OpenAI's $500B Data Center Push: Georgia Site and Nvidia Talks

New details are emerging on the scale and geography of OpenAI’s infrastructure ambitions, with a massive Georgia data center and a near-finalized Nvidia-backed financing structure adding substance to what had been a high-level story.

Key points

• OpenAI is close to finalizing a $500 billion data center deal with Nvidia’s backing, according to the New York Times, with Nvidia guaranteeing $250 billion of that sum.
• A specific site has now been identified: OpenAI plans to spend over $30 billion on a data center near Savannah, Georgia, one of the largest single-site commitments in the company’s history.
• Critics are drawing parallels to dot-com era circular financing, with Jim Cramer and Reuters Breakingviews both flagging risks in Nvidia’s role as both chip supplier and financial guarantor.
• OpenAI is among the tech companies lobbying Washington against curbs on open-source AI models, aligning with rivals on that narrow policy question even as competition intensifies.
• Chinese AI labs including Moonshot AI’s Kimi K3 are drawing direct comparisons to OpenAI’s models, adding competitive pressure to OpenAI’s infrastructure bet.

What’s new on the Georgia site?

Yesterday’s coverage established the broad contours of the Nvidia financing discussions. What’s new today is the geographic specificity. OpenAI’s planned facility near Savannah, Georgia represents a concrete commitment inside the larger $500 billion headline figure, and it matters for a few reasons.

Savannah sits near major port infrastructure, has access to large parcels of land, and Georgia has been aggressive in attracting large industrial and tech investments with tax incentives. A $30 billion-plus spend on a single campus would rank among the largest data center investments ever announced in the US Southeast. For investors tracking OpenAI’s capital deployment ahead of any potential public offering, this is the kind of committed, shovel-ready spending that signals the infrastructure buildout is real rather than aspirational.

It also raises questions about power sourcing and permitting timelines. Data centers at this scale require gigawatts of capacity, and Georgia’s grid will face pressure. Those details have not yet been reported.

How real is the $500 billion Nvidia deal, and what are the risks?

The structure being reported deserves scrutiny. Nvidia is reportedly guaranteeing $250 billion in data center funding for OpenAI. That means Nvidia, which sells the chips that fill those data centers, is also helping finance the buildings that house them. The customers buy from Nvidia, and Nvidia backs the capital that makes those purchases possible.

Bloomberg, Reuters Breakingviews, and CNBC have all flagged the circular nature of this arrangement. Jim Cramer went further, calling it reminiscent of how companies collapsed during the dot-com bubble. Whether that analogy holds depends heavily on whether AI infrastructure spending is ultimately matched by revenue. OpenAI has been growing revenue rapidly, but the capital outlays being discussed here are measured in the hundreds of billions.

For now, the deal is not finalized. The word “close” appears in the NYT’s framing, and the discussions remain ongoing. Investors should treat the $500 billion figure as a target under negotiation, not a signed commitment.

The competitive backdrop

OpenAI’s infrastructure bet is partly a response to competitive pressure that is not letting up. Chinese lab Moonshot AI’s Kimi K3 model is drawing comparisons to OpenAI’s offerings on coding, reasoning, and knowledge tasks. This follows DeepSeek’s disruption last year, which showed that frontier-quality models could be built at a fraction of the cost US labs have been spending.

Microsoft CEO Satya Nadella has argued the US retains a structural advantage, though the sources do not specify precisely what that advantage is beyond the broader ecosystem. OpenAI’s response to the cost-efficiency challenge from Chinese competitors appears to be scale: build more compute capacity than anyone else and bet that raw infrastructure leads to model superiority. That is a defensible thesis, but it is an expensive one, and it depends on continued access to capital at the terms currently being negotiated with Nvidia.

On the policy front, OpenAI has joined a broad coalition of tech companies urging Washington to avoid restrictions on open-source AI models. That alignment with companies it competes with directly, including Anthropic, reflects a shared interest in keeping regulatory constraints limited even as the competitive dynamics between them remain sharp. Nothing in the reported policy positions changes OpenAI’s near-term business outlook materially.