Anthropic reports $17B in Q3 revenue, commitments of $895B in revised S-1
Anthropic’s revised S‑1, revealing $17B in Q3 revenue alongside a GAAP net loss driven largely by convertible note revaluations and nearly $900B in compute commitments, is prompting sharp scrutiny of its planned ~$2T IPO valuation. Commenters highlight that while 47% quarter‑over‑quarter growth and positive gross margins are extraordinary at this scale, growth is already decelerating and margins are under pressure from cheaper open‑weight and Chinese models. Many see Anthropic as a potential long‑term winner in AI, but question whether its thin moat, capital‑intensive economics, and reliance on continued hypergrowth can justify anything close to its current pricing, especially for retail investors.
Q3 financials & accounting quirks
- Reported Q3 revenue and ARR growth are very high (ARR up ~47% QoQ), but GAAP net loss is enormous due mainly to revaluation of convertible notes and derivative liabilities.
- Several commenters debate whether these “paper losses” should be treated as real economic losses; concern that dismissing them is like ignoring stock‑based compensation.
- Others argue GAAP treatment is weird but designed to prevent tricks; valuation increases naturally make convertibles more valuable and can flip to “gains” if valuation falls.
Growth trajectory & “tokenmaxxing”
- Revenue growth slowed from ~100%+ QoQ to ~47% QoQ. Some see this as still “eye‑popping”; others say the sharp deceleration is a bad signal for a $1–2T IPO.
- Debate over how much Q3 is still inflated by earlier “tokenmaxxing” (unconstrained usage) vs. the start of spend optimization and caps in big companies.
- Disagreement on whether we’re seeing normal S‑curve behavior or early signs that AI demand is below expectations.
Moat, competition & open‑weight models
- Many argue LLMs are becoming commodities; switching costs between providers are low and developers often can’t tell models apart in day‑to‑day work.
- Open‑weight and especially Chinese models are framed as “good enough” for most coding and many tasks, often at a tiny fraction of Anthropic’s prices.
- Counterpoint: Anthropic is trying to build a moat via integrated enterprise tools, workflows, and brand, not just raw models. Unclear if that moat will be durable.
Compute costs, commitments & business model
- There is heavy concern about ~$895B in compute commitments and very high Q3 “compute and infrastructure” cost growth (faster than revenue).
- Some say Q3 margins were hurt by expensive spot GPUs and will improve when long‑term contracts start; others fear Anthropic may be locked into overpaying if demand shifts to cheaper models.
- Several liken the lab business to capital‑intensive industries (airlines, chip fabs) rather than high‑margin software.
IPO valuation & investing debate
- Many see Anthropic as a great growth story but a bad stock at mooted $1.5–2T valuations; some think it’s really a ~$200–300B business pretending to be a $2T one.
- Others argue that, given current run‑rate and growth, hitting $300–400B+ revenue by 2030 isn’t impossible if growth merely decays gradually and margins hold.
- Retail investors are repeatedly warned against FOMO: outcome distribution ranges from huge upside (if near‑term AGI + Anthropic “wins”) to catastrophic downside (overcommitment to compute, commoditization, open‑model competition).