OpenAI’s $40 Billion Run Rate Is Real — and It Still Doesn’t Answer the Only Question That Matters for Its IPO

Bloomberg reports OpenAI is on track for an annualized revenue run rate above $40 billion — a genuine doubling in roughly eight months, and a number that says everything about demand and nothing about the cost of meeting it.

What Happened

According to Bloomberg, OpenAI is on track for an annualized revenue run rate exceeding $40 billion — approximately double the run rate of more than $20 billion that CFO Sarah Friar said the company closed 2025 with. Three honest caveats belong in the first paragraph, not the last: a run rate annualizes a recent month’s pace, not a closed fiscal year; the figures are reported by Bloomberg and partly company-sourced, not audited results; and a number surfaced just ahead of a planned public offering carries an obvious incentive to be surfaced in its most favorable form.

With that framing in place, the growth itself is not in dispute. President Greg Brockman told staff internally that the run rate grew more than twenty percent month-over-month in July. The acceleration is broad across product lines: AI coding software, consumer subscriptions, and a nascent advertising business are all contributing. This is not the run-rate optics of a startup annualizing one anomalous month — doubling a more-than-$20 billion base in roughly eight months at more than 20 percent monthly growth is a rate of expansion that reflects genuine, paid, recurring demand at scale.

The timing is deliberate. Bloomberg’s report comes as OpenAI advances its IPO plans, and the run rate is precisely the kind of momentum metric that shapes the pre-offering narrative. Public markets will eventually require a different set of disclosures — gross margin after compute costs, cash burn, and the size of the multi-year infrastructure commitments sitting beneath the revenue line. The $40 billion figure answers none of those questions, which is partly why it is the figure being surfaced now.

The Structural Read

There are three lenses worth applying here, and none of them are the revenue number itself.

Revenue vs. the cost to produce it — the Fifth Bottleneck is money. The meaningful metric for a frontier AI lab is not top-line revenue but the gap between that revenue and the compute expenditure required to generate it. OpenAI’s $40 billion run rate sits against a buildout obligation measured in the hundreds of billions: the training infrastructure, the inference capacity, the long-term data center commitments that make the revenue possible in the first place. A frontier lab is, structurally, a tenant that owes an enormous and largely fixed sum for the infrastructure it runs on — and the question that decides its fate as a public company is whether the top line can grow into those obligations before the bills compound. This is the Fifth Bottleneck, money, made concrete: the AI buildout has a capital structure problem that revenue growth alone cannot resolve, because the costs scale with the capability that generates the revenue. The gap — not the gross number — is the metric. For deeper context on why capital structure is the constraint that matters most at this layer of the AI stack, see The Fifth AI Bottleneck and The First AI Financial Meltdown.

The IPO narrative engine. A run rate is the most flattering true thing a pre-IPO company can say about its revenue trajectory. It is true — the annualization of a recent month is real math applied to real numbers — and it presents momentum in its most compressed, striking form. This is not cynicism; it is how pre-offering narratives work, and investors understand the convention. What matters is recognizing that the metric has been selected for its strength: it captures the acceleration while sidestepping the margin, the burn, and the size of the commitments sitting behind the top line. Public markets will eventually demand all of those, as they did with every previous generation of high-growth, deeply unprofitable technology companies. The pattern is well-established enough that comparing the OpenAI IPO calculus to the dynamics facing Anthropic is instructive — see Anthropic’s IPO and Valuation: What Public Markets Will Actually Test.

Acceleration is the signal; the level is the headline. The $40 billion figure will dominate coverage because round large numbers dominate coverage. The more structurally meaningful data point is the more than twenty percent month-over-month growth Greg Brockman cited for July, combined with the product mix shift toward AI coding and subscriptions. A growth rate above twenty percent monthly at this revenue base would, if sustained, compound into a figure that genuinely changes the cost-gap calculus — because faster revenue growth against largely fixed infrastructure obligations improves the unit economics structurally over time. The level is a snapshot. The acceleration, and the product composition driving it, is the variable worth tracking. This is analogous to the dynamic visible at the app layer: run rates can look spectacular and still be insufficient if the cost curve beneath them is steeper than the revenue curve above.

Three Implications

The Bottom Line

91,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

Sources: bloomberg.com · pymnts.com · qz.com · sherwood.news · semafor.com