OpenAI tops $40B run rate — coding tools are paying the bill

The number moved fast enough to be worth reading carefully. OpenAI's annualized revenue run rate has passed $40 billion, roughly double where it sat at the end of 2025, Bloomberg reported on August 13. The growth driver named first is AI coding software. If you buy tokens from anyone, that detail tells you more about your next renewal than the headline figure does.

What actually happened

Per Bloomberg's reporting, the run rate accelerated in recent months on three things: AI coding tools, subscription sales, and a young advertising business. Co-founder and president Greg Brockman said the annual revenue run rate grew more than 20% month-over-month in July. CFO Sarah Friar has previously put the end-of-2025 figure at more than $20 billion annualized.

For context on the competitive picture: Anthropic is on a similar trajectory, with CNBC reporting a projected $10.9 billion in Q2 revenue and a possible first operating profit — also driven heavily by agentic coding. Both companies are moving toward public listings. We're not going to guess at timing or terms; the filings will say.

One thing we'll flag rather than assert: a run rate is an annualized snapshot of a recent period, not a year of collected revenue. It's a real signal about demand. It is not a bank statement.

Why it matters for your business

Two readings, and they point in opposite directions.

The optimistic one: revenue at this scale means the AI vendors you depend on are no longer pure cash-burn bets. That's genuinely good if you've built a business process on top of an API. Vendors with revenue stay alive.

The one we'd actually plan around: coding is carrying the growth. When a single category drives the curve, pricing follows the curve. We've already watched per-seat plans convert to usage-based billing across this market — GitHub moved Copilot toward token-metered credits, and multiple vendors have repriced mid-year. A company doubling its run rate ahead of a public listing has every incentive to keep optimizing revenue per customer, and you are the customer.

So do the boring things now, while you have leverage. Know your actual monthly token spend per workflow, not per company — if you can't name which automation costs what, you can't negotiate or cut it. Keep contracts short. Build against an abstraction (an API gateway, a router, a proxy you control) so swapping a model provider is a config change instead of a rewrite. And run the arithmetic on what a 30% price increase does to each automation you've deployed. The ones that still clear the bar are your real portfolio. The ones that don't were never economics — they were a subsidy, and subsidies end when the IPO paperwork starts.

Sources: Bloomberg, CNBC.