OpenAI Just Committed $30 Billion to a Single Data Center — And It Needs 3.2 Gigawatts to Run It

When I think about the scale of what OpenAI is building right now, I genuinely struggle to wrap my head around it. $30 billion. 3.2 gigawatts of power. One data center campus. This is not a tech company anymore — this is an infrastructure nation.

Project Camellia: OpenAI's First Self-Built Data Center

OpenAI has announced plans to build a massive data center campus in Effingham County, Georgia, near Savannah — and this one is different from everything they've done before. It's the first data center OpenAI will design and build entirely themselves, rather than renting space from Microsoft Azure or other cloud providers. They're calling it Project Camellia, and the numbers are staggering.

The campus will run on 3.2 gigawatts of electricity — a level of power consumption that rivals small countries. OpenAI has already secured a contract with Georgia Power for that capacity, with the first several hundred megawatts expected to come online in 2028 and the rest phasing in through 2032. At full scale, the project could exceed $30 billion in total investment.

For reference: the entire US nuclear power fleet generates about 100 gigawatts. OpenAI wants roughly 3% of that just for one campus.

Why Georgia?

Savannah's surrounding area has become one of the fastest-growing industrial corridors in the American South, thanks to the port of Savannah, available land, and competitive energy deals. OpenAI chose the Savannah Gateway Industrial Hub, about 45 minutes outside Savannah itself, for this campus. That tracks — the region has seen massive investment from EV manufacturers, chip fabs, and logistics companies over the past few years, and the existing infrastructure makes it easier to bring a project of this scale online faster than building from scratch in a more congested area.

There are local concerns, of course. Reports indicate OpenAI is trying hard to quell local opposition to the project — which likely means water usage, land displacement, and noise issues that always accompany developments of this magnitude.

Presence: OpenAI Goes Deep Into the Enterprise

While the data center headlines grabbed attention, OpenAI also officially launched "Presence" — its enterprise AI agent platform. Presence is designed to deeply integrate AI models with a company's internal data, policies, and business processes, powering automation across customer service, sales, IT support, and more.

This is a significant strategic move. OpenAI has been growing its enterprise revenue rapidly, but Presence represents a pivot from "ChatGPT for companies" to "AI agents embedded in your entire operation." It's a direct shot at Salesforce, ServiceNow, and a dozen other enterprise software giants who've been trying to bolt AI onto their existing platforms.

Instead of a company using ChatGPT to draft emails, Presence means AI agents that can understand your company's internal policies, access your proprietary databases, and autonomously handle entire workflows — from onboarding a new employee to resolving a customer complaint end-to-end, without a human in the loop.

The $750 Billion Picture

The Georgia announcement fits into a much larger pattern. Reports indicate OpenAI has raised its projected compute spend through 2030 to roughly $750 billion — up from a previous estimate of $600 billion. That's not a typo. Seven hundred and fifty billion dollars in compute investment over the next four years.

To put that number in context: the entire US defense budget for 2026 is around $895 billion. OpenAI is planning to spend nearly as much on servers and power infrastructure as the US military spends on everything — in four years.

What This Means for the AI Race

We're entering a phase of the AI race where the moat isn't just your model — it's your infrastructure. Google, Microsoft, Amazon, Meta, and now OpenAI are in an arms race for compute capacity that the rest of the industry simply cannot match. If you can run more inference, train bigger models, and power more agents simultaneously, you win.

The question I keep coming back to is: who pays for all of this in the end? At $750 billion in capex, the revenue math has to work out to something truly extraordinary. Either AI becomes the most profitable software category in history, or someone is going to have a very rough day when the bills come due.

Project Camellia is a bet that OpenAI will be one of the winners. With 3.2 gigawatts and $30 billion on the table, they're going all in — and they need Presence and whatever comes after it to make the economics work.

What's your experience? Drop a comment below! 👇 Are you impressed by OpenAI's infrastructure ambitions, or does this level of spending worry you about the sustainability of the AI boom? And do you think enterprise AI agents like Presence will actually deliver on their promise?