OpenAI is in talks to lease a massive 10-gigawatt data center in Ohio, with financial backing from Nvidia, according to The Information. The report signals another huge escalation in the compute arms race, and it puts two of the most important names in AI on the same buildout. If the deal closes, it would rank among the largest single data center commitments the industry has ever seen.
What stands out here is the scale. Ten gigawatts isn’t a normal data center number. It’s closer to the output of a fleet of nuclear reactors.
What 10 gigawatts actually means
Most large data centers today are measured in tens or hundreds of megawatts. A single gigawatt is already enormous. Ten of them is a different category entirely.
To put it in perspective:
- One gigawatt roughly powers 750,000 homes.
- Ten gigawatts is comparable to the peak electricity demand of a mid-sized US state.
- A buildout like this can’t run on the existing grid without major new power generation and transmission.
That’s why these projects increasingly come bundled with their own energy strategy: gas turbines, nuclear deals, and long-term utility contracts. Power, not chips, has become the real bottleneck for AI.
Why Nvidia’s involvement matters
Nvidia backing the deal is the part worth watching. The company already sells the GPUs that fill these buildings. Now it’s increasingly helping finance the buildings themselves.
This is a pattern. Nvidia recently committed to a large investment tied to OpenAI’s infrastructure ambitions, and it has taken stakes in other compute players. By helping fund the data centers, Nvidia helps guarantee demand for its own hardware. It’s a circular arrangement that critics have flagged, and supporters call it the only way to move fast enough.
For OpenAI, the backing helps solve the central problem of this era: the capital required to keep scaling is staggering, and no single company wants to carry it alone.
How this fits the bigger picture
This Ohio talk doesn’t exist in a vacuum. It lands on top of OpenAI’s Stargate effort, the multi-hundred-billion-dollar infrastructure push it announced alongside SoftBank and Oracle. The company has been signing compute and chip deals at a pace that would have looked absurd a year ago.
The status quo before all this was simpler. AI labs rented capacity from cloud providers like Microsoft, Google, and Amazon. Now the leading labs are going direct: securing land, power, and custom-built sites on their own terms.
The shift tells you something. OpenAI is behaving less like a software company and more like a heavy-industry operator, one that needs energy, real estate, and steel as much as it needs research talent.
Why it matters for the industry
Three takeaways stand out for anyone tracking AI:
- Compute is the moat. The companies that lock up power and chips now will set the pace for the next several years of model development.
- Energy is the new constraint. Expect more deals tied to nuclear, natural gas, and grid expansion. Ohio and the broader Midwest are becoming prime territory because of available land and power access.
- The money is getting circular. Chipmakers funding their own customers raises real questions about how much of this demand is organic versus engineered. Watch how regulators and investors respond.
What to expect next
This is still described as talks, so nothing is signed. Large leases like this can take months to finalize, and the power commitments alone require utility and regulatory sign-off.
Still, the direction is clear. The biggest AI players are racing to secure gigawatt-scale capacity, and they’re pulling chipmakers, utilities, and entire regional economies along with them. If you build, invest, or operate anywhere near AI infrastructure, the ground is shifting fast.
For the full details on the Ohio negotiations and Nvidia’s role, see the original report from The Information.