OpenAI signs agreement with US National Laboratories

OpenAI announced an agreement with the US Department of Energy covering all 17 of its National Laboratories, intended to give laboratory scientists access to OpenAI’s reasoning-focused models for research use. OpenAI said the agreement could eventually reach up to 15,000 scientists across the system.

The most concrete piece of the deal was infrastructure: OpenAI and Microsoft would deploy an o1-series reasoning model on Venado, an Nvidia supercomputer housed at Los Alamos National Laboratory, as a shared resource for researchers at Los Alamos, Lawrence Livermore and Sandia. OpenAI framed the intended applications around basic science — materials science, renewable energy, astrophysics — rather than weapons work specifically, though Los Alamos and Lawrence Livermore both carry out nuclear-weapons-related research as part of their core mission. OpenAI said access to its models would be conditioned on review of specific use cases and on consultation with OpenAI staff holding security clearances, a structure intended to keep the company informed about how its models were being applied inside classified or dual-use research settings.

The agreement fit a broader pattern in OpenAI’s public positioning in the opening weeks of the second Trump administration, alongside a similar Los Alamos bioscience partnership and OpenAI’s push for closer integration with US government infrastructure and its “AI Action Plan” advocacy — part of a wider argument, echoed across the industry, that domestic government adoption of American AI models mattered strategically against Chinese competition. It marked one of the more direct connections yet drawn between a frontier lab’s commercial models and the national-security research infrastructure of a nuclear-weapons state.