Summary
Ed Zitron (Where’s Your Ed At) deconstructs the revenue reporting practices of OpenAI and Anthropic, arguing that “annualized recurring revenue” (ARR) figures routinely reported by tech press and leaked by companies are systematically misleading. Using historical ARR reports, Zitron calculates actual revenue trajectories and argues both companies are far from their projected figures. His bottom line: OpenAI made ~$3.6B in 2024 and ~$5.3B in 2025 through July; Anthropic made ~$1.5B in 2025 through July.
Key Points
- ARR (annualized recurring revenue) = one month’s revenue × 12; not a GAAP metric, not used by public companies
- Companies leak ARR figures specifically because they sound larger and “annual” — misleading reporters and investors
- OpenAI in 2024: $3.6B actual revenue (vs. projected $12.7B ARR for full year 2025)
- Anthropic in 2025 through July: ~$1.5B actual revenue
- OpenAI growth rate slowing: from 12.7% monthly to 9.54% monthly; hitting $12.7B target requires 14% monthly growth without fail
- Bloomberg reports Anthropic sees “$9 billion annualized by year-end” — Zitron calls this “total bullshit, especially as this number was leaked as Anthropic is fundraising”
- OpenAI reported “$12B ARR” and “$13B ARR” within days of each other — suggests they are using non-standard calculation (current subscribers × 12, not last month × 12)
- Both companies burn more than $5B each per year; revenue vs. burn gap is enormous
- Valuations: OpenAI $300B, Anthropic $170B if a current round closes
Newsletter Angles
- The financial engineering behind AI “success”: ARR is a legitimate metric systematically being used to create false impressions. This is the accounting equivalent of the Mechanical Turk Pattern — the illusion of performance
- The gap between burn rate and revenue is the story the tech press isn’t covering: both companies are spending $5B+ per year and bringing in much less. They are not self-sustaining businesses
- This has political implications: the Anthropic-White House conflict, the Stargate investment, and AI regulation debates all assume these companies are thriving. The actual numbers tell a different story about who needs whom
Entities Mentioned
- OpenAI — primary subject; ~$3.6B 2024 revenue, $300B valuation, burning $5B+/year
- Anthropic — secondary subject; ~$1.5B 2025 revenue through July, $170B valuation target, burning $5B+/year
Concepts Mentioned
- Tech-State Conflict — the political positioning of both companies exists in context of massive financial fragility
- Mechanical Turk Pattern — the ARR deception is an information asymmetry play, analogous to presenting AI as autonomous when it isn’t
Quotes
“ARR is only really used by startups (and other non-public companies). It’s not considered a GAAP-standard accounting practice, and public companies generally don’t use it because they have to report actual figures.”
“Bloomberg is now reporting that Anthropic sees its revenue rate ‘maybe [going] to $9 billion annualized by year-end,’ which, to use a technical term, is total bullshit, especially as this number was leaked as Anthropic is fundraising.”
Notes
Zitron is a tech journalist with a critical perspective on AI industry narratives. His methodology — reconstructing actual revenue from historical ARR reports — is imperfect by his own admission (lacking some months of data). The analysis is directionally important even if some specific figures are estimates. Published August 2025; Anthropic’s ARR figures have since risen significantly per the Sherwood News report from October 2025.
⚠️ Contradiction: The Sherwood News report (October 2025) cites Anthropic at $7B ARR heading toward $9B by year-end — consistent with the “total bullshit” number Zitron dismissed in August. Actual ARR acceleration may have outpaced Zitron’s skepticism, though his critique of the metric’s misleading use remains valid.