Nvidia found $1B under the couch to help secure American scientific computing dominance
Commitment comes as GPUzilla prepares to fortify Uncle Sam's arsenal with at least seven AI-optimized supers
In a strategic move to bolster American scientific computing, Nvidia announced a $1 billion commitment on Thursday, representing roughly 1/60th of its quarterly profits. The investment, unveiled in Washington, DC, is aimed at funding AI research and development over the next five years. The focus will be on areas including quantum computing, healthcare, and energy security.
This aligns with Nvidia's long-standing goal of merging AI with high-performance computing. While its primary focus remains on AI training and inference, the company has consistently introduced innovative CUDA libraries for AI-assisted quantum computing, healthcare, drug discovery, and physics simulation. The commitment is part of the US government's Genesis Mission, initiated by the Trump administration, which seeks to leverage AI to drive scientific discoveries that ensure the nation's technological dominance.
This represents Nvidia's return to US supercomputing on a large scale. The Department of Energy's (DoE) previous flagship supercomputers, Summit and Sierra, were launched in 2018. Although Nvidia had been supplying smaller supercomputers to the US government since 2021, their latest offering, the 100,000 Blackwell-GPU-based Solstice system for Argonne National Laboratory, is set to be a significant leap forward.
The system, developed in collaboration with Oracle, is primarily designed for AI workloads that don't require ultra-high precision computation, utilizing 16, eight, or even four bits of precision. This strategic positioning positions Nvidia to meet the goals of the Genesis Mission. Nvidia's shift towards precision-flexible computing comes as the company focuses more on higher throughput at precisions used by lucrative AI workloads.
In response, the DoE's largest supercomputer, expected to be powered by AMD in Oak Ridge National Laboratory's Discovery system, will still offer superior performance in pure FP64 computations. Meanwhile, China's LineShine computer currently leads the Top500 ranking of the world's fastest computers, with 2.2 exaFLOPS measured out of its 2.7 exaFLOPS theoretical performance.
Written by urgent.news from The Register Science's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
