AI Money-Burning Race Intensifies: OpenAI to Increase Compute Spending Forecast to $750 Billion, AMD Vying for AI Server Orders

BlockBeats News, July 22nd, OpenAI has increased its expected computing resource expenditure by 2029 from $600 billion to around $750 billion. The company has also launched the Camellia Data Center project in Georgia, USA, with an initial investment of $20 billion. It is expected to receive 3.2 gigawatts of power supply from 2028 to 2032, and the total project cost may exceed $30 billion.

Previously, OpenAI had signed a 6-gigawatt data center capacity agreement with Oracle and expanded its years-long collaboration with AWS to $138 billion. Moreover, it plans to add $250 billion in cloud computing investment to Microsoft Azure. OpenAI's CFO, Sarah Friar, expressed concerns that if revenue growth cannot match expenditures, the company may struggle to sustain the cost of expanding computing power.

On the other hand, AMD has reached a multi-billion-dollar server collaboration agreement with Anthropic. Starting from 2027, Anthropic will purchase up to 2 gigawatts of AMD Instinct MI450 series chips; once the project achieves specific deployment goals, AMD will invest up to $5 billion in Anthropic. Currently, Anthropic uses Google TPU, Amazon Trainium, and NVIDIA GPU simultaneously. This collaboration will further diversify its chip supply sources.

Furthermore, Elon Musk's SpaceXAI is constructing at least one large-scale data center in Texas to expand its AI computing power beyond its existing facility in Memphis, Tennessee.