Summary: OpenAI unveiled Project Camellia on July 22, 2026 as a 3.2 GW Georgia data center project with phased service from 2028 through 2032 and committed private financing. Its rate protection, closed-loop cooling, community benefits, and annual audits form a public accountability template, but the project is early and the financial and jobs figures remain OpenAI claims or estimates.
On July 22, 2026, OpenAI unveiled Project Camellia: a long-horizon data center project in Effingham County, Georgia, backed by a 3.2 GW power contract with Georgia Power Company, with service arriving in phases between 2028 and 2032, and financed — per the commitment — entirely with private capital. For anyone tracking AI infrastructure, the close read matters less for the scale than for the method: the announcement translates “how does a hyperscale data center earn its social license” into an auditable checklist, where power rates, the grid, water, community benefits, and audits each carry a mechanism instead of a press-release flourish.
3.2 GW and the Two Layers of “No Pass-Through”
The power contract itself holds no surprises: 3.2 GW, phased delivery, service between 2028 and 2032. Where the real design effort went is the price guarantee, and it is doubled up. The contractual layer: OpenAI covers the full cost of the infrastructure and electric service needed to power the project, rather than letting the existing rate structure absorb the hit. The regulatory layer: under Georgia Public Service Commission rules, those costs are barred from being passed to existing ratepayers in the first place. Stack the two, and “Georgia households will not see their electricity bills rise because of this project” graduates from verbal pledge to rule-backed statement.
Notice the symmetry, too. OpenAI sought neither a special rate nor a subsidy; it pays its own way in exchange for certainty. Set against the friction that often defines large-load-to-grid relations, the posture is deliberately cooperative — the price is internalized cost, and the reward is watching the most common objection (higher bills) get dismantled at the source.
Peak-Shaving by Design: The Data Center as a Grid Participant
The second power commitment is the rarer breed. OpenAI describes Project Camellia as belonging to the first wave of data centers engineered to proactively dial down their own consumption during high-demand periods — curbing load before residential customers feel anything, in support of Georgia Power’s grid reliability. Directionally, the facility is rewritten from pure load into a participant the grid can lean on: cooling schedules and the timing slack of deferrable batch work both count as movable levers.
What the announcement leaves open are the operating specifics — what trips a curtailment, by how much, and how it gets verified. That is precisely question number one for the annual audits to chase: a design aspiration to shed load and a verifiable operating obligation are two very different animals.
Closed-Loop Cooling: No Water Competition with the Community
The water answer is just as concrete. Cooling runs on a closed-loop design that recirculates water instead of pulling and discharging it endlessly — the announcement’s own analogy is a car radiator: you top off losses, you never drain and refill. OpenAI further commits that water will not be diverted away from local communities; day-to-day demand shrinks to workplace scale — restrooms, kitchens, cleaning and maintenance — on par with an office building staffed at the same headcount. Measured against the evaporation-heavy cooling disputes that trail data centers around, this erases the biggest source of water-footprint argument while the project is still on the drawing board.
$80M in Community Benefits and Up to $71M in Codex Credits
The community package is where the announcement turns most concrete (every figure below is an OpenAI claim or estimate, not independently verified as of the announcement date):
| Item | Content | Amount / scale |
|---|---|---|
| Community benefits | Schools, public safety, healthcare, housing, workforce training, veterans, small businesses | $80M (over the project’s life) |
| Codex education credits | Georgia college, community college, and technical school students | Up to $71M |
| Taxes | State and local; expected to be the county’s largest taxpayer | Hundreds of millions (estimate) |
| Jobs | Construction and permanent on-site roles; local contractors prioritized | Thousands (estimate) |
The Codex credits merit a longer look: eligible students receive $100 apiece through their ChatGPT accounts, spendable on Codex, the agentic coding tool OpenAI builds into ChatGPT. For OpenAI the outlay doubles as education spend and ecosystem spend — the next developer generation grows up inside its toolchain; for Georgia students it is a zero-cost entry ticket to hands-on agent work. Past the dollar figures, OpenAI also commits to giving local contractors and businesses first call, keeping construction-period value inside the county.
Accountability: Annual Independent Audits and a Community Compact
Commitments grow teeth only when something verifies them. OpenAI pledges a publicly released audit each year from an independent firm — whether the four core promises (rates, grid support, water, community benefits) hold up now gets a checkable annual baseline. On the procedural side sits participatory governance: following the July 23 public open house, resident input and the conversations after it will shape a Georgia Community Compact translating community priorities into concrete commitments and accountability measures.
The sequencing is deliberate: host the open house, collect the input, then write the input into a document — not finalize first and explain later. What the Compact finally contains, and how much force it carries, remains to be seen; still, the pairing of a documented compact with external annual audits already puts the project a step ahead of most peers.
Three Takeaways for AI Infrastructure Watchers
First, the siting logic is itself an argument. The parcel sits inside the Savannah Gateway Industrial Hub, a pre-existing industrially zoned warehouse development; the comparison OpenAI draws is not “data center vs. empty land” but “data center vs. the warehouses that would otherwise occupy the same ground” — less traffic, less water, stronger long-term tax revenue, more durable on-site jobs. Choosing warehouses as the counterfactual stacks the social-cost comparison in the data center’s favor.
Second, the funding structure leads: the project is entirely privately financed, and OpenAI points to its Abilene, Texas campus already running at scale as the existence proof, softening the suspicion that this is more promise than delivery.
Third — the limits. The project remains in early development, with substantial work outstanding on infrastructure, phasing, design, financing, and the operating model; every figure above is an OpenAI claim or estimate, and “largest taxpayer in the county” is explicitly an expectation. For builders the practical lesson is the pattern, not the numbers: as the politics of siting AI compute keep tightening, folding rate guarantees, water design, education investment, and public audits into the first announcement is shifting from differentiator to entry ticket. If Abilene and Camellia keep delivering, this checklist becomes the baseline future projects get measured against.