OpenAI has made the boldest infrastructure move in its history: the company announced 'Project Camellia,' its first data center campus designed, financed, and operated entirely in-house, located in Effingham County, Georgia, near Savannah. The initial investment stands at $20 billion with the potential to exceed $30 billion across construction phases running from 2028 to 2032. The campus will span 1,400 acres and will be powered by 3.2 gigawatts secured through a contract with Georgia Power — enough electricity to power roughly 2.4 million U.S. homes at full capacity. For small and medium-sized businesses relying on OpenAI's APIs, GPT-5.6, Codex, or ChatGPT Enterprise, this announcement has direct implications: more compute capacity means greater availability, fewer interruptions, and — over time — significantly lower prices.
What Is Project Camellia and Why Does It Matter Now?
Until now, OpenAI operated on Microsoft Azure infrastructure. Project Camellia marks a strategic shift: OpenAI will build and control its own compute at massive scale, partially decoupling from third-party providers to run the most advanced models — including the GPT-5.6 family (Sol, Terra, and Luna) and future o-series models. The campus will be located at the Savannah Gateway Industrial Hub, and OpenAI is already seeking financing and construction partners to execute the project in phases. Additionally, the company committed $80 million in community benefits — covering schools, public safety, healthcare, workforce training, and housing — alongside $71 million in Codex AI credits for every student at a Georgia college. This is not just a real estate play: it's a market signal that OpenAI intends to control the AI infrastructure of the next decade.
"When the company building the models also controls the data centers running them, enterprise access to AI enters a new era of scale, stability, and pricing power."
Davarion Group & LabsReal Impact for SMBs
- 01Greater API reliability: With 3.2 GW of owned capacity, OpenAI will dramatically reduce the bottlenecks causing 429 rate-limit errors and high-latency periods — critical for businesses automating customer service or workflows via GPT.
- 02Lower costs in the medium term: Owning infrastructure removes the margin OpenAI currently pays Azure. Historically, when major labs control their own compute, API prices drop 40–80% within 12–24 months of capacity going live.
- 03Long-term vendor confidence: A $30B physical infrastructure commitment signals that OpenAI plans to exist and grow for decades. For SMBs building automation stacks on OpenAI APIs, this removes the key risk of depending on a provider without durable infrastructure.
- 04Recommended action now: Begin architecting your automation workflows on GPT-5.6 models today. As Project Camellia comes online in 2028, these models will become faster and cheaper — building on them now gives you compounding ROI.
Project Camellia reveals a pattern we've already seen with Amazon, Google, and Microsoft: companies that build their own infrastructure are the ones that ultimately set market prices. With 3.2 GW, OpenAI will operate one of the world's largest AI facilities, surpassing the individual data centers of many competitors. For mid-sized businesses currently weighing whether to build automation on OpenAI or open-source models, this announcement tilts the calculus: the price stability, response speed, and scalability OpenAI will offer post-2028 will be difficult to match for providers without equivalent owned infrastructure.
At Davarion Group & Labs, we help businesses in Houston TX and across Latin America build autonomous AI agents on OpenAI's APIs and other leading AI platforms. If your company wants to get ahead of the curve — implementing today the automation workflows that Project Camellia will make even more powerful tomorrow — our team can design, develop, and deploy custom AI agent solutions tailored to your operations. Visit us at davarion.com to start your digital transformation before your competitors do.
