Anthropic Stops Handing Claude Credits One Scientist at a Time: NSF, Sloan and Simons Will Now Distribute Them

Anthropic has redrawn the distribution model of its AI for Science program. Since the program launched in May 2025, researchers applied for free Claude API credits themselves. Now, in an expansion announced on October 28, the company is routing credits through twenty scientific foundations, institutions and research groups that will decide who gets them.

The first tier consists of six funders and scientific societies that will hand credits to researchers inside their own grant programs: the Alfred P. Sloan Foundation, the Simons Foundation, Schmidt Sciences, the Open Source for Science Fund at Renaissance Philanthropy, the American Chemical Society's CAS division — and the US National Science Foundation, a roughly $9 billion agency that accounts for about a quarter of federal support for basic research at American colleges and universities.

A second tier of institutions will deploy the credits internally. It includes the Space Telescope Science Institute, which runs science operations for the Hubble, Webb and Roman observatories; the Wellcome Sanger Institute, one of the world's largest genomic data producers; the Francis Crick Institute with its more than 100 research groups; Imperial College London's Department of Life Sciences; EMBL, the European Molecular Biology Laboratory; and the Gladstone Institutes in San Francisco.

Six specific projects receive credits directly: FlyWire at Princeton, which produced the first complete wiring map of an adult fruit fly brain; gnomAD at the Broad Institute, an open reference pooling DNA data from more than 800,000 people; AstroPath at Johns Hopkins, which applies astronomy calibration methods to tumor imaging; LEAP at Columbia, an NSF-funded climate modeling center; Tau Ceti, a shared autoformalized mathematics library in Lean; and Australia's Centre for Population Genomics, a joint venture of the Garvan Institute and Murdoch Children's Research Institute.

Two existing collaborations anchor the expansion. The Howard Hughes Medical Institute's Janelia Research Campus is co-developing the Model Hardware Standard, a shared specification in research preview that lets AI agents safely operate laboratory instruments. The Allen Institute is building teams of AI agents to analyze open brain, cell and immune datasets — work Anthropic says can compress months of manual analysis into hours.

Anthropic did not disclose how many credits each partner receives or what they are worth in dollars, and it says many collaborations are at an early stage. The strategic logic, though, is explicit. Research is one of the few high-value settings that genuinely requires long context, long chains of reasoning and tool use — and a federal agency's participation doubles as an endorsement for a company that has been pushing hard into government work. OpenAI fields a dedicated science team and Google carries DeepMind's AlphaFold halo; Anthropic's answer is infrastructure: make Claude the tool a generation of scientists already uses every day.

What separates this from a discount program is what the projects are supposed to leave behind: open datasets, open-source software, shared libraries and technical standards that outlive any credit balance. Whether those public goods actually materialize will determine if this round of AI-for-science patronage is remembered as marketing — or as infrastructure.