Sarah Friar, CFO of OpenAI, seems on CNBC’s Squawk Field on August 20, 2025.
CNBC
OpenAI will make 2026 its 12 months of “sensible adoption,” the bogus intelligence startup’s finance chief stated in a weblog Sunday.
“The precedence is closing the hole between what AI now makes doable and the way individuals, firms, and nations are utilizing it daily,” OpenAI CFO Sarah Friar wrote. “The chance is massive and rapid, particularly in well being, science, and enterprise, the place higher intelligence interprets straight into higher outcomes.”
Within the weblog, Friar laid out how OpenAI views its technique for driving monetization of its providers, like ChatGPT, whereas securing the compute essential to energy these merchandise, saying that the AI lab’s income straight tracks with the provision of its technical infrastructure. OpenAI’s compute grew from 0.2 gigawatts in 2023 to about 1.9 GW in 2025, whereas the corporate’s annual income run charge grew equally from $2 billion in 2023 to greater than $20 billion final 12 months, Friar stated.
“That is never-before-seen development at such scale,” she wrote. “And we firmly imagine that extra compute in these intervals would have led to quicker buyer adoption and monetization.”
Friar’s weblog comes as OpenAI and the tech trade’s deal with AI comes beneath scrutiny for the huge investments crucial to construct out the info facilities and safe the power and elements required to energy this cutting-edge expertise that has but to ship a lot income returns for buyers.
Amongst these key offers is an settlement between OpenAI and AI chipmaker Nvidia introduced in September. Beneath that settlement, Nvidia stated it might commit $100 billion to assist the AI startup because it builds and deploys at the least 10 GW of Nvidia’s programs.
A gigawatt is a measure of energy, and 10 GW is roughly equal to the annual energy consumption of 8 million U.S. households, based on a CNBC evaluation of information from the Power Info Administration.
However in November, Nvidia informed buyers there was “no assurance” that its settlement with OpenAI would progress past an announcement and to an official contract stage.
“Securing world-class compute requires commitments made years prematurely, and development doesn’t transfer in a superbly clean line,” wrote Friar, including that the system requires self-discipline.
Three years in the past, OpenAI relied on a single compute supplier, however now it really works with a diversified ecosystem, Friar stated.
“We will plan, finance, and deploy capability with confidence in a market the place entry to compute defines who can scale,” she wrote.
OpenAI’s enterprise mannequin ought to scale with its providers, Friar stated.
“As intelligence strikes into scientific analysis, drug discovery, power programs, and monetary modeling, new financial fashions will emerge,” she wrote.
The weblog comes after OpenAI final week stated that it deliberate to start testing advertisements for some ChatGPT customers within the U.S. That transfer comes as OpenAI gears up for a possible public debut this 12 months.
“Monetization ought to really feel native to the expertise,” Friar wrote. “If it doesn’t add worth, it doesn’t belong.”
