Anthropic Tops OpenAI as Two Firms Capture About 89% of Nearly $80 Billion in AI Startup Revenue

A May 18, 2026 analysis finds 34 AI startups generating almost $80 billion in annual revenue — up 112% in six months — with Anthropic overtaking OpenAI largely on demand for its AI coding tools and the two firms together taking roughly 89% of the total.

Anthropic has overtaken OpenAI in reported revenue, a shift driven largely by demand for its AI coding tools, according to a May 18, 2026 analysis that found 34 AI startups now generate nearly $80 billion in annual revenue — a 112% increase versus six months earlier. That rapid growth matters because it has concentrated revenue at the top of the emerging AI stack, reshaping where value and investment are flowing.

Beyond the two leaders, the dataset shows a small group of mid‑sized players gaining traction: Perplexity, ElevenLabs and Cognition have each crossed the $500 million revenue mark. Those companies sit beneath the duopoly, signaling that a handful of firms are establishing sustainable commercial footholds even as most revenue pools toward model makers. Headline totals are muted by contractual arrangements that divert receipts before they hit startups’ books. Anthropic splits revenue with cloud partners including Amazon and Google, while OpenAI is committed to paying Microsoft roughly 20% of its revenue through 2030. These deals reduce headline revenue for the companies themselves and reshape margins for downstream builders.

The analysis also flags heavy cost dynamics: Anthropic and OpenAI together burn through more than $30 billion a year, mostly on training compute. That capital intensity helps explain why the two firms capture about 89% of revenue among the top startups in the dataset. Investors such as Sequoia interpret the numbers as evidence that long‑term value will accrue primarily to model makers rather than pure application companies, influencing funding and go‑to‑market expectations.

For product teams and independent app builders the implications are concrete. Revenue‑share deals and platform cuts will compress net takeaways; very large training bills make models capital intensive; and standalone application companies may need differentiated IP or direct partnerships with model providers to sustain growth. The near‑term outlook points to continued consolidation, aggressive compute optimization and careful contract planning when launching new products.