Anthropic's IPO prospectus shows AI vision, surging costs

Commenters focused on Reuters' coverage of Anthropic's IPO prospectus, centering on the disclosed 2025 figures - a near-$42 billion GAAP loss, a roughly $34 billion accounting charge, huge future infrastructure obligations and about a quarter of revenue tied to two customers. Some users (simonw, jfrbfbreudh) argued those 2025 numbers are misleading without 2026 data and said 2026 revenue growth has been reported as enormous; others (vmg12, htrp, theqO) noted Reuters only saw the 2025 filings or that claims are largely company-released and not audited. Several commenters (sensanaty, camdenreslink, DebtDeflation) expressed shock at scale and concentration risk, while others asked for detailed breakdowns of inference margins, training costs and the nature of the accounting charge (kamranjon, lokar).

Opinion split over whether this signals a doomed business model or a normal phase for transformative tech. Some compared LLM economics to capital‑intensive industries like airlines and telecom - transformative but poor ROIC (2ss, throw0101a, pid-1) - while others doubted the breadth of real use beyond coding and NLP (elzbardico, runarberg). A few suggested the IPO/leak could be strategic and that convertible-debt revaluations explain the charge (AntiRush, FuckButtons), whereas pessimists warned the company may be passing risk to public markets (knuppar).