The $20 Billion Gap: OpenAI's Real Annualized Revenue Is ~$50B, Not the $70B Markets Priced In

On October 8, 2026, the Financial Times dropped a number that quietly recalibrates the entire AI investment thesis: OpenAI has recently told investors that its annualized revenue was approaching $50 billion at the end of September — far less than the $70 billion figure that was widely reported and, more importantly, widely priced in.

The gap is not a rounding error. It is a $20 billion discrepancy — roughly 29% of the headline number — on the single metric that private markets, debt underwriters, and IPO watchers have been using to justify the most aggressive capital raise in corporate history.

Nine days that moved $20 billion

The timeline is what makes this story remarkable. On September 29, Axios reported — and Reuters confirmed — that OpenAI’s annualized recurring revenue was “approaching $70 billion,” up more than 70% since the start of the third quarter and driven by a more-than-twofold increase in enterprise sales since July. The figure ricocheted through every corner of the market. Oracle shares rose 4% on the news, because OpenAI’s growth is a direct read on Oracle’s cloud revenue. Sacra, the private-market research firm, baked $70B into its estimates, noting it implied OpenAI had added $50 billion in annualized revenue in just nine months.

Nine days later, the FT’s reporting says the number OpenAI itself is putting in front of investors is close to $50 billion. As the FT put it, the key figure was “approaching $50bn in September, far less than the $70bn that was widely reported.”

Neither number is necessarily a lie. They may not even describe different underlying businesses. But the difference between them — and the speed with which the market adopted the higher one — says a great deal about how AI economics are being measured, marketed, and monetized right now.

Why “annualized revenue” is such a slippery metric

“Annualized revenue” or “run rate” is, by construction, an extrapolation: take the most recent period’s revenue and multiply it out to a year. That makes it exquisitely sensitive to whatever window you choose. If September included a burst of one-off enterprise deals, API usage spikes, or the initial bookings from a major distribution partnership, a single strong week or month can be annualized into billions of dollars of “recurring” revenue that has not actually recurred yet.

This is not a hypothetical concern for OpenAI. Critics have flagged for over a year that the company’s run-rate announcements tend to coincide with its strongest recent windows. The distinction between “annualized revenue approaching $70B” and “ARR near $70B driven by enterprise sales more than doubling since July” matters enormously: the former is a snapshot, the latter is a claim about a durable growth curve. If the September figure was lifted by a cluster of large enterprise contracts signed in a short window, the honest annualized picture may genuinely be the ~$50B that OpenAI is now reportedly showing investors behind closed doors.

There is also a simple mechanical possibility: the $70B figure may have come from a different measurement basis — perhaps a forward-looking projection, a bookings number, or a single-source leak — that was never the same metric as what OpenAI’s finance team tracks. Axios attributed its report to sources familiar with a briefing; the FT attributes the $50B figure to what OpenAI itself told investors. When a leak and a disclosure disagree by $20 billion, the disclosure should win.

What the real numbers look like

Zoom out, and even the lower figure tells a staggering growth story. OpenAI exited 2025 at roughly $20 billion in annualized revenue. In August 2026, Bloomberg reported its run rate had topped $40 billion. A September close near $50 billion implies the company roughly 2.5x’d its revenue in nine months — a pace no software company in history has sustained at this scale. By comparison, Anthropic’s annualized revenue surged from $9 billion at the end of 2025 to $30 billion by end of March 2026, and it is now valued at $965 billion.

But the growth rate is only half the equation. The FT reported last month that OpenAI expects $278–280 billion in negative free cash flow between 2026 and 2030, with expenses projected to massively outrun revenues even as revenue grows tenfold — from roughly $36 billion this year to a projected $350 billion by 2030. Against that plan, the difference between $50 billion and $70 billion of current revenue is the difference between a company that is covering an ever-smaller fraction of an enormous burn, and one that has a credible bridge to the projections underpinning its fundraising.

Why $20 billion matters to everyone else

The reason this discrepancy lands so hard is that OpenAI’s growth number has become infrastructure for the entire financial system around AI:

• The IPO narrative. OpenAI closed its March 2026 round at an $852 billion post-money valuation, is reportedly in talks for a new round above $1.2 trillion, and its CFO has said the company “will be a public company in 2027.” Every dollar of ARR disclosed on that roadshow will be benchmarked against the $70B that markets already believe. Coming in $20 billion light reframes the valuation conversation before it starts.
• The debt buildout. Blackstone is syndicating a $60 billion Broadcom-Anthropic chip debt package, and Broadcom has lined up $50 billion in financing so OpenAI can buy its custom accelerators. Credit underwriters price these deals off growth assumptions. A 29% haircut to the anchor metric flows directly into covenant terms, pricing, and appetite for the next tranche.
• Supplier stocks. Oracle, Nvidia, TSMC, and the whole AI supply chain trade on OpenAI’s ability to convert revenue into compute purchases. TSMC just posted a record NT$1.494 trillion quarter precisely because AI accelerator orders keep beating estimates. If the growth engine’s fuel gauge reads lower than advertised, every downstream read gets recut.

The larger lesson

There is a well-worn pattern in tech financing cycles: the metrics that circulate fastest in private markets are the ones nobody audits. “Annualized revenue” is not GAAP revenue. It is not audited, it is not filed anywhere, and — as of this week — the same company appears to have two versions of it in circulation, separated by $20 billion, within a nine-day span.

None of this means OpenAI’s business is faltering. Growing from $20 billion to roughly $50 billion in annualized revenue inside a year is one of the fastest ramps in the history of capitalism, and even the lower number would justify a sense of awe. But it does mean the market has been underwriting trillions of dollars of valuation and hundreds of billions of debt financing against an unaudited, extrapolated, and — it now appears — optimistic figure.

Until OpenAI files an S-1 and publishes audited financials, every ARR number you read about the company should carry an implicit error bar. As of this week, we know that error bar is at least $20 billion wide.

Watch for three things next: whether Axios or Reuters substantively respond to the FT’s reporting; whether OpenAI’s next financing round documents (or its 2027 IPO preparation) disclose audited revenue; and whether the credit markets syndicating AI infrastructure debt begin demanding more conservative growth assumptions than the ones that circulated in September.

---

Figures in this article are drawn from Financial Times, Reuters, Axios, and Bloomberg reporting as of October 8, 2026. OpenAI has not publicly confirmed its current annualized revenue.