OpenAI's $34B 2025 burn lands on IPO investors' desks

breakthroughFinanceFounder

Tuesday, June 16, 2026

Conviction

High

Time horizon

This quarter

Risk

AI cost base is structurally higher than IPO marketing implies

The News

The Financial Times reported Monday that OpenAI spent $34 billion in 2025 on artificial intelligence development and marketing, as expenses continued to surge before a planned IPO this year; the ChatGPT maker spent about $19 billion on research and development and nearly $6 billion on sales and marketing, as well as other costs . Reuters confirmed the figures and noted earlier reports showed the company booking about $13 billion in revenue in 2025; the company clocked a net loss of about $39 billion in 2025 . On June 8, 2026, OpenAI confidentially filed an S-1 registration statement with the SEC; the company is targeting a valuation of up to $1 trillion .

The Read

$34B of spend against $13B of revenue is not a "scaling losses" story — it's a structural margin question that the IPO prospectus will have to answer head-on. The interesting number isn't the $34B; it's the $19B on R&D, which is roughly 1.5x revenue. Most pre-IPO software companies show R&D at 15–25% of revenue; OpenAI is at 146%. The market has been told to think of this as Amazon-in-1999, but the analog breaks down where it matters: Amazon's infrastructure investment compounded into a moat (AWS) that throws off cash. OpenAI's infrastructure investment is rented from Microsoft, Oracle, and Nvidia — three counterparties whose stock prices are already pricing in the next decade of that spend. A trillion-dollar IPO requires investors to underwrite that the unit economics flip before the compute commitments do. Inference costs have reportedly quadrupled in 2025; in response, OpenAI is considering significant price reductions for its API offerings to stay competitive against rivals like Anthropic — which is the opposite of what a margin-recovery story looks like.

Do This Week

Treasury and FP&A teams with material OpenAI exposure (direct API spend over $1M annual, or equity exposure via Microsoft/Nvidia/Oracle) should run a counterparty-stability scenario. Model the implied gross margin at current API pricing against the $34B cost base, then ask: at what revenue ramp does OpenAI stop needing another $50B private round between now and an IPO close? If that number is above your forecast, you have a vendor-stability question — not a vendor-pricing one.

For Finance