NVIDIA is partnering with the national science foundation to support the Allen Institute for AI´s OMAI project, a mid-scale research infrastructure effort to build a fully open multimodal Artificial Intelligence ecosystem for scientific discovery. The contribution centers on NVIDIA HGX B300 systems and the NVIDIA AI Enterprise software platform, hardware and software designed to accelerate model training and inference at scale. The partnership is framed as a public-private investment in U.S. technology and research capacity.
The technical donations include systems built on NVIDIA Blackwell Ultra GPUs and high-bandwidth memory and interconnects, hardware described as optimized for the largest models and most demanding workloads. The support will be allocated to research teams at the University of Washington, the University of Hawaii at Hilo, the University of New Hampshire and the University of New Mexico. NVIDIA and the national science foundation say the resources will enable training, evaluation and deployment of multimodal language models that can ingest text, images, tables and graphs.
A central aim of OMAI is openness: making model code, training data, documentation and interrogation tools accessible to researchers at low or zero cost. That openness is positioned as a practical advantage for scientific work because researchers can trace model behaviors back to training instances and systematically study how data shapes emergent capabilities. Ai2 leaders frame the effort as both an accelerator for domain science and as a contribution to the science of model development, especially for early-career researchers who will receive training and tooling to inspect datasets and models.
The initiative dovetails with federal priorities. Organizers link OMAI to the White House ´Winning the AI Race: America’s AI Action Plan´ and say the project advances goals around domestic leadership in Artificial Intelligence research and federal support for data center infrastructure and technology export. Stakeholders stress that models and tools provided through the project will serve as national research infrastructure, contingent on sustained compute and collaborative stewardship between government, academia and industry.
