The geography of intelligence is moving south, and it's taking $30 billion with it.
The Summary
- OpenAI is building a data center near Savannah, Georgia with $30+ billion in planned investment, marking one of the largest AI infrastructure commitments to date
- This isn't just expansion. It's OpenAI betting that AI demand will outpace even ChatGPT's explosive growth trajectory
- The location signals a shift: AI infrastructure is chasing power, cooling, and tax incentives, not proximity to Silicon Valley
The Signal
$30 billion is nation-state money. For context, that's roughly what Taiwan spent building a single advanced semiconductor fab. OpenAI is now spending at the scale of countries, not startups. This is the clearest signal yet that training and running frontier AI models requires physical infrastructure at a magnitude most people still don't grasp.
Georgia makes sense when you map the constraints. The state offers industrial power rates, access to nuclear and natural gas baseload generation, and coastal proximity for cooling infrastructure. Savannah's port means equipment can arrive without truck bottlenecks. Most importantly, Georgia's tax incentives for data centers are among the most aggressive in the U.S. OpenAI isn't picking Savannah for the peaches.
"AI infrastructure is chasing power, cooling, and tax incentives, not proximity to Silicon Valley."
The timing matters. OpenAI is building this as inference costs remain stubbornly high and as competitors like Anthropic, Google, and X all race to scale their own compute. Every major AI lab is now in an arms race measured in gigawatts, not just parameters. The company that can serve more requests, faster, with lower latency wins commercial contracts. Enterprise customers don't care about your AGI philosophy. They care if your API goes down during their product launch.
This also reshapes the agent economy. If you're building AI agents that need to make millions of API calls per day, the location and capacity of your provider's data centers directly impacts your unit economics. A $30 billion facility means OpenAI expects to serve orders of magnitude more agent traffic than exists today. They're building for a world where:
- Every company runs dozens of specialized agents
- Inference demand exceeds training demand
- Latency-sensitive applications require regional compute
- Real-time AI becomes table stakes, not a luxury feature
The second-order effects hit regional economics hard. Construction jobs, long-term data center operations roles, and the supply chain around cooling, power management, and networking all concentrate in Savannah now. This is how AI wealth gets distributed geographically. Not through remote work. Through data centers that need to be near cheap power and fiber lines.
The Implication
If you're building AI agents or products that rely on frontier models, watch where the compute goes. Latency matters. Provider stability matters. OpenAI spending $30 billion means they expect you to send a lot more requests than you're sending today.
For cities and states, this is the new economic development playbook. Offer power, fiber, tax breaks, and permitting speed. The AI labs will come. The question is whether local communities get jobs and tax revenue, or just higher power bills and stressed grids.