OpenAI finance chief sets a test for whether AI spending pays off
Sarah Friar says companies should judge AI by completed, reliable work per dollar, as CFOs take a larger role in AI strategy.
By Daniel Okafor · Business Editor
3 min read
OpenAI CFO Sarah Friar has laid out a scorecard for judging whether corporate AI spending is producing economic returns. Her argument, published in an OpenAI blog post, centers on measuring the work AI completes, rather than relying on older software yardsticks such as seat counts, active users or renewals.
Friar wrote that finance leaders face a basic test: whether the value of AI-generated work is rising faster than the cost of producing it. She said that requires a broader view than narrow measures such as cost per token.
The measure Friar emphasized is “useful intelligence per dollar.” She framed it around four questions: whether AI is doing work that matters, what a successful task costs, whether users can trust the result, and whether each additional dollar creates more value as usage expands.
The four-part scorecard
Friar said companies should begin by defining what counts as successful AI-completed work, including a quality threshold. They should then calculate the full cost of producing that work and divide it by the number of successful tasks to arrive at a cost per completed task.
The next test is reliability. Friar said leaders must determine whether people can depend on AI outputs in real operations. Over time, she wrote, companies should look for high-quality completed work to grow faster than total cost, while quality stays steady or improves.
Compute is central to that equation for OpenAI, according to Friar. She wrote that OpenAI’s role is to improve the economics with each generation of models by making them more capable, quicker, more dependable and cheaper for the work customers need done.
Compute spending remains central to OpenAI’s plans
Fortune reported that OpenAI, as a private company, does not provide formal capital spending guidance. The company’s Stargate initiative, announced in January 2025, set out a plan to invest as much as $500 billion over about four years in large-scale AI infrastructure in the United States.
According to Fortune, the first phase targeted roughly $100 billion, while the broader plan aimed to accelerate toward 10 gigawatts of U.S. capacity by 2029. Fortune reported that OpenAI has already passed a milestone in that buildout as demand for AI continues to rise.
Fortune also reported that OpenAI’s initial public offering could happen as soon as this summer or as late as 2027, citing reports. The company is valued at $852 billion and is nearing the $1 trillion range, according to Fortune.
CFOs take a larger strategy role
The discussion comes as finance chiefs take on more responsibility for strategic decisions, including major long-term AI spending. Fortune reported that CFOs have traditionally led capital allocation and investor communications, but are now more often expected to help set company strategy with CEOs.
McKinsey recently held its 24th annual Global CFO Forum, bringing together about 100 finance chiefs from more than 30 countries. Andy West, a senior partner at McKinsey and global co-leader of its Strategy and Corporate Finance practice, told Fortune he informally asked attendees whether strategy now reports to them.
About two-thirds raised their hands, West told Fortune. He said the share would have been under one-third five years ago.
West also told Fortune that AI has been part of the forum’s discussions for several years. Last year, he said, finance leaders were still testing AI; this year, the discussion shifted toward companywide transformation.
This story draws on original reporting from Fortune.