Morning Edition · Thursday, July 23, 2026Published at 2:03 AM EDT · New York
OpenAI Unveils $20 Billion, 3.2-Gigawatt Georgia Data Center It Will Design and Build Itself
Project Camellia would span 1,400 acres near Savannah, drawing power from Georgia Power in phases from 2028 to 2032, and is what OpenAI calls its first self-built facility.
OpenAI announced Project Camellia, a $20 billion data center campus in Effingham County, Georgia, that the company says it will design and build itself rather than lease from a cloud partner. The campus is planned as four buildings across 1,400 acres at the Savannah Gateway Industrial Hub, about 12 miles from the port of Savannah.
The binding constraint is electricity, not chips. DataCenterDynamics reports the site is designed for 3.2 gigawatts, with Georgia Power delivering capacity in phases from 2028 through 2032. OpenAI says it will fully fund the required grid infrastructure so existing ratepayers do not subsidize the build, and it is pairing the project with an $80 million community fund and up to $71 million in Codex credits for local students.
Owning the facility changes OpenAI's cost structure. A self-built, self-powered campus lets the company control the performance-per-megawatt tradeoff directly and amortize capital over the life of the hardware instead of paying a cloud margin. It also concentrates a large, single-tenant power demand onto one regional utility, which is where disputes over local rates and permitting will arise.
What this means
OpenAI moving from renting cloud capacity to owning generation-scale infrastructure is the clearest signal yet that frontier labs treat power procurement, not chip supply, as the deciding input. The exposed parties are regional utilities and ratepayers who absorb the grid buildout, and the beneficiaries are Nvidia and the electrical-equipment and construction supply chain that a 3.2-gigawatt campus creates demand for. The channel is capital expenditure and long-dated power contracts that lock in demand through 2032.
What to watch
- Whether Georgia Power's rate filings show the cost of new capacity falling on OpenAI alone or spreading to other customers, which decides how politically durable these self-built campuses are.
- Whether other labs announce self-owned rather than leased facilities, a shift that would confirm compute ownership is becoming a strategic advantage rather than an operating expense.
Observations to monitor, not financial advice.
Synthesized from: OpenAI · DataCenterDynamics
Part of a tracked trend
Enterprises Build Owned Accelerator Clusters
Regulated industries increasingly buy dedicated frontier-scale accelerator infrastructure rather than renting cloud compute, sustaining demand for the newest Nvidia platforms and making performance-per-megawatt the deciding spec.
More from this edition
- Judge Approves Anthropic's $1.5 Billion Settlement Over Pirated Books Used to Train Claude
- New Papers Argue Safe Language Models Do Not Compose Into Safe Multi-Agent Systems
- NVIDIA Commissions the U.S. Military's First DGX GB300 at the Naval Postgraduate School
- OpenAI Launches Presence, an Enterprise Platform for Voice and Chat Agents
- Google Commits $40 Million to the Energy Department's Genesis Mission as 270 Projects Begin
- Anthropic Lets Users Teach Claude a Skill by Recording Their Screen
- Tesla Will Record Grünheide Workers' Movements to Train Optimus
- A New Benchmark Tries to Measure Power-Seeking in Frontier Models
- Study Finds Supervised Fine-Tuning Collapses Behavioral Diversity in Sequential Decisions
- Researchers Propose Subquadratic Operators That Keep a Global View of Multi-Dimensional Data
- NVIDIA Unveils a Synthetic Video Detector That Scores 1080p Footage in 22 Milliseconds