Anthropic’s third quarter results mean IPO math and history are not on its side, analysts say

Anthropic is seeking a $1.5 trillion–$2 trillion IPO valuation, implying 43–57 times its approximately $35.3 billion in revenue over the past four quarters; analysts questioned whether growth and profits can justify that price. In the third quarter, revenue rose nearly 47% quarter over quarter, but compute and infrastructure expenses climbed about 77% to $9.9 billion, and adjusted operating margin fell to 15% from 19%. The company had $131.8 billion in cash, equivalents and short-term investments at September 30, while a study found 14 historical IPOs with similarly high price-to-sales ratios underperformed the broader market by an average 15.4 percentage points over holding periods of up to three years.

Analysis

The key issue is not one quarter’s margin decline; it is whether incremental compute spend creates durable, monetizable capability or merely buys capacity ahead of demand. If short-term capacity remains a recurring feature, reported revenue growth can coexist with weaker cash conversion and a lower eventual margin ceiling—undermining the long-dated profit assumptions embedded in a very high IPO multiple. Conversely, if that spend unlocks product adoption and later inference efficiency, today’s margin pressure may be investment timing rather than deteriorating economics.

The near-term read-through to NVIDIA is directionally supportive for accelerator demand, but not automatically for long-run earnings quality: urgent capacity purchases can pull demand forward, while financing and customer-economics constraints may eventually cap deployment. Alphabet and Microsoft could benefit if buyers favor platforms that bundle models with distribution and infrastructure; Anthropic’s results alone do not establish share gains or losses for either. Apple has no clear direct read-through here.

For the IPO, the payoff is asymmetric: cash reserves reduce immediate funding risk, but do not remove execution or valuation risk. A lower-end offer price may still leave little room for forecast misses. The contrarian point is that quarterly margin noise is not decisive if model capability drives step-change adoption; watch revenue conversion and cash economics rather than gross margin alone. Over 1–3 months, pricing and updated financial disclosure are catalysts; over 6–18 months, compute commitments, cash generation and evidence of improving efficiency determine whether premium growth converts into profits. Falsifiers: sustained acceleration in revenue alongside improving operating cash generation would weaken the bear case; continued margin compression and rising capacity commitments would strengthen it.

Key Decisions for Investors

• Treat the IPO as a valuation-discipline event, not a momentum allocation: do not underwrite the upper end of the proposed range; use roughly $1.4 trillion (about 40x the cited trailing sales) as a ceiling for reassessment, not a fair-value target. Verify updated trailing revenue and offer terms before acting.
• Avoid a standalone short before listing or without confirmed borrow and liquidity. If listed, consider an underweight/avoid stance at a premium valuation; reassess after the first reported quarters and any lockup-related supply, rather than assuming a smooth short-term decline.
• Keep NVIDIA as a watch rather than a direct trade on this report. A near-term capacity scramble could support demand, but seek confirmation in company guidance and broader AI-capex data before adding exposure; rising urgency without evidence of customer monetization would worsen the risk/reward.
• Track Anthropic’s operating cash generation, short-term compute spend as a share of total compute costs, and forward capacity commitments. Improving cash conversion alongside growth would falsify the margin-pressure thesis; further margin deterioration or new borrowing to fund capacity would reinforce it.