Nvidia Corp. and the U.S. National Science Foundation have committed a combined $152 million—$77 million from Nvidia and $75 million from NSF—to the Allen Institute for AI. The money will fund the Open Multimodal AI Infrastructure to Accelerate Science project, a national effort to build fully open artificial-intelligence models aimed at speeding scientific research and reinforcing U.S. technological leadership. Ai2 plans to develop domain-specific large language and multimodal models trained on scientific literature, with all training data, code and evaluation methods released publicly. Nvidia will supply its HGX B300 computing systems and AI Enterprise software to power model training and deployment at scale. NSF said the investment is its first large outlay for AI software infrastructure and aligns with priorities in the White House AI Action Plan. Ai2 expects a major release within 18 months and will share datasets and benchmarks throughout development to keep the research community involved.