Nvidia Pledges $1 Billion to U.S. Science, a Sum The Register Calculates at One-Sixtieth of a Quarter's Profit

On Thursday, Nvidia announced $1 billion in commitments to advance U.S. science over the next five years. The announcement came at an event in Washington, D.C. called “Science: A New Golden Age,” per Nvidia’s release. It is a large number. It is also, as The Register’s Tobias Mann noted, roughly 1/60th of Nvidia’s quarterly profits. A billion dollars, in other words, is what Nvidia finds when it checks between the cushions.

The Register’s headline puts it more colorfully: Nvidia “found $1B under the couch.” We won’t improve on that, but we will note that, in fairness, the couch is very large.

First, the facts as stated by the company and the coverage. Nvidia says the money will go toward AI-driven research and development in areas including quantum computing, healthcare, and energy security. Per its release, it will support higher-education research institutions, U.S. quantum computing efforts, and cloud service providers that serve U.S. government missions. Nvidia also says it collaborates on several Phase 2 awards under the Genesis Mission, a U.S. government program for AI-accelerated science, in quantum computing, fusion, accelerator design, and microelectronics, and that it invites members of the science community to use its compute and AI infrastructure.

CEO Jensen Huang said, “With a $1 billion investment, NVIDIA is putting advanced Super Intelligence in the hands of America’s scientists.” Yahoo Finance’s account quotes him calling the Genesis Mission “launching a new golden age of American discovery.” Separately, the Department of Energy announced $159 million for 12 new Phase 2 Genesis Mission projects, including a digital twin for Commonwealth Fusion Systems’ SPARC fusion device, per the same coverage. This article takes no position on the program itself; the story here is the number and its arithmetic.

The Register provides useful historical context on the real product underneath the pledge: supercomputers. Nvidia’s last flagship Department of Energy machines, Summit and Sierra, were commissioned in 2018. Perlmutter, built on A100 GPUs, followed in 2021 and was notable but smaller. In the years since, AMD-based systems such as Frontier and El Capitan have been much larger. Nvidia’s 2025 announcement that it would support seven new systems across Argonne and Los Alamos, including what it calls the department’s largest scientific supercomputer at Argonne, is thus best read as a return to the genre.

There’s also a nerdy and charming technical subplot. Per The Register, Nvidia’s GPUs have shifted toward the lower-precision number formats AI workloads prefer and away from the double-precision FP64 that traditional simulation loves. For the upcoming Vera Rubin platform, Nvidia leans heavily on FP64 emulation based on the Ozaki scheme, which the article calls worthwhile for mostly-AI workloads that occasionally need higher precision, though not without compromises.

Put plainly: the chip company is investing a billion dollars in science, and part of its pitch is a very clever way to get the right answer from hardware that increasingly prefers approximately the right answer, quickly. The science, meanwhile, would like to know whether the seventh decimal place is going to be a problem. It’s a pledge about precision, made at a company currently optimizing for fewer digits.

The Register adds a competitive note. Oak Ridge’s Discovery system, expected online in 2029, will use AMD MI430X GPUs and Venice CPUs, built by HPE’s Cray division and aimed at FP64 performance, with estimated peak performance of 3.3 to 8.5 exaFLOPS depending on a facility power upgrade. And China’s LineShine leads the Top500 ranking, having debuted with 2.2 exaFLOPS measured. Nobody in this story is standing still, including the people whose job it is to stand and look at the chart.

Yahoo Finance reports Nvidia stock fell about 0.5% midday Thursday, to near $235, after a record high of $243 on Tuesday; it’s up about 25% this year. A billion-dollar philanthropic gesture moved the stock by, give or take, the amount of a rounding preference. This is the first time in recorded history a billion dollars has been described as “in line with the noise.”

Let’s be fair. A billion dollars over five years is, in absolute terms, a lot of money for science, and none of the coverage suggests the commitment is anything but real. The fraction is a framing choice by a reporter, and a good one: it tells you how much money there is now in the part of the economy that sells shovels to the gold rush. You could fund an entire national program from the change in one company’s pocket and still have enough left over for a very nice jacket.

Sources: