OpenAI just raised the stakes on its infrastructure buildout in a big way. The company announced Wednesday that it will spend $750 billion on infrastructure through 2030, roughly 25% more than it estimated earlier this year, according to TechCrunch AI, which cited a Wall Street Journal report. That’s not a typo. Three-quarters of a trillion dollars, and the spending is accelerating even as OpenAI’s flagship Stargate data center project appears to have stalled.
Here’s what the number actually buys, and why it should get your attention.
🏗️ First stop: Georgia
The opening move is Project Camellia, a $20 billion data center campus northwest of Savannah. TechCrunch AI reports the development will span 1,400 acres and pull at least 3.2 gigawatts of power from Georgia Power, the regional utility. That generating capacity is expected to come online between 2028 and 2032.
OpenAI says it will “pay the full cost of the infrastructure and electric-service costs” for the site. That matters because Georgia’s Public Service Commission passed a rule last year to stop utilities from passing large-user costs onto regular ratepayers. OpenAI also agreed to cut its power draw by up to 1 gigawatt when the grid is under strain, and it’s collecting a 50% property tax break for 15 years from Effingham County.
⚡ The power question nobody’s answering
Neither OpenAI nor Georgia Power has said how Camellia will actually be powered. TechCrunch AI asked both and didn’t get a reply. But the regulatory filings tell a story.
- In December, Georgia Power got approval to produce an additional 9,885 megawatts, and expects all of it contracted by the end of 2026.
- The OpenAI deal alone accounts for about a third of that new capacity.
- Most of it comes from natural gas. Georgia Power plans to build or buy about 5.8 gigawatts of gas generation, a quarter of it from more polluting simple-cycle turbines.
- The rest comes from grid-scale batteries and solar.
The upshot: this deal will more than double Georgia Power’s natural gas fleet. So much for the clean-energy framing that usually surrounds AI announcements.
🏎️ Built fast, questions later
Electricity is slated to start flowing in 2028, but OpenAI hasn’t said when the first GPU turns on. It could be sooner. The company recently hired Brett Mayo to lead data center construction, and his resume is telling. Mayo previously ran xAI’s Colossus data center in Memphis, which was built in record time.
That speed came at a cost. Colossus is now the subject of a lawsuit from the NAACP and the Southern Environmental Law Center, which allege it has run dozens of unpermitted natural gas turbines while claiming exemption from federal clean air rules. Hiring the person who oversaw that project signals OpenAI wants Camellia standing quickly, and it’s a hint about the environmental fights that may follow.
💡 Why this matters
What stands out here isn’t just the dollar figure. It’s the direction. OpenAI is raising its spending target at the same moment Stargate, its marquee data center effort, looks stuck. That tells you the company is rerouting rather than retreating, and it’s willing to plant capacity wherever it can secure power and tax breaks fastest.
A few things worth tracking:
- Power is the real bottleneck. Compute isn’t limited by chips alone anymore. Gigawatts, grid contracts, and utility approvals now gate how fast anyone can scale. Camellia is a case study in that shift.
- The clean-energy story is thinning. Doubling a utility’s gas fleet to feed one AI campus is going to draw scrutiny from regulators, environmental groups, and ratepayers.
- Speed invites lawsuits. The Colossus playbook got results and got sued. Expect the same tension around Camellia.
For anyone building on OpenAI’s models, the practical read is straightforward. The company is committing capital at a scale that assumes demand keeps climbing for years. That’s a bet on you needing far more compute, and it’s laying the physical groundwork now to sell it.
More details are available at the original TechCrunch AI report.