OpenAI Held Back by Computing Power Shortage

OpenAI Faces Tough Choices

Sarah Friar, CFO of OpenAI, recently revealed that the company is forced to forgo certain opportunities due to a lack of computing power. This situation compels OpenAI to make "difficult choices" as the demand for its artificial intelligence solutions exceeds the available computing capacity.

OpenAI has had to scale back its projects, including Sora, to redirect its resources toward its core AI products. Friar explained that the company has turned down certain opportunities this year due to insufficient computing power to support them. In an interview with Cathie Wood, CEO of ARK Invest, she stated, "We are making very difficult choices right now and things that we are not pursuing because we do not have enough computing power."

Growing Pressure Until 2026

The computing power issue is particularly concerning for OpenAI in 2026, as the demand for AI continues to grow globally. Friar mentioned that she spends a lot of time looking for last-minute computing solutions for that year.

Greg Brockman, president of OpenAI, also addressed this pressure during an interview on the "Big Technology Podcast." He emphasized that the company is struggling to meet demand, illustrating a growing constraint in the AI industry: even the most advanced companies are limited by access to the computing power necessary to train and run models. "If you don't have computing power, you don't have revenue. That's one thing I know for sure," asserted Friar.

Strategies and Trade-offs

The computing shortage forces OpenAI to make strategic compromises. Brockman indicated that the company is focusing on a limited number of essential use cases, such as a personal AI assistant and tools capable of solving complex tasks, as it "cannot handle them all" with current resources.

This dynamic is already influencing product decisions. OpenAI has paused certain initiatives, including the discontinuation of its video app Sora, to concentrate its resources on revenue-generating AI products. The company, which serves around 900 million consumers and over 1 million businesses, recently raised $122 billion to secure future computing capacity.

"We cannot build computing power quickly enough to meet demand," said Brockman, describing the "very painful decisions" regarding product launches and resource allocation. OpenAI is making "multi-year commitments" to ensure future capacity, according to Friar.

A Challenge Shared by the Industry

Other AI companies are facing similar constraints. Anthropic, for example, recently tightened usage limits on its Claude model during peak hours, demonstrating that even leading model manufacturers are struggling to keep up with rising demand.

For now, a fundamental constraint remains: even in the age of AI, it is impossible to scale without the necessary hardware to support it.