OpenAI's Revenue Run-Rate Gap: ~$50B vs the ~$70B Circulated — and What It Did to the Capex Chain
On October 8, 2026, the Financial Times reported (verified by Reuters) that OpenAI told investors its annualized revenue was "approaching $50 billion at the end of September, roughly $20 billion below the nearly $70 billion figure that had been widely reported." The gap is mostly methodology, not a demand break: investors had adjusted OpenAI's figures upward to make them comparable with Anthropic's, which counts the full value of sales through cloud partners (who keep ~16 cents of every dollar generated through those relationships); OpenAI excludes some partner revenue. CFO Sarah Friar's remark that the run-rate climbed ~70% between end-June and end-September, applied to the adjusted base, is what produced the ~$70B figure. (TheStreet, Oct 11, 2026)
Markets reacted the same day: Nvidia -2.9%, Oracle -5.5%, Micron -4.8%, with AMD, Broadcom and Intel also falling sharply; the Nasdaq closed down ~1.3% (TheStreet, citing MarketWatch). AMD fell 6.4% over three sessions after a record close. The sensitivity is structural: Reuters also reported that Broadcom and Oracle are seeking $50 billion in OpenAI financing — the same OpenAI whose revenue base was just marked down ~29% in circulated figures.
The counterweight, two days later: Bloomberg (Oct 9) reported OpenAI expects at least $70B in annualized revenue by year-end. The underlying growth is real — OpenAI entered 2026 at ~$20B annualized (vs ~$6B in 2024) and reached ~$50B by September; Q2 quarterly revenue was ~$6.7B vs Anthropic's ~$11.5B, with Anthropic's annualized passing $65B in July. At a ~$1.4T valuation, OpenAI is ~28x the newly reported run-rate vs ~20x the old figure.
What it means
OpenAI sits at the end of much of the capex chain this topic tracks — Stargate builds, the Broadcom/Oracle financing, custom-ASIC volumes — and its revenue base is now the single most-watched number in the buildout. The Oct 8 selloff showed the market will reprice the entire chain on a disclosure-methodology change alone, before any actual demand data changed. It also strengthens the case for the standardized financials an IPO would bring — the 2027 window this note tracks (Altman: an IPO this year would be "ill-timed"). Until then, private run-rate claims will keep moving public markets, and the gap between the $50B (end-Sept) and $70B (year-end target) figures is itself a testable demand signal for Q4.