OpenAI Wants to IPO. Its Own CFO Says It Shouldn’t.

OpenAI is preparing to file its IPO registration in the coming weeks, targeting a September listing that could value the business at more than $1 trillion — and its own CFO is not convinced the company is ready.

Working with Goldman Sachs and Morgan Stanley on a draft prospectus, OpenAI is aiming to go public as early as September, according to Reuters. The company had raised $122 billion earlier this year, likely Silicon Valley's largest-ever funding round. The S-1 filing, when it comes, will set a benchmark that shapes every AI lab's fundraising options for the years ahead — and the numbers underneath the headline are what make the timing tight.

Sarah Friar has reportedly been frozen out of key financial meetings and has reportedly told colleagues she does not believe OpenAI can go public in 2026, according to Business Insider. OpenAI disputes the characterization of the meeting freeze and the IPO-readiness concern. The pushback is on record; so are the reports. OpenAI projects it will burn through roughly $25 billion in 2026 — against roughly $24 billion to $25 billion in annual revenue, dollar for dollar — according to the company's own disclosures and investor communications as reported by Reuters. Those figures are projections, not audited results, and the methodology behind them will be clarified when the S-1 filing is public. The gap is not a crisis in private markets, where patient capital has sustained frontier AI labs since 2022, but it will face a different set of questions on a public balance sheet. Public market shareholders tend to ask for something that private investors have not: a path to profits.

Four things will determine whether the IPO happens on schedule.

First, Anthropic has crossed a threshold. In April, for the first time, Anthropic's tools overtook OpenAI's in spending among business customers, per Ramp data tracking roughly 50,000 companies. Anthropic is now generating $30 billion in annualized revenue, against OpenAI's $24 billion to $25 billion — a gap that did not exist eighteen months ago and that closed not because Anthropic grew slowly, but because its coding agent Claude Code set the industry standard. It is worth noting that OpenAI and Anthropic use different revenue accounting standards — OpenAI may count net platform fees where Anthropic counts gross — which means the raw comparison is an open question rather than a settled measurement.

Second, the Musk trial verdict cleared a legal obstacle on May 18th — a jury found for OpenAI, removing a lawsuit that had complicated the IPO timeline. But the verdict did not come clean. Multiple witnesses called Sam Altman a liar under oath. His lawyers called it irrelevant. Public market investors will form their own view.

Third, the trial record surfaced something that will require explicit disclosure in the S-1: Altman has a personal stake in chipmaker Cerebras, which has a major commercial relationship with OpenAI. S-1 regulations treat related-party transactions as material disclosures that require a specific description of the financial relationship, not just the existence of a stake. Investors reviewing the filing will see the Cerebras entanglement alongside the perjury testimony — both are credibility data points, but only one is a legal disclosure requirement.

Fourth, SpaceX files first. Its IPO — targeting a $1.75 trillion valuation — will set the benchmark for how public markets price frontier AI businesses this year. Whether anchor investors have signaled appetite for another AI listing before SpaceX has even priced is an open question — one the S-1 itself will answer.

If OpenAI prices and holds, it will do something no private funding round has done: establish a public-market benchmark for what a frontier AI business is worth at dollar-for-dollar burn. Investors who bought the IPO will then apply that same math to every lab that comes after. Anthropic will face harder questions when it files. So will every later-stage private company that has been valued on the assumption that the burn was temporary. The patient capital era did not end when the last venture check cleared — it ends when the first public shareholder demands to know when it stops.