Exa Labs has raised $250 million in a Series C led by Andreessen Horowitz, valuing the San Francisco startup at $2.2 billion. That is more than triple the $700 million valuation it carried when Benchmark led its $85 million round last September.
The pitch is simple and the timing is sharp. When an AI agent needs to look something up on the web, it does not type a query into Google. It calls a search API. Exa wants to be that API.
The company, founded in 2021 by Harvard roommates Will Bryk and Jeffrey Wang, built its search engine from scratch using neural embeddings rather than keyword matching. That means it can handle the long, natural language queries that AI agents generate, the kind that break traditional search engines designed for three-word human inputs.
The customer list tells you who is buying
Exa powers the @web search feature inside Cursor, the AI coding tool. It also provides search for Cognition, HubSpot, Monday.com and OpenRouter, along with more than 400,000 developers. The company says its API is now the default for developers building AI agent workflows that need web data.
Andreessen Horowitz partner Sarah Wang framed the opportunity in stark terms: agents will search the web more than humans this year, and soon the volume will be orders of magnitude greater than all Google searches combined. If that prediction holds, the company supplying the search infrastructure sits in a position of considerable leverage.
A crowded and fast-moving market
Exa is not alone. Parallel Web Systems, led by former Twitter CEO Parag Agrawal, raised $100 million at a $2 billion valuation from Sequoia. Tavily and TinyFish are chasing the same niche. Perplexity has built a consumer-facing AI search product that has attracted investment from sources as varied as Jeff Bezos and Cristiano Ronaldo.
The competitive dynamics are unusual. Google dominates human search but has an advertising model to protect. OpenAI owns the conversational interface through ChatGPT but cannot make search its sole focus. That gap, between the incumbents' priorities and the emerging demand from AI agents, is where startups like Exa are positioning themselves.
The infrastructure thesis
The investment case for Exa is an infrastructure play. Every AI agent that reasons, plans and acts on a user's behalf needs access to fresh, accurate web data. As agent workloads scale, the volume of programmatic search queries could dwarf consumer search within a few years.
Related reading
- Washington tightens its grip on AI as Google, Microsoft and xAI agree to government safety checks
- Demis Hassabis says AGI is missing two critical ingredients
- Google redesigns the search box for the first time in 25 years and bets the whole thing on AI
The risk is that Google, Microsoft or OpenAI decide to build or acquire their way into the space. Exa's advantage is that it has spent five years building indexes and retrieval systems purpose-built for machine queries. Replicating that from inside an organisation optimised for human search is harder than it looks.
A $2.2 billion valuation for a search API startup would have sounded absurd two years ago. In a world where AI agents are about to become the web's most prolific users, it starts to make sense.