OpenAI's $21 Billion Loss Fuels Doubts About AI Lab Economics

Audited losses challenge venture valuations

OpenAI lost $20.9 billion on $13.07 billion in revenue during 2025, spending roughly $1.60 for every dollar earned, according to audited financial documents obtained by Ed Zitron and verified by the Financial Times. Total costs approached $34 billion, and a one-time charge related to the company's for-profit conversion pushed the reported net loss to $38.5 billion.

The figures surfaced days after OpenAI filed confidentially for an initial public offering, setting up the first major test of frontier AI economics in public markets. The company closed a $122 billion funding round in March 2026 at an $852 billion valuation, pricing that assumes durable value will accrue to the model layer rather than infrastructure or application tiers.

Why it matters

Venture capital has bet hundreds of billions on the thesis that large language model developers will capture and defend margin at scale. The convergence of leaked financials, infrastructure warnings, and operator criticism in a single week suggests that thesis faces material execution risk—and public investors will price that risk within months.

Operators and suppliers add skepticism

Palantir CEO Alex Karp described the token-based business model as "insane" during a televised interview, arguing that enterprises surrender competitive advantages while extracting little lasting value. His comments reflect growing concern that the model layer may become a commoditized, low-margin business subject to price competition.

Oracle, a key compute provider for OpenAI, warned the SEC that its massive Stargate data center project faces financing hurdles and potential customer nonpayment risks. The filing adds supply-chain doubt to demand-side questions about unit economics.

Intellectual property friction is also mounting. Anthropic's recent product launch affected Figma, illustrating how model capabilities can disrupt adjacent software categories and complicate partnership economics.

Even Alphabet is raising $80 billion specifically for AI infrastructure, underscoring the capital intensity required to compete at the frontier. The scale of investment across the sector has prompted analysts to question whether current burn rates are sustainable absent a clear path to positive unit economics.

Public markets will force transparency

OpenAI's confidential IPO filing means audited results, customer concentration, and margin structure will soon face public scrutiny. The offering will test whether growth-stage venture multiples translate to public equity valuations when operating losses exceed $20 billion annually.

The skeptics questioning frontier AI economics have shifted from short sellers to operators, auditors, and the labs' own regulatory disclosures. That broadening base of credible critics arrives as the sector prepares for its first major liquidity event, forcing a reckoning between private valuations and public-market discipline.

Details were first reported by Ed Zitron and confirmed by the Financial Times.