OpenAI's Revenue Gap Lands: $50B vs the $70B Signaled — the Measurement-Basis Defense and the $1.4T Ask
The demand-anchor story this note tracks (prior revision: the Oct 8 FT/CNBC $50B run-rate disclosure that knocked 3–8% off the AI complex) got its full reconciliation and the market's verdict this cycle.
The number and the gap. The Guardian (Oct 9): "OpenAI has revealed that it is making about $20bn less in projected revenue than it had recently indicated to investors... The ChatGPT-maker company has told investors that its revenues for this year would reach $50bn (£37bn), a projection based on sales up to the end of September. However, this is significantly less than the $70bn that it had signalled in information provided to investors last month" (Guardian).
The measurement-basis defense. "The discrepancy arose from attempts by OpenAI investors to provide a more direct comparison with how projected revenue is measured by Anthropic, which hit $65bn in forecast revenue by the end of July. Anthropic includes the revenue from sales via cloud partners, such as Amazon's AWS and Google Cloud, while OpenAI does not1." Part of the gap is definitional — but the definitional cleanup itself produced the lower comparable number, and it is the comparable number the market now prices.
The market verdict (Thursday Oct 8): "the tech-led Nasdaq closing down by 1.4%. Chip giant Nvidia fell 2.9%, Oracle was down 5.5% and Micron declined 4.8%."
The ask continues anyway. "OpenAI is in early-stage talks to raise $30bn in a funding round that values the business at about $1.4tn" — roughly 50% above the $852B March valuation — still without a finalized term sheet (Guardian; Traded, Oct 12). Altman has delayed an IPO until next year citing safety concerns; Anthropic's IPO is expected "as soon as next month" — the S-1 remains the hard-data event for the demand base (The Demand Base Gets Marked Down Before the IPO: Investors Temper Anthropic From $2T to ~$1.5T).
Meaning for the capex question. The demand base's headline growth is being marked to a comparable measurement basis — not collapsing — but every forward commitment (the $1.4T ask, the $500B financing platforms, the debt-funded chip orders) now gets underwritten against a $50B-this-year OpenAI rather than $70B. That is precisely the scrutiny Huang spent this week rebutting (Equity Layer: BofA's Capital-Cycle Math — $1T Consensus Capex, Token Prices -50%, and an 8-Day SOX Streak That Only NVDA Survived), and it sharpens the open deceleration watch: the next hard datum is Anthropic's S-1, not blogger ARR charts.
-
An instance of AI demand is a measurement convention before it is a market fact. — The $20B 'gap' is produced by measurement-basis choices — cloud-partner pass-throughs and comparability cleanups, not customer events, set the revenue figure markets underwrite. ↩︎