OpenAI Raises 2030 Compute Spending Forecast to $750 Billion, Launches $20 Billion Data Center Project

OpenAI has significantly increased its projected spending on computing power through 2030 to approximately $750 billion, up from an earlier estimate of $600 billion, signaling a dramatic escalation in the capital required to sustain its AI model development and deployment. The revised forecast, reported by The Wall Street Journal, reflects a series of new deals with cloud providers as the company continues to secure the massive computational resources necessary for training and running advanced AI systems. This marks the second major adjustment to OpenAI’s compute budget this year, following CEO Sam Altman’s earlier public statement that the company planned to spend $1.4 trillion, which caused external unease and prompted CFO Sarah Friar to clarify that the actual projected spending through 2030 was around $600 billion.

On the same day, OpenAI announced a $20 billion investment to launch the construction of a data center project named "Project Camellia" in Savannah, Georgia, marking its first foray into building its own data center as the lead designer and primary builder. This move represents a pivotal shift in OpenAI’s business model, transitioning from a compute tenant that rents chip capacity from providers such as Oracle and Amazon Web Services to an infrastructure developer that controls its own hardware and facilities. Sachin Katti, OpenAI’s Vice President of Compute Strategy, explained the rationale behind the shift, stating that after going through two or three generations of compute iterations, the company has learned many lessons on how to design these facilities to cut costs and speed up construction, which is why it is stepping up to lead the design and entire process.

OpenAI’s goal with Project Camellia is to standardize its AI data center designs and publicly release them for other companies to use, effectively open-sourcing hardware, networking, and data center designs in a model similar to the Open Compute Project co-founded by Meta in 2011. To support this vertical integration, OpenAI has reportedly hired Brent Mayo as Head of Data Center Construction and Delivery. Mayo left xAI earlier this year, where he was one of the core architects of Elon Musk’s data center construction projects and led the rapid procurement, commissioning, and deployment of large-scale AI chip clusters for xAI’s Colossus supercomputing campus in Memphis. At OpenAI, Mayo will be responsible for ensuring that cloud partners advance construction on schedule while also participating in advancing the new facility in Georgia.

This strategic shift towards proprietary infrastructure represents a significant realignment of OpenAI’s long-term operational strategy and capital allocation. By standardizing designs and leading construction, the company aims to optimize costs and deployment speed, reducing its reliance on third-party cloud providers and gaining more control over the compute resources that are critical to its AI ambitions. The increased compute spending forecast, now at $750 billion, underscores the sustained demand for AI model development and the escalating costs associated with scaling AI infrastructure, a trend that is reshaping the competitive landscape of the AI industry.