Nvidia just wrote a check bigger than most countries' GDP to guarantee someone else's electric bill.
The Summary
- Nvidia agreed to guarantee up to $105 billion in conditional lease obligations for OpenAI's new data center campus in Pike County, Ohio, covering an initial 4.25 gigawatts of compute capacity with an option for 3.75 more.
- Nvidia is also investing $1.5 billion into SB Energy, a SoftBank Group energy company, presumably to ensure the power actually flows.
- This is the first time a chipmaker has backed another company's infrastructure buildout at nation-state scale, signaling that AI compute is now too capital-intensive for even OpenAI to finance alone.
- The deal turns Nvidia from supplier into co-investor, locking OpenAI into its hardware ecosystem while spreading the financial risk of the biggest AI buildout yet attempted.
The Signal
OpenAI is building what it calls an "AI factory" in Pike County, Ohio. The initial 4.25 gigawatts of IT power capacity represents roughly the equivalent of four large nuclear reactors running continuously. For context, the entire state of Ohio currently has about 28 gigawatts of total installed electric generation capacity. OpenAI is planning to consume a measurable fraction of a state's power grid for a single campus.
The numbers reveal how AI training has become an infrastructure problem, not a software problem. The $105 billion guarantee from Nvidia isn't cash upfront. It's a conditional backstop on lease payments. OpenAI leases the facility. Nvidia guarantees the landlord gets paid even if OpenAI can't cover it. The chipmaker is functionally underwriting OpenAI's ability to operate at scale.
"This is the first time a chipmaker has backed another company's infrastructure buildout at nation-state scale."
Why would Nvidia take that risk? Three reasons:
- It locks OpenAI into buying Nvidia chips for the next decade, eliminating customer flight risk to AMD or custom silicon
- It prevents OpenAI from collapsing under the weight of its own capex, which would crater demand for high-end GPUs
- It establishes Nvidia as the de facto central bank of AI infrastructure, the only entity with balance sheet depth to finance compute at this scale
The $1.5 billion investment in SB Energy is the tell. Nvidia isn't just guaranteeing the lease. It's buying into the energy supply chain. SB Energy was founded as part of SoftBank Group, which has its own AI ambitions and capital allocation chaos. Nvidia is now a stakeholder in ensuring the power grid can physically deliver what the lease promises. Chips don't work without electricity. Nvidia is integrating backward into the power supply.
The Ohio location matters more than it seems. Pike County isn't a tech hub. It's rural Appalachia with cheap land, low taxes, and proximity to natural gas pipelines and the PJM Interconnection grid. The buildout is expected to reshape the regional economy, creating construction and operations jobs in an area that hasn't seen this kind of capital inflow since the coal boom. OpenAI gets political cover. Ohio politicians get ribbon-cutting photo ops. Nvidia gets a customer too big to fail.
"OpenAI is planning to consume a measurable fraction of a state's power grid for a single campus."
The option for an additional 3.75 gigawatts suggests this is phase one. If the initial buildout works, the campus could eventually approach 8 gigawatts of IT capacity. At that scale, you're not building a data center. You're building a city that only machines live in. The implications for energy markets, real estate, and regional infrastructure are enormous. Someone has to build substations, upgrade transmission lines, and manage thermal load. That's not a tech company's core competency. It's a utility's.
The Implication
If Nvidia is willing to guarantee $105 billion in lease payments, the company believes AI training demand will justify the risk. That tells you more about the roadmap for GPT-5 and beyond than any OpenAI blog post ever will. They're not building this campus to run chatbots. They're building it because the next generation of models requires compute that doesn't exist yet.
Watch what happens to Ohio's energy policy over the next 18 months. If this campus gets built, every other state will start courting AI infrastructure with tax breaks and grid access. Compute is the new manufacturing. Nvidia just proved it's willing to finance it directly. That changes the game for every AI lab trying to compete with OpenAI. You either find a benefactor with a $105 billion balance sheet, or you optimize for smaller models. There's no middle path anymore.