OpenAI is once again picking up speed in expanding its artificial intelligence (AI) infrastructure after a period of slowing down. The company plans to increase its AI compute infrastructure investment to about 1.1 quadrillion won by 2030 and is pushing to strengthen its AI infrastructure by building ultra-large data centers.
On the 22nd (local time), foreign media including TechCrunch said OpenAI announced it will launch "Project Camellia" to build a long-term data center with a capacity of 3.2 GW (gigawatts) in Effingham County, Georgia. One gigawatt is equivalent to the generation capacity of one nuclear power plant and can supply electricity to about 1 million households.
The project is drawing attention because OpenAI, which had delayed "Stargate," a U.S. AI infrastructure build-out pursued with SoftBank and Oracle, and had withdrawn from overseas data center projects in the United Kingdom and Norway, is returning to large-scale infrastructure investment. OpenAI will first invest $20 billion in Project Camellia, and more than $30 billion (about 44 trillion won) is expected to be spent to complete the 3.2 GW data center.
This time, unlike Stargate, OpenAI will take a more active lead across the project, including facility planning and design. Sachin Katti, OpenAI vice president, told Bloomberg News, "We have learned many lessons about design approaches that can reduce expense and increase speed as we build compute systems over two to three generations," adding, "We are now directly leading the design and the entire build process."
As the AI infrastructure expansion gains momentum, the scale of investment is also growing. The Wall Street Journal (WSJ) reported that, citing sources, OpenAI raised its forecast for AI compute infrastructure-related expenditure through 2030 to $750 billion (about 1.1 quadrillion won) from the previous $600 billion. OpenAI has also reportedly hired Brent Mayo, who oversaw data center expansion at Elon Musk's AI corporations xAI.
Concerns are emerging about the massive investment. Sarah Friar, OpenAI chief financial officer (CFO), is said to have internally raised concerns that it may be difficult to bear the hefty AI compute infrastructure contract expense if revenue does not increase as quickly as expected.