OpenAI Bets on a 3.2 GW Compute Campus in Georgia
OpenAI said it is moving ahead with the Project Camellia data center project in Effingham County, Georgia, and plans to lock in 3…
On July 22, OpenAI put down a clear marker for the next phase of the AI race near Savannah, Georgia. The company announced that it is moving ahead with the Project Camellia data center project and has signed a long-term power arrangement with Georgia Power: a phased total of 3.2 GW of electricity from 2028 through 2032.
What happened
Project Camellia is located within the existing Savannah Gateway Industrial Hub. A county notice from Effingham County said the project involves about $20 billion in capital investment, will create 400 long-term jobs, and expand the local tax base. OpenAI said the project is still in an early development stage, and that infrastructure, construction timing, design, financing, and the operating model have not yet been finalized.
The power arrangement is the centerpiece of the investment. Georgia Power said the project is expected to require about 3,200 MW of electricity and has signed a 25-year agreement. As part of the arrangement, OpenAI has committed to provide up to 1,000 MW of flexible load during peak demand periods, meaning it will proactively reduce some of its electricity use when the grid is under maximum strain. Georgia Power said this will help the system avoid building new generation resources for short peak-demand windows.
OpenAI also outlined a series of community commitments, including covering the infrastructure and electric interconnection costs needed to serve the facility, providing $80 million in community support over the project’s life cycle, and using a closed-loop water system. Effingham County has scheduled a public information meeting for the evening of July 23, where residents can ask questions about power, water use, jobs, and land development.
Why it matters
This is not an ordinary data center. 3.2 GW means AI infrastructure is entering local power markets as a large industrial load. In the past, investors discussing AI often focused on chip purchases, model revenue, and cloud orders; now grid access, long-term power contracts, demand response, and local tax incentives are becoming equally important capex variables.
Project Camellia also points to a new infrastructure financing logic: model companies are no longer just renting cloud resources, but locking in future compute supply years in advance through long-term contracts, dedicated campuses, and local partners. For OpenAI, this helps ease compute shortages that limit model training and deployment; for utilities, data centers bring stable large customers, but also require stricter risk segregation and load management.
The project’s regional impact is equally direct. Georgia wants the investment, jobs, and tax revenue, while local communities need to ensure that the added costs of power, land, and water do not get passed on to residents. OpenAI has said the company will cover the infrastructure and electric service costs associated with the project, but how those commitments are written into contracts and audited will determine their credibility.
What to watch next
First, how the 3.2 GW buildout will be phased, and whether the actual construction scale changes with financing and model demand. Second, whether the 1,000 MW demand response can be reliably executed during peak periods, and how the Georgia Public Service Commission and utility customer protection mechanisms will oversee it. Third, whether the project’s tax incentives, land development, and environmental review can withstand public scrutiny. Finally, whether OpenAI replicates a similar self-built compute model in other states will affect the competitive landscape for U.S. AI infrastructure and could further deepen the interdependence among power, chips, cooling, and data center financing.
Sources
Information only. No investment, legal, tax, or financial advice.