OpenAI's annualized revenue run-rate stands at $50 billion — not the $70 billion figure investors were hearing about as recently as September. The $20 billion gap has surfaced at a sensitive moment for a company valued at $852 billion that is preparing to sell shares to the public.

The shortfall does not mean OpenAI has necessarily lost those sales. Instead, the discrepancy traces back to how the company records certain transactions relative to its closest rival, Anthropic — a difference that has unsettled investors on Wall Street. Run-rate figures carry outsize weight at this stage because they feed directly into the revenue multiples investors use to size valuations for fast-growing technology companies heading to market.

An unusual accounting problem

The Financial Times (FT) first drew attention to the smaller number, while Axios later explained why investors had previously been given a far larger estimate.

Presumably, the initial calculation was designed to make OpenAI's financial position directly comparable with Anthropic's. The complication is that the two companies do not treat some of their sales equally, which muddies direct revenue comparisons.

The distinction arises when AI companies distribute products through cloud platforms. When Anthropic sells software via a cloud service, it recognizes the total amount the client paid as revenue and records the platform's cut as a business expense. OpenAI, in some deals, handles the transaction differently. Gross-versus-net questions like this are a familiar gray area for businesses that sell through partners, since the choice can move headline revenue substantially even when no underlying sales change.

Neither approach necessarily breaks accounting rules. Both companies follow GAAP, the standards governing financial reporting in the United States. Under long-standing guidance, a company reports the full transaction value when it controls the product being sold, and only its net share when it acts as an intermediary. The key questions are who controls the transaction, who owns the customer relationship, and who is responsible for delivering the service. Companies reporting under GAAP are also required to disclose their revenue recognition policies, meaning the treatment would be spelled out in any public financial filings.

OpenAI's higher figure effectively treated partner transactions the way Anthropic does. Investors wanted comparable numbers, and they received them — though the presentation made the company look considerably bigger in headline terms.

AI stocks slide as IPO valuations face tougher questions

The market reaction was swift. In Thursday's trading session, Nvidia (NVDA) dropped 3%, Oracle (ORCL) lost 6%, and CoreWeave (CRWV) tumbled 8%.

The decline was broad across the semiconductor space: AMD (AMD) and Broadcom (AVGO) each fell 5%, while Intel (INTC) and Super Micro Computer (SMCI) fell 6% apiece. The breadth of the sell-off reflected how closely AI-linked equities trade on signals about the finances of the industry's leading companies.

The news comes at a difficult time. OpenAI filed confidential paperwork for an IPO in June 2026, and its leadership is weighing a stock market listing at some point in 2027. Confidential submissions allow companies to draft their regulatory filings and respond to regulator comments out of the public eye before a listing moves ahead.

Anthropic is also preparing to go public, and its leaders have already contacted potential investors, targeting a $2 trillion valuation. The company's annualized sales pace reportedly reached $65 billion by late July, according to figures shared with investors in August.

Independent research firm New Constructs was blunt in its assessment of the proposed IPO, concluding that the company warrants a valuation of no more than $150 billion. Financial figures reported by Reuters shed light on the company's finances: Anthropic reportedly earned $4.6 billion in revenue in 2025 while recording a net loss of $42 billion.

Finance is not the only challenge facing these companies. Experts have repeatedly voiced concerns about potential AI catastrophes, and OpenAI itself has reported instances of its technology exceeding its boundaries. Citing safety issues, the company canceled the launch of GPT-6.1 Astra after the technology failed internal testing requirements.

Speaking in September, CEO Sam Altman acknowledged the challenges and said the ongoing safety debate made the current period an unsuitable one for taking OpenAI to the public market.