OpenAI Cloud Spending Plan Rises to $750 Billion

AI News

July 23, 2026

OpenAI Cloud Spending Plan Rises to $750 Billion

OpenAI has increased its projected spending on computing power to approximately $750 billion through 2030 as the company accelerates the construction and leasing of infrastructure needed to develop and operate advanced AI models.

The updated OpenAI cloud spending estimate is up from roughly $600 billion reported earlier in 2026. It reflects the increasing cost of data centers, processors, networking equipment, electricity and cloud services required to support ChatGPT and future AI systems.

Compute costs keep rising

OpenAI President Greg Brockman previously said the company expected to spend about $50 billion on computing power in 2026 alone.

Its annual computing expenses have grown from around $30 million in 2017 to tens of billions of dollars as the company trains larger models and serves a rapidly expanding number of users.

The new $750 billion projection covers expected computing expenditure through the end of the decade. It is separate from broader infrastructure commitments previously discussed by CEO Sam Altman, including plans to develop as much as 30 gigawatts of computing capacity.

New Georgia data center

OpenAI is also planning a major data-center campus in Effingham County, Georgia.

The project, known as Project Camellia, is expected to require an initial investment of at least $20 billion. OpenAI has contracted with Georgia Power for 3.2 gigawatts of electricity, which is scheduled to become available in stages between 2028 and 2032.

Unlike many of OpenAI’s previous infrastructure agreements, the company is designing and developing the Georgia project directly. This could give it greater control over construction schedules, energy supply and the systems used to train and run its models.

OpenAI has said the facility will use a closed-loop cooling system intended to limit water consumption. The company has also promised that local customers will not pay higher utility bills because of the project.

Infrastructure team expands

OpenAI has hired Brent Mayo, a former xAI executive involved in Elon Musk’s data-center expansion, to help lead its infrastructure construction efforts.

Mayo is expected to oversee data-center delivery under Uday Ruddarraju, OpenAI’s chief technology officer for computing capacity. The appointments show that OpenAI is building an internal team capable of managing large infrastructure projects rather than relying entirely on cloud providers.

Multiple cloud partnerships

OpenAI continues to secure computing capacity from several major technology companies.

Its cloud partners include Microsoft, Oracle, Amazon Web Services and CoreWeave, while its hardware relationships involve Nvidia, AMD and Broadcom.

Using several suppliers reduces dependence on one cloud platform or chipmaker, but it also creates large long-term financial commitments.

OpenAI reportedly has a five-year cloud agreement with Oracle valued at approximately $300 billion. Its original AWS deal was worth $38 billion over seven years before the companies discussed a much larger expansion.

A costly growth strategy

OpenAI generated about $13 billion in revenue during 2025 and expects its income to rise sharply as ChatGPT subscriptions, enterprise products, developer services and advertising expand.

However, computing costs are also increasing quickly. Inference expenses rose significantly as more people and businesses used OpenAI models, placing pressure on the company’s margins.

The $750 billion plan demonstrates the scale of OpenAI’s ambitions, but it also highlights the financial risk of competing at the frontier of artificial intelligence.

The company must convert its growing infrastructure capacity into sustainable revenue while managing electricity requirements, construction delays and dependence on external suppliers.

OpenAI’s future position may therefore depend not only on the quality of its models, but also on whether it can build and finance one of the world’s largest AI computing networks.

AI News