OpenAI Project Camellia — $20 Billion Georgia AI Data Campus
- Jul 23
- 4 min read
OpenAI has unveiled plans for one of the largest computing projects in American history: Project Camellia, a roughly $20 billion artificial-intelligence data center campus spread across 1,400 acres in Effingham County, Georgia, with some reports putting the eventual investment above $30 billion once every phase is complete. The announcement instantly makes the quiet county outside Savannah one of the most important construction sites in the global AI race.
The scale of the project is difficult to overstate. Project Camellia is designed to draw up to 3.2 gigawatts of power — roughly the output of three nuclear reactors and enough electricity to supply well over two million homes. The campus, planned for the Savannah Gateway industrial corridor, would rank among the largest single-site computing facilities ever attempted anywhere in the world, dwarfing most existing hyperscale footprints.
The Georgia campus is only one piece of an infrastructure blitz that keeps accelerating. Alongside the data center announcement, OpenAI launched a new enterprise agent platform this week, and the company has reportedly raised its projected compute spending through 2030 to roughly $750 billion, up from an earlier estimate of $600 billion — a figure that rivals the inflation-adjusted cost of the entire Apollo program and exceeds the annual GDP of most countries.
For Georgia, the project promises thousands of construction jobs, hundreds of permanent technical positions, and a transformed tax base for Effingham County. State and local officials courted the deal as the biggest economic-development win for the region since Hyundai’s electric-vehicle plant landed nearby, and the Savannah area’s deepwater port access, fiber routes and available industrial land made it a natural finalist for the site selection.
Not everyone is celebrating. Effingham County residents have already begun pushing back, packing local meetings with questions about water consumption, electricity rates, construction noise, and what a 1,400-acre industrial campus will do to the rural character of their communities. Local television coverage this week captured a county wrestling with the classic data-center bargain: enormous headline investment, comparatively modest permanent employment, and very large utility footprints that neighbors live with forever.
The power question looms largest. Georgia’s grid is already straining under a wave of data-center demand, and a single 3.2-gigawatt customer would rank among the biggest electricity users in the entire Southeast. Utility analysts expect the project to require new generation capacity, new transmission lines, and years of negotiation over who ultimately pays for the buildout — the same questions that have followed hyperscale campuses in Virginia, Texas and Arizona.
Water is the second flashpoint. Modern AI campuses can consume millions of gallons a day for cooling, and coastal Georgia’s aquifers are already managed under strict withdrawal limits. OpenAI has said its newer facility designs rely heavily on closed-loop and air cooling to slash consumption, but county officials say detailed, binding environmental commitments will be central to the permitting fights ahead.
Project Camellia also lands in the middle of a raging Wall Street debate about AI capital expenditure. The same week OpenAI confirmed the Georgia campus, Alphabet’s earnings showed another jump in AI capex that helped knock the Nasdaq down more than 2% in a single session. Investors are increasingly split between those who see a generational infrastructure buildout and those who see an arms race outrunning any plausible near-term revenue.
OpenAI’s answer to the skeptics is demand. The company argues that compute scarcity — not customer appetite — is the binding constraint on its business, with enterprise adoption of AI agents growing faster than capacity can be provisioned. The new enterprise agent platform announced alongside Camellia is aimed squarely at that market, betting that autonomous software workers become the next major line item in every corporate budget.
The competitive context is just as stark. Microsoft, Google, Amazon, Meta and Anthropic are all racing to lock up power, land and chips through the end of the decade, and Nvidia’s newest GB300 systems have become objects of open geopolitical struggle — the White House this week accused China’s Moonshot AI of obtaining them through third countries. Compute has become strategic national infrastructure, and Camellia is America doubling down on home soil.
History offers reasons for both optimism and caution. The railroad and telecom booms built infrastructure that powered decades of growth — and bankrupted many of the companies that built it. Whether Project Camellia becomes Georgia’s Hoover Dam or its Iridium moment depends on questions no one can yet answer: how fast AI revenue compounds, how cheap intelligence becomes, and whether the grid can keep pace with the ambition.
What happens next: county permitting hearings through the fall, utility filings detailing the power arrangements, and a groundbreaking targeted for next year, with the first computing halls expected online before the end of the decade. Residents will get their say at public hearings, regulators will get their filings, and the market will keep score every quarter.
One thing is already certain — the center of gravity of the AI buildout is no longer confined to Silicon Valley or northern Virginia. It is pouring concrete in places like Effingham County, Georgia, one gigawatt at a time.
Comments