Anthropic is asking Wall Street to believe in AGI

The Verge
Anthropic's S-1 filing reveals strong AI economics and ambitious AGI bets ahead of a potential $1.5–2 trillion IPO.

Summary

Anthropic has filed its S-1 to go public on the Nasdaq at a valuation potentially between $1.5 trillion and $2 trillion. In this Decoder interview, SemiAnalysis analyst Joey Brookhart breaks down the filing. Anthropic has scaled rapidly to over $70 billion in annualized recurring revenue with strong gross margins, up from deeply negative inference margins two years ago, driven by token throughput gains across a diverse chip fleet (Trainium, TPU, GPU). The company rents compute rather than owning it, with large commitments to hyperscalers and neoclouds, and derives nearly half its business through cloud marketplaces like AWS Bedrock and Google Vertex. A headline $80 billion loss is a non-cash accounting charge from warrants tied to its rising valuation, not an operating loss. The central tension is Anthropic's AGI ambitions: it is structured to keep investing heavily in training and accept thinner margins, betting that AI will diffuse into white-collar labor and new verticals such as cybersecurity and legal. Brookhart weighs risks including competition from open-weight models and rival labs, customer concentration, potential dilution, and whether revenue will arrive fast enough to justify massive compute commitments.

(Source:The Verge)