OpenAI's Leaked 2025 Numbers and Your Enterprise Contract
OpenAI’s audited 2025 financial statements — obtained by independent journalist Ed Zitron and independently reviewed by the Financial Times — show the company tripled revenue to $13.07 billion while its operating loss reached $20.92 billion. Research and development expenses alone outpaced all revenue. For enterprise operators in contract negotiations or approaching renewal, these are the first concrete numbers showing why OpenAI’s IPO is a profitability necessity — and what that means for token prices going forward.
Key takeaways
- OpenAI revenue grew from $3.7B (2024) to $13.07B (2025) — roughly 3.5× growth; operating loss reached $20.92B.
- R&D expenses ($19.18B) exceeded all revenue; $10.59B of that went directly to Microsoft.
- Operating loss as a share of revenue improved: 237% in 2024 → 160% in 2025 — but the absolute gap remains massive.
What Do These Numbers Mean for Enterprise Pricing?
Ars Technica’s analysis of the leaked documents confirms two directional signals operators need to internalize:
The cost floor is real and growing. OpenAI’s cost of revenue — the direct cost of running inference — grew from $2.65B in 2024 to $7.5B in 2025, nearly tripling as usage scaled. Sales and marketing grew from $1.11B to $5.73B over the same period. These costs do not disappear at IPO; they are the operating reality the company’s public valuation has to justify.
The path to profitability runs through pricing. Fortune’s review of the documents confirms the ratio is improving, but the absolute gap is still $21B operating losses against $13B revenue. Getting to break-even by 2030 — OpenAI’s stated investor target — requires either significant volume growth, price increases, or both. Operators should not assume current API pricing is permanently stable.
The $10.59B paid to Microsoft in R&D costs means OpenAI’s model training is deeply underwritten by Azure infrastructure. Operators accessing OpenAI through AWS Bedrock are still paying OpenAI prices that need to cover those Microsoft payments.
What Should Enterprise Teams Do Before the Public S-1 Drops?
When the public S-1 prospectus drops, pricing strategy, revenue targets, and cost disclosures become public commitments — and the negotiating window narrows.
What to do now:
- Treat the leaked numbers as a credible benchmark. Use the $13B revenue / $21B operating loss reality as the context for any pricing-protection or cost-escalation conversation with your OpenAI account team.
- Ask about enterprise price commitments before the S-1 goes public. Renewal windows that close before the prospectus are still in the private-company pricing environment — which is more negotiable than post-IPO quarterly guidance pressure.
- Ask whether current token pricing is contractually fixed or subject to change. June reporting on OpenAI price-cut deliberations showed OpenAI watching Anthropic’s moves. The financials show the pressure in the other direction: cuts are hard to sustain.
- Flag the Microsoft R&D entanglement to your procurement team. Operators building multi-cloud AI strategies should understand that OpenAI’s cost basis is structurally tied to Azure even when accessed through other channels.
For operators, the near-term move is not to assume OpenAI will collapse under these losses — the revenue trajectory is real. It is to use the current window to lock in favorable pricing terms before public-market quarterly guidance makes contract flexibility harder. Watch for the public S-1 release date.
FAQ
Are these financial documents official OpenAI disclosures? No. These are leaked audited statements obtained by independent journalist Ed Zitron and independently reviewed by the Financial Times. OpenAI has not confirmed or denied them. Audited financials carry more weight than draft estimates, but operators should treat specific figures as directionally reliable rather than guaranteed S-1 disclosures.
Does a $21B operating loss mean OpenAI is at risk of failing? Not directly. The ratio is improving (237% → 160%) and OpenAI continues to raise capital. The concern for enterprise operators is not solvency — it is that the gap between frontier AI costs and enterprise prices must close before OpenAI reaches public-market profitability targets, which creates pricing pressure over the contract horizon.