OpenAI's $750B infrastructure bet tilts toward Georgia as Stargate stalls

OpenAI’s $750B capex revision signals infrastructure acceleration

OpenAI raised its total infrastructure spending forecast to $750 billion through 2030—a 25% increase from earlier projections—according to The Wall Street Journal, as reported by TechCrunch AI on July 22. The revised target arrives as the company’s Stargate data center initiative appears to have stalled, repositioning OpenAI’s near-term expansion around Project Camellia, a $20 billion data center campus in Effingham County, Georgia, northwest of Savannah.

The Georgia facility will occupy 1,400 acres and is designed to draw at least 3.2 gigawatts of peak power from Georgia Power, the regional utility. According to TechCrunch AI, the electric generating capacity is expected to become available between 2028 and 2032, though the source does not specify when GPU infrastructure will be operational. OpenAI has committed to bearing all infrastructure and electric-service costs, shielding existing customers from rate increases under a Georgia Public Service Commission rule enacted in 2025 that blocks utilities from passing costs associated with new large users consuming more than 100 megawatts.

Power supply and regulatory constraints

Georgia Power received approval in December to expand generating capacity by 9,885 megawatts; the OpenAI contract accounts for approximately one-third of that contracted capacity, per TechCrunch AI’s analysis of regulatory filings. The source notes that neither OpenAI nor Georgia Power has publicly disclosed the specific power generation mix for Project Camellia. However, regulatory documents filed with the Georgia PSC indicate that Georgia Power’s new capacity will derive primarily from natural gas—roughly 5.8 gigawatts of new natural gas generation, representing more than a doubling of the utility’s existing natural gas fleet. The remainder will come from grid-scale battery storage and solar installations.

OpenAI will also reduce its power draw by up to 1 gigawatt during peak grid-demand periods, a demand-response provision that benefits regional grid stability. Additionally, the company is receiving a 50% property tax abatement from Effingham County for 15 years, according to TechCrunch AI’s reporting on filings with the Effingham Herald.

Why This Matters

The $750B capex commitment reflects OpenAI’s determination to secure dedicated, geographically distributed compute infrastructure independent of cloud providers. However, the multi-year timeline (electricity not available until 2028–2032) creates operational risk: frontier model training and deployment schedules may outpace facility completion. Grid operators and utility regulators in the Southeast now face unprecedented load forecasts; a 3.2+ GW facility sets a precedent for how regional grids accommodate AI workload clustering. Teams evaluating AI vendor selection and geographic expansion must account for power-market constraints and utility-approval timelines—a process that has historically taken 18–36 months. The reliance on natural gas for the majority of new generating capacity also signals that OpenAI’s scaling priorities currently override carbon-intensity optimization, a calculus that may shift if regulatory frameworks (such as EPA emissions standards or state-level grid decarbonization mandates) tighten.