OpenAI plans Project Camellia with 3.2GW Georgia power deal

OpenAI has announced Project Camellia, a long-term data center project in Effingham County, Georgia, backed by a 3.2 gigawatt power contract with Georgia Power. The company says power delivery is expected in phases from 2028 through 2032, making the site less of a routine campus expansion and more of a direct bet on future frontier-model capacity.

The announcement matters because compute is now one of the hardest constraints in AI. Model releases, enterprise agents, video generation, scientific workloads, and long-running coding systems all compete for the same power, land, networking, cooling, and hardware supply chain.

Project Camellia shifts OpenAI closer to owning the compute stack

OpenAI describes Project Camellia as a data center it is designing and developing in Effingham County. That wording is important: this is not only a cloud-capacity purchase or another third-party hosting note. It points to more direct control over where OpenAI's future systems run, how capacity is staged, and how infrastructure tradeoffs are explained to local communities.

The 3.2GW figure is the clearest signal. For context, that is far beyond a single building or normal enterprise colocation footprint. It places the project in the same conversation as the largest AI infrastructure programs, where electricity procurement, grid timing, permitting, local incentives, and water design become product constraints.

OpenAI says the Georgia Power supply will arrive between 2028 and 2032. Readers should treat that as a staged infrastructure runway, not immediate model capacity. The practical impact will show up over years as OpenAI tries to translate long-lead utility commitments into usable training and inference supply.

OpenAI is leading with community promises because data centers are political now

The company is trying to frame Camellia as a local development plan, not only an AI compute project. OpenAI says it will work with Effingham County on jobs, training, community investment, education access, and infrastructure accountability. It also says residents should not see higher electricity bills because of the project and that it will curtail its own power use during peaks before local supply is affected.

That language is not decoration. Data center proposals are meeting harder resistance across the United States as residents question grid strain, noise, water use, tax incentives, and whether local benefits match the scale of public and environmental costs. Axios notes that OpenAI is positioning the Georgia project as a model for responsible data center development during that backlash.

OpenAI also says Project Camellia will use a closed-loop water cooling system intended to limit ongoing water demand. The claim is useful, but it still needs local scrutiny over construction-phase use, backup systems, discharge rules, utility planning, and long-term reporting.

The unresolved question is how 3.2GW gets powered

The biggest open detail is the energy mix. OpenAI named Georgia Power and the phased delivery window, but the initial announcement does not fully answer what generation, transmission upgrades, storage, demand-response arrangements, or clean-energy procurement will support the load.

TechCrunch reported that OpenAI and Georgia Power had not immediately provided specifics on how the project would be powered. Until those details are public, Camellia is best read as a confirmed capacity and siting move with unresolved energy-policy consequences.

That uncertainty matters for developers and AI buyers because power delivery is now tied to product availability. If infrastructure buildouts lag, high-demand models can remain throttled, expensive, waitlisted, or regionally uneven. If they land on schedule, large labs can support more video, agents, research, and enterprise workloads without every launch turning into a capacity event.

What LinkLoot readers should watch next

The next useful signals are not marketing slogans. Watch for utility filings, Georgia Power resource plans, county incentive documents, construction milestones, water-permit details, and any OpenAI disclosure on how Camellia connects to its broader cloud and Stargate-style infrastructure commitments.

For builders, the headline is simple: frontier AI is becoming an infrastructure business. Model quality still matters, but the winners also need power contracts, local approvals, data center execution, and enough capacity to make advanced tools available at usable prices.

Sources and methodology

This post uses OpenAI's Project Camellia announcement as the primary source and compares it with Axios, TechCrunch, and local Effingham Herald coverage. External reports were treated as factual inputs only. The main confirmed facts are the Effingham County location, the 3.2GW Georgia Power contract, the 2028-2032 phased delivery window, and OpenAI's stated local-ratepayer, curtailment, jobs, and water-cooling claims. The energy mix and full local impact remain open verification points.