THE MOST capital-intensive startup in history has a chief financial officer who isn't sure the math works. Sam Altman has committed OpenAI to spend more than $600 billion renting cloud servers over the next five years, and has privately told associates he wants the company public by the fourth quarter of this year. Sarah Friar, the former Goldman Sachs analyst hired to make all of this financially credible, has told colleagues she doesn't believe OpenAI will be ready in time — and isn't convinced the spending is justified.
The tension surfaced in reporting by The Information this week, revealing that Friar has been excluded from some conversations about financial plans, including a discussion on server spending with one of OpenAI's top investors. Her absence was, by one attendee's account, noticeable and awkward. In an unusual organizational move for a company of OpenAI's scale, Friar stopped reporting directly to Altman last August and instead began reporting to Fidji Simo, head of applications — who herself went on medical leave last week. The CFO of an $852 billion company now reports to a position that is temporarily vacant.
None of this has slowed the financial machinery. OpenAI closed a $122 billion funding round last week at an $852 billion valuation, the largest private raise in history. Amazon contributed $50 billion, with Nvidia and SoftBank putting in $30 billion apiece. The company says it now generates $2 billion in revenue per month and has 900 million weekly active users. By any conventional metric, the business is thriving.
Burn notice
But the conventional metrics are doing heavy lifting. OpenAI projects burning more than $200 billion before it starts generating positive cash flow, a figure that privately doubled from earlier estimates when investors were briefed in February. Gross margins came in lower than projected last year because the company had to purchase expensive compute at the last minute to meet surging demand — a problem that sounds like success until you notice the margin compression it creates. Revenue is growing at roughly 3x year-over-year, but the spending commitments are growing faster still, locked in years ahead through contracts that bear little resemblance to traditional cloud deals.
This is the structural concern Friar has raised. OpenAI's cloud arrangements require it to help finance data center construction — pre-ordering capacity for 2028, 2029, and 2030, and in at least one case with Oracle, sharing in cost overruns if projects exceed budget. These are not the flexible, pay-as-you-go cloud contracts that enterprise customers are accustomed to. They are forward bets on demand curves that do not yet exist, secured with capital from the same companies supplying the infrastructure. Amazon's $50 billion investment, for instance, includes $35 billion that is contingent on OpenAI going public or achieving artificial general intelligence — making the IPO not a liquidity event but a structural prerequisite for the funding to fully materialize.
Friar's skepticism finds an unlikely echo in Dario Amodei, the Anthropic CEO who left OpenAI in 2021 and has since become Altman's chief rival. Amodei said publicly in February that miscalculating the pace of revenue growth by even a year or two could prove catastrophic for companies making these kinds of data center commitments. His phrasing was pointed: he suggested some competitors had not, as he put it, written down the spreadsheet — had not fully internalized the risk. Anthropic itself plans to spend $19 billion on training and inference in 2026, roughly matching its own revenue, but Amodei has been vocal about sizing commitments to demonstrated demand rather than projected demand.
The parallel to previous Silicon Valley clashes is instructive. Airbnb experienced a similar rift between visionary CEO Brian Chesky and risk-conscious CFO Laurence Tosi, who departed in 2018 over disagreements about strategy and pace. The departure triggered an executive exodus and delayed the IPO by two years. Chesky ultimately succeeded — Airbnb went public in 2020 and is now worth over $80 billion — but the delay was costly and the analogy cuts both ways. The founder's instinct was eventually vindicated, but only after the cautious finance executive's concerns proved at least partially correct about timing.
Both OpenAI and Anthropic are now racing toward Q4 2026 listings, with Altman privately expressing a desire to reach public markets before his former colleagues. Anthropic has engaged Goldman Sachs and JPMorgan, with bankers expecting a raise exceeding $60 billion. If both companies go public within weeks of each other, institutional investors will face an extraordinary allocation decision between two unprofitable behemoths burning cash at historic rates while growing revenue at historic rates. The question is whether the growth or the burn defines the story — and whether the CFO who warned it might be too soon, or the CEO who insisted it couldn't be soon enough, better understood the moment.
Friar, for her part, posted photos of the finance team on LinkedIn after the $122 billion round closed, congratulating Altman for his foresight. In the language of corporate communications, it read like alignment. In the language of organizational dynamics, it read like something else entirely. ■
For more, join 75,000 subscribers getting tech's favorite brief
