Nvidia has spent the last two years selling the picks and shovels of the AI boom. Its latest move, announced Monday, shows it’s now willing to buy a stake in the mine itself.

The company said it will invest $1.5 billion in SB Energy, the data center and power developer tied to SoftBank and OpenAI, according to a company blog post and a report from TechCrunch. On paper, it looks like a straightforward equity investment. In practice, it locks Nvidia in as the exclusive compute supplier for one of the largest AI infrastructure projects currently under construction in the United States.

What the deal actually does

The investment secures Nvidia’s position as the sole provider of compute infrastructure for OpenAI’s Ports-Pike data center, being built near Cincinnati, Ohio. Beyond the equity stake, Nvidia has also committed to extending up to $105 billion in credit to help finance the buildout — a facility that, according to SEC filings reviewed by TechCrunch, could eventually scale from an initial 4.25 gigawatts to as much as 8 gigawatts of capacity.

To put that in perspective: a single gigawatt is roughly enough to power several hundred thousand homes. An 8-gigawatt data center campus would sit in the same league as some of the largest power-consuming industrial complexes in the country — except this one exists purely to run AI models.

SB Energy isn’t a new name in this story. Its existing investor base already includes both SoftBank and OpenAI, which tells you how tightly interwoven the capital structures around frontier AI labs have become. Notably, SoftBank had previously held roughly $5.8 billion worth of Nvidia stock — a position it exited in November, reportedly to redeploy that capital into other AI bets. That Nvidia is now writing a check back into a SoftBank-linked entity is a neat illustration of how circular the money flows have gotten across this ecosystem: chipmaker invests in infrastructure developer, infrastructure developer builds capacity for AI lab, AI lab’s backer had just cashed out of the chipmaker’s own stock.

The land has history — and the power plant has a price tag

One detail worth sitting with: the Ports-Pike site sits on land owned by the U.S. Department of Energy that was previously used to enrich uranium for America’s nuclear arsenal and its Navy submarine fleet. That a facility once built for Cold War-era weapons production is being repurposed for AI compute says something about how national infrastructure priorities have shifted — quietly, and without much public debate.

SB Energy also plans to build a 9.2-gigawatt natural gas power plant on the site to feed the data center. That plant alone is expected to cost around $33 billion, a figure that reflects a broader trend rather than a one-off. Gas power plant construction costs have surged 66% over the past two years, according to a BloombergNEF report covered earlier by TechCrunch, driven almost entirely by data center demand. Turbine costs, which can account for up to 30% of a new plant’s price, have reportedly climbed even faster, and waitlists for the equipment are said to be stretching into the early 2030s.

Why this matters beyond one project

The Nvidia–SB Energy deal is a useful data point for anyone trying to understand the current phase of the AI buildout — because it shows the industry moving past the “who has the best model” conversation into a much more physical, capital-intensive one: who owns the land, who finances the power plant, and who guarantees the compute supply chain end to end.

A few things stand out:

• Chipmakers are becoming infrastructure investors. Nvidia isn’t just fulfilling GPU orders anymore; it’s underwriting the energy and real estate that its chips depend on. That blurs the line between supplier and stakeholder in ways worth watching closely.
• Natural gas is the default fallback, even as it gets more expensive. As TechCrunch has reported, rising costs haven’t dampened enthusiasm for gas-fired power among AI infrastructure builders, even though alternatives like renewables paired with long-duration storage are gaining traction elsewhere, including at Google.
• Energy economics are becoming an AI story. Analysts have warned that as more gas plants like this one come online, they’ll compete directly with export markets for the same fuel — a dynamic that could triple natural gas prices in some U.S. regions, according to earlier TechCrunch reporting on hyperscaler energy strategy.

For a market obsessed with model releases and benchmark scores, this is a reminder that the AI race is, underneath it all, an energy and infrastructure race — and the companies writing the biggest checks aren’t just the ones building smarter models. They’re the ones securing the gigawatts to run them.