12 Gigawatts Is No Small Target: NVIDIA and OpenAI Have Turned AI Compute into an “Electricity Arms Race”

NVIDIA and OpenAI continue to ramp up their investments in hyperscale AI infrastructure. The latest project involves approximately 12 gigawatts of NVIDIA computing capacity, with room for further expansion in the future. It is worth noting that public information varies by source: the first phase of the Ohio project officially announced by OpenAI involves approximately 8 gigawatts of IT capacity, while market reports indicate that the overall plan could reach approximately 12 gigawatts or even higher.

What does 12 gigawatts mean? Simply put, this is no longer about “buying a few GPUs to run models,” but about building an AI industrial city. Data centers, electricity, chips, networks, and cooling systems—none of them can afford to fail. AI competition is evolving from “whose model is smarter” to “who can secure more electricity, more GPUs, and more stable infrastructure.”

For NVIDIA, this is naturally a highly lucrative opportunity. OpenAI needs to continuously expand its training and inference capabilities, while NVIDIA happens to control GPUs, networking, and the entire AI infrastructure ecosystem. The larger the project, the more significant the demand for NVIDIA’s chips and systems. NVIDIA even expects the related opportunity to generate revenue potential worth hundreds of billions of dollars by 2030.

But what the market should truly be wary of is that AI infrastructure has entered the era of “super capital expenditure.” NVIDIA is not only selling the picks and shovels, but is also beginning to participate in financing and infrastructure planning. In the latest Ohio project, NVIDIA plans to invest $1.5 billion in SB Energy and provide up to approximately $105 billion in guarantee support.

This means that NVIDIA is gradually evolving from a mere chip supplier into a key organizer of the AI infrastructure ecosystem.

So the real story behind 12 gigawatts is not simply how many GPUs OpenAI wants, but that the AI industry is moving from a software competition into an integrated war of “compute + energy + capital.”

AI companies used to compete on algorithms; now they compete over whose power outlet is thicker. It has to be said that in this round of AI competition, even electricians are starting to have the characteristics of technology stocks. #英伟达与OpenAI锁定12吉瓦算力