OpenAI is turning down some opportunities this year because it doesn’t have enough computing power to support them, according to its CFO Sarah Friar.
“We’re making some very tough trades at the moment and things we’re not pursuing because we don’t have enough compute,” Friar told ARK Invest CEO Cathie Wood in an interview released this week.
Friar said the issue is particularly acute in 2026, as demand for AI continues to surge globally.
“Today, I do spend a lot of time trying to find any last-minute compute available here in 2026,” she said.
OpenAI President Greg Brockman echoed that pressure in an interview on the “Big Technology Podcast” released Wednesday, saying the company has struggled to keep up with demand.
Their comments highlight a growing constraint across the AI industry: even the most advanced companies are being limited by access to the computing power needed to train and run models.
“If you do not have it [compute], you do not have revenue. That is one thing I know for sure,” Friar said.
Compute limits force trade-offs
The shortage is forcing OpenAI to make strategic trade-offs.
Brockman said the company is prioritizing a small number of core use cases, including a personal AI assistant and tools that can solve complex tasks, because it “can’t possibly get to all of them” given current compute limits.
That dynamic is already shaping product decisions. OpenAI has pulled back from some initiatives — including discontinuing its video app Sora — as it focuses resources on core, revenue-generating AI products.
The company, which now serves around 900 million consumers and more than 1 million businesses, Friar said, is raising huge sums. It recently completed a $122 billion funding round, in part to secure future compute capacity.
“We cannot build compute fast enough to keep up with demand,” Brockman said, describing what he called “very painful decisions” about what to launch and where to allocate resources.
OpenAI is making “multi-year commitments” to secure future capacity, Friar said.
Other AI companies appear to be facing similar constraints.
Anthropic recently tightened usage caps for its Claude model during peak hours, a sign that even leading model makers are struggling to keep up with surging demand.
For now, a basic constraint remains: even in the age of AI, you can’t scale without the hardware behind it.
