The OpenAI–Anthropic Convergent Bets

Four scaling — as explored in the emerging fifth paradigm of scaling — paradigms — pre-training, chain-of-thought, test-time compute, agentic loops — stack rather than sequence. The agentic loop is where the stack compresses into usable capability. Coding is the ideal sandbox because it has ground truth. Both labs know this. Both are pivoting their entire product surface around the agentic coding wedge.

The barbell that separated them for three years — Anthropic’s enterprise focus without infrastructure — as explored in the economics of AI compute infrastructure — , OpenAI’s infrastructure without enterprise focus — is closing:

Anthropic: run-rate revenue now exceeds $30 billion, up from approximately $9 billion at the end of 2025 — roughly 3.3× growth in four months

OpenAI: ~$25B ARR against an $852B valuation from the March round