A new analysis from The Information reveals that the AI startup market has become a two-company race faster than almost anyone predicted. Thirty-four leading AI startups collectively generate nearly $80 billion in annualized revenue - and Anthropic and OpenAI account for 89 cents of every dollar.
The combined share of Anthropic and OpenAI has surged to 89 percent of annualized revenue among the 34 leading AI startups tracked by The Information. That figure is up 4.5 percentage points from six months earlier, and up 36.8 percentage points from early 2023 when both companies began full-scale services. DigitalToday
The $80 billion in annual revenue among the group is up 112 percent in just six months. Anthropic recently passed OpenAI in revenue, largely driven by its AI coding tools. Beyond the two leaders, Perplexity, ElevenLabs, and Cognition have each crossed the $500 million mark in annualized revenue. The Decoder
What the Numbers Actually Mean
Before taking these figures at face value, it is worth understanding how they are constructed. The revenue figures for Anthropic and OpenAI are somewhat inflated. Anthropic shares revenue with cloud providers Amazon and Google, which resell its models to their respective customers. OpenAI has to hand 20 percent to Microsoft through 2030. Strip out pass-through revenue and the concentration is still significant, but less extreme than the headline number suggests. MediaPost Publications
What matters more than the exact figure is the direction. Every six months, the two leaders are capturing a larger share of a rapidly growing market. That is the definition of a winner-take-most dynamic - and it has major strategic implications for every company in the AI supply chain.
The Sequoia Argument and Why It Matters
Investment firm Sequoia Capital noted that the value resides in the application layer rather than the models, arguing that tools people actually use will win and that foundation models will become commoditized over time. MediaPost Publications
This is a live debate with real stakes. The data currently supports the opposite of Sequoia's thesis - model makers are capturing more revenue, not less. But the argument that application-layer companies will eventually dominate is not wrong on its face. It just may take longer than expected.
Despite the growth, Anthropic and OpenAI together burn through more than $30 billion a year, mostly on training costs. At some point, those economics have to change. The companies building on top of their models have lower infrastructure costs and often stronger direct customer relationships. The race is not over. The Decoder
For executives evaluating where to build AI-dependent workflows, the concentration of revenue in two providers is a vendor risk worth managing. Diversification across model providers is not just a cost strategy - it is a resilience strategy.
