Computing Demand Explodes! OpenAI Boosts AI Infrastructure Investment, Cloud Spending Forecast to Hit $750 Billion by 2030

According to The Wall Street Journal, citing sources familiar with the matter, OpenAI has increased its projected spending on cloud computing and data center infrastructure from an earlier estimate of approximately $600 billion to around $750 billion by 2030—an increase of $150 billion. This reflects the company’s ongoing expansion of computing power (compute) required for AI model training and inference, as well as its accelerated efforts to build proprietary data centers to reduce dependence on external cloud service providers.

The report indicates that this upward revision in capital expenditure is primarily driven by recent long-term agreements signed with multiple cloud service providers to secure substantial computing resources for AI model development and operations over the coming years. Compute investment has become one of the top strategic priorities under CEO Sam Altman’s leadership and is a key topic as the company prepares for a potential initial public offering (IPO).

Sources reveal that the high level of compute spending has sparked internal discussions. CFO Sarah Friar has expressed concerns about the scale of long-term compute contracts, warning that if revenue growth fails to meet expectations, the company may struggle to sustain such massive infrastructure expenditures.

This week, OpenAI announced it will invest $20 billion to launch 'Project Camellia,' an AI data center initiative in Effingham County, Georgia. This marks OpenAI’s first large-scale data center project led and designed in-house, rather than simply leasing compute capacity from cloud providers.

Sachin Katti, OpenAI’s Vice President of Compute Strategy, stated that the company has signed a power supply agreement with Georgia Power, aiming to secure 3.2 gigawatts (GW) of electricity between 2028 and 2032—enough to support large-scale AI computing demands.

Katti added that OpenAI has completed land acquisition for the site and is currently selecting partners to handle construction and operations. The company has held multiple discussions with local government, school districts, and community representatives, and the project has received support from regional economic development agencies and local authorities.

The 3.2 GW of power allocated for the Georgia facility will be additional to existing collaborations and will not affect other ongoing large-scale cloud partnerships.

Currently, OpenAI has a roughly 6 GW data center compute partnership with Oracle (ORCL-US), most of which is still under construction. In late February, the company expanded its multi-year agreement with Amazon Web Services (AWS), a subsidiary of Amazon.com (AMZN-US), to $138 billion over eight years. This includes 2 GW of computing resources powered by AWS’s in-house Trainium AI chips.

Additionally, last year, OpenAI committed an additional $250 billion to purchase cloud computing services from Microsoft’s (MSFT-US) Azure cloud platform, though the spending timeline was not disclosed.

To accelerate data center development, OpenAI has recently expanded its infrastructure team. Sources indicate the company has hired Brent Mayo, who previously worked on the supercomputer project at Elon Musk’s xAI, as Head of Data Center Construction and Delivery.

Mayo led the construction of xAI’s Colossus supercomputer center in Memphis, Tennessee, overseeing the rapid deployment of large-scale AI chip clusters. After leaving xAI earlier this year, he joined OpenAI and will oversee partner-led data center projects while also contributing to Project Camellia.

Mayo will report to Uday Ruddarraju, who was recently promoted to Chief Technology Officer of Computing Capacity at OpenAI. Ruddarraju also worked on the xAI Colossus project and joined OpenAI last year, now reporting directly to President Greg Brockman.

Over the past year, OpenAI has continuously adjusted its infrastructure strategy. In January 2025, Altman announced the 'Stargate' initiative at the White House, originally planning a collaboration with SoftBank Group and Oracle to invest up to $500 billion over four years in U.S. AI infrastructure. However, progress has been slower than expected.

Now, OpenAI is restarting its self-built data center strategy, aiming to gain greater control over AI infrastructure rather than relying entirely on third-party cloud providers.

Notably, Altman previously stated publicly that the company might invest as much as $1.4 trillion in AI compute, sparking market concerns over OpenAI’s financial burden. Friar later downplayed the statement, telling investors that the actual spending plan through 2030 was around $600 billion. The latest forecast of $750 billion shows that investment in AI infrastructure continues to expand rapidly.

Market analysts suggest that OpenAI’s significant increase in compute investment reflects how the global AI race has shifted from model capabilities to foundational infrastructure such as data centers, power, and chips. As tech giants like OpenAI, Microsoft, Amazon, Alphabet, and Meta Platforms continue to expand their AI data centers, demand for GPUs, AI servers, high-bandwidth memory (HBM), power equipment, and liquid cooling systems is expected to remain strong over the next several years.