OpenAI plans to spend $30B+ on a massive new data center in Georgia, securing 3.2GW of energy, with several hundred MWs set to come online starting in 2028
Context & Ripple Effects
The Georgia project turns OpenAI’s earlier compute commitments into a site-specific build: the company had said it signed contracts for 10GW of US AI compute capacity, and now has identified 3.2GW of energy capacity for one facility. Initial capacity is not expected until 2028, underscoring the long lead time between contracting compute and operating it.
It also sits alongside a higher projected cloud-spending plan through 2030, making the Georgia commitment evidence that OpenAI’s infrastructure strategy is moving from broad capacity targets toward large, capital-intensive deployments.
First-order effects
- OpenAI takes on a $30B-plus infrastructure commitment and a 3.2GW power allocation, with several hundred megawatts scheduled to begin coming online in 2028.
- The project gives OpenAI a defined path to add dedicated capacity, while tying part of its future compute availability to delivery of the Georgia facility and its energy infrastructure.
Second-order effects
- A commitment of this scale increases the importance of power delivery, construction, and equipment execution for OpenAI’s compute roadmap; delays in any of those inputs would affect when contracted capacity becomes usable.
- Other AI developers and cloud providers seeking comparable scale face a more constrained competition for large, long-duration power-backed data-center capacity, rather than compute alone.
Third-order effects
- AI infrastructure is increasingly being planned like utility-scale industrial capacity: large power commitments and multiyear construction timelines become strategic inputs to model development and service growth.
- If such projects continue to proliferate, the sector’s advantage may shift further toward organizations able to finance and execute long-lived infrastructure commitments, though actual utilization will determine how durable that advantage is.
The trend: The story is one data point in AI compute becoming a power-constrained, long-duration infrastructure business rather than a purely cloud-procurement market.