Our Take
Anthropic is banking on capital appetite and unproven use cases—not demonstrated returns—to sustain growth at a $965 billion valuation.
Why it matters
The company's IPO filing coincides with real corporate concern about AI spending productivity (Uber and others have flagged wasted budgets). Investors need to know whether the revenue acceleration reflects actual business value or front-loaded enterprise appetite for models that haven't proven durable ROI.
Do this week
Finance leaders: audit your AI spend from the past 12 months against documented productivity gains before Anthropic's S-1 drops, so you can reality-check vendor growth claims against your own numbers.
Anthropic filed for IPO as revenue soars to $47 billion annualized
Anthropic announced a confidential IPO filing on the back of extraordinary growth: annualized revenue crossed $47 billion in May 2026, a jump from roughly $9 billion at the end of 2025 (company-reported). The filing follows a $65 billion fundraise at a $965 billion valuation announced last week, which was oversubscribed by multiple investors.
Co-founder Daniela Amodei framed the IPO decision as a capital strategy at the Bloomberg Tech conference. "It's a really big upfront cost to train the models and to serve inference on them," she said. "My guess is that over time, the sort of core set of companies that are working to advance the frontier are just going to need access to capital, and I think the public market is very well suited to that."
Anthropic's compute strategy remains outsourced. Unlike OpenAI and xAI, the company has not built its own data centers. Instead, it partnered with xAI for capacity at $1.25 billion per month (disclosed in SpaceX's S-1 filing). Amodei said the company prefers to risk undersupply rather than overcommit to compute.
The bet: corporate adoption will eventually justify the burn
Anthropic's revenue curve is steep, but the company is entering public markets during a period of corporate skepticism about AI returns. Uber and other large enterprises have publicly stated that not all AI spending has yielded measurable productivity gains, raising questions about whether corporate budgets for frontier models will sustain at current rates.
Amodei acknowledged the gap. She expects use cases in coding, financial services, legal work, and health care to remain primary drivers, but conceded that "as the business community gets more familiar with the tools, we're all going to learn together." Translation: current revenue reflects early adoption and exploration, not settled, repeatable customer outcomes.
The IPO filing is thus a bet on two things: (1) that private capital markets will fund AI infrastructure and model training for long enough for enterprises to mature their deployment strategies, and (2) that once they do, the TAM and unit economics will justify valuations that currently rest on growth momentum rather than proven ROI per customer.
How to read this: separate the growth signal from the use-case signal
Anthropic's revenue number is real and verifiable. What it measures is less clear. Five-figure ARR jumps often reflect volume of new contract bookings, not depth of value captured per customer or durability of those contracts past year one.
Before the S-1 arrives, practitioners should ask: Are we seeing sustained expansion from existing customers, or churn masked by new logos? Is the $47 billion run rate supported by multi-year commitments, or one-year pilots? Anthropic's IPO prospectus will force disclosure of these metrics. Compare them to your own customer cohorts and retention rates before accepting the growth narrative wholesale.