OpenAI's $50 Billion Correction Resets the Anthropic Comparison

OpenAI has told investors its annualized revenue is "approaching $50 billion," some $20 billion below the $70 billion figure reported in late September. TechCrunch on October 8, 2026 attributed the lower number to Financial Times reporting.

The $70 billion figure came from information previously shared with investors by OpenAI, according to the Financial Times. That figure was devised through attempts by OpenAI's own investors to produce a direct comparison with Anthropic's annualised revenues. Axios had previously reported that OpenAI's annualized revenue was approaching $70 billion, a report relayed by Reuters on September 29, 2026.

The comparison problem is definitional. OpenAI and Anthropic calculate annualized revenue differently, with Anthropic counting sales made by its cloud partners while OpenAI does not. For readers who track cloud consumption and API resale, that distinction matters. Gross merchandise flow through hyperscalers is not the same as first-party recognized revenue.

The September reports put OpenAI's annualized recurring revenue as up more than 70% since the start of the third quarter of 2026. Enterprise sales increased more than twofold since July 2026, according to that same Reuters account. A separate Yahoo Finance summary of the Axios reporting on September 29 gave the same growth rate for annualized recurring revenue since the start of the quarter.

The revision lands at the end of a steep revenue ramp. The Information reported that annualized revenue doubled to $3.4 billion since late 2023, with most revenue from subscriptions and API access. It later reported OpenAI reached $12 billion in annualized revenue after roughly doubling revenue in the first seven months of the year, with ChatGPT breaking 700 million weekly active users. By March 4, 2026, The Information reported OpenAI had topped $25 billion in annualized revenue, a 17% increase from $21.4 billion.

Other checkpoints fill out the curve. The Information reported OpenAI generated about $5.7 billion in revenue in the first quarter. It reported OpenAI was on track to top $13 billion in 2025 revenue, up from around $4 billion the prior year. Leaked 2025 financials showed it made about $13 billion but spent significantly more. Bloomberg reported on August 14, 2026 that OpenAI was on track to generate more than $40 billion in annualized revenue.

Anthropic's numbers have moved in parallel. The Information reported in May that Anthropic's annualized revenue recently neared $45 billion. Bloomberg reported on August 18, 2026 that Anthropic's annualized revenue topped $65 billion.

Valuation and financing activity has moved faster than recognized revenue. OpenAI raised $122 billion in a March funding round. It completed an employee share sale at a $500 billion valuation, as reported by Bloomberg on October 2, 2025. On October 8, 2026, Bloomberg reported OpenAI was in talks to raise $30 billion at a $1.4 trillion valuation, with Tiger Global poised for a $5 billion paper profit on its OpenAI investment. Reuters on September 29, 2026 also reported, citing Bloomberg News, that OpenAI was targeting $30 billion in funding at a $1.4 trillion valuation. OpenAI said it would purchase $250 billion worth of cloud services. Its IPO, previously rumored for 2026, has been pushed off until early 2027.

The broader context here is accounting, not just growth. Annualized revenue is a snapshot calculation. It takes a recent month or quarter and extends it across a year. It is sensitive to timing, to consumption bursts, and to what gets included. Channel sales through cloud partners can inflate a top-line comparison if one vendor includes them and the other does not.

In this author's view, worth flagging is how familiar this pattern feels. New platform shifts routinely produce a period where private valuations, committed infrastructure spend, and reported revenue use different yardsticks. I watched my children move from PCs to phones to cloud services without ever asking about run rates or partner resale. Enterprise buyers are now doing the equivalent at scale, buying tokens, seats, and API capacity. The finance layer struggles for a time to describe it consistently. The long arc still points toward wider use. What changes with this correction is not the direction of adoption. It is the precision of the comparison.