will be available to us in the future has increased significantly since then. The research community now has to consider this as well,” Kaski said.

Luty said he considered it unlikely that the Trump administration would intervene in Google’s services in Europe in the same way as in the Anthropic case. He did not directly answer whether a similar service shutdown could occur in protected cloud services, but described it as unlikely.