There never seems to be enough power to feed AI’s growing hunger.
On Monday, Nvidia and OpenAI announced a partnership to develop upwards of 10 gigawatts of AI data centers, powered by millions of the chip giant’s GPUs. As part of the deal, Nvidia will progressively invest $100 billion in OpenAI with each gigawatt deployed, with plans for the first to come online in the second half of 2026.
To put into perspective just how much power this deal is aiming to create, 10 gigawatts is enough to power roughly 8 million U.S. homes. The project will take up between 4 and 5 million GPUs, Nvidia CEO Jensen Huang told CNBC Monday, which is twice the amount the company shipped in 2024.
“There’s never been an engineering project, a technical project of this scale, ever,” Huang told CNBC. Still, Huang noted that this is “the first 10 gigawatts,” and may only be the beginning.
The bet marks these firms' ambitions that AI will completely upend the future of computing, with Huang claiming that every computing experience will eventually be “touched by AI” somehow as it breaks into “just about every single industry, every single use case we can imagine.”
For Nvidia, deals like this enable the company to "pick winners and losers in the AI race," Scott Bickley, advisory fellow at Info-Tech Research Group, told The Deep View. It allows the chip giant to build a "defensive moat" against competitors, while also potentially tempering OpenAI's ambitions in developing its own custom chips, he said.
"(Nvidia) would love to lock in the most important player in this space, OpenAI, and enhance their goal of becoming the central platform of the AI era, not dissimilar to (Microsoft's) goals in the PC era."
But getting to that future is going to require a lot of building, a problem OpenAI CEO Sam Altman recognizes. He told CNBC that in addition to investing in research and increasing product adoption, “we have to figure out how to do this unprecedented infrastructure challenge.”
Along with cementing OpenAI and Nvidia as the industry’s kingpins, this deal marks an attempt at solving that challenge – and an expensive one, at that.
It’s no secret that AI is a power hog. According to the International Energy Agency, data center power consumption is projected to account for almost half of the growth in electricity demand in the U.S. by 2030. But tech giants, once committed to bold climate initiatives, have become less keen on going green as AI ambitions cloud their vision. With AI models growing more carbon-intensive as they get smarter, a future where every computing interaction is “touched by AI” could have major environmental consequences.