Anthropic on track for >$1B quarterly operating profit in Q3, says SemiAnalysis
Anthropic is on course to become the first frontier AI lab to turn a substantial operating profit, according to a July 2026 SemiAnalysis report modelling the company's finances ahead of its confidential IPO filing (1 June 2026).
- Q1 2026 — roughly break-even, not yet profitable: revenue of ~$4.8B, but the company spent about 71 cents on compute for every dollar of revenue, leaving it around break-even before research, payroll and other costs (per the Wall Street Journal).
- Q2 2026 — first-ever operating profit: revenue jumped ~130% to ~$10.9B and Anthropic projected its first operating profit of ~$559M, as the compute-to-revenue ratio fell to ~56 cents. The figure includes model-training costs but excludes stock-based compensation. SemiAnalysis puts the Q2 EBIT margin at ~36%.
- Q3 2026 — profit set to exceed $1B: SemiAnalysis projects over $1B in GAAP EBIT (~6% margin), with annualised recurring revenue climbing from ~$9B at end-2025 to over $60B.
Why the margins are turning: SemiAnalysis attributes the swing to Anthropic's API-first, usage-based model (75-85% of ARR) and sharply improved inference efficiency, lifting blended gross margin into the mid-60% range and API gross margin above 80% — versus a negative ~94% blended gross margin in 2024. On the most recent All-In Podcast, investor Gavin Baker argued Anthropic is worth ~$3T today, citing reports that it runs ~85% gross margins on inference and forecasting revenue "well over $100B" for 2026, potentially $200-300B by 2028.
Why it matters: the numbers sit at the centre of the debate over subsidies and the economic viability of the AI sector. Anthropic has told investors it does not necessarily expect to sustain profitability quarter-on-quarter given planned infrastructure spending, and the figures are model-based estimates rather than audited results — but if they hold, they would be the clearest evidence yet that investing in AI compute can be profitable at scale.
Read more: SemiAnalysis · Yahoo Finance / FT · CNBC · All-In Podcast (Gavin Baker)
