Exa raises $250M Series C at $2.2B valuation for AI search

What's the deal? Exa (formerly Metaphor), a San Francisco-based startup building search infrastructure for AI agents, has closed a $250M Series C led by Andreessen Horowitz. The round values the company at $2.2B — more than tripling its $700M valuation from its Series B just eight months ago.

Why now? The rise of agentic AI has created urgent demand for machine-readable web data. Traditional search engines optimise for human clicks and ad revenue; Exa's API uses vector embeddings to deliver structured, high-fidelity results to LLMs in milliseconds. As autonomous AI programmes proliferate, whoever owns the data layer they read from holds a critical chokepoint.

Exa has positioned itself as the "picks and shovels" play in this shift — not competing with consumer tools like Perplexity AIDealroom has a profile for this one. Try Dealroom →, but powering the infrastructure beneath them. CEO Will Bryk describes the product as translating the web for machines.

What could go wrong? Google is not sitting still. Its AI Overviews and pivot toward semantic search aim to close the gap. If incumbents crack machine-optimised search without cannibalising their ad models, Exa's moat narrows fast. And at $2.2B, the company must convert developer traction into durable, scaled revenue.

The signal: The round underscores a broader conviction that AI's data infrastructure needs to be rebuilt from scratch. As LLMs move from chatbots to autonomous agents, the companies that control how AI reads the web will wield outsized influence — and attract outsized capital.

Read more: Bloomberg

Image credit: Andreessen Horowitz