Exa company profile: valuation, funding & revenue

This Exa company profile covers the company's valuation, revenue, funding rounds, investors and history. Exa works in artificial intelligence, with a focus on web search infrastructure for AI. It was founded in 2021 and is headquartered in San Francisco, California, United States. Its latest valuation on record is US$2.2 billion (May 2026). Every figure links to a public source, and anything the company hasn't disclosed is marked as such.

Revenue

Funding history

Exa timeline

Key dates from the sources on this page, oldest first. Metric snapshots are listed after the funding events.

• 2021 — Exa is founded in San Francisco by Will Bryk (CEO) and Jeff Wang.
• 2025 — Series B: US$85M, valuing the company at US$700M. Investors: Benchmark (lead, reported).
• May 2026 — Series C: US$250M, valuing the company at US$2.2B. Investors: Andreessen Horowitz (lead); Benchmark, Lightspeed Venture Partners, Y Combinator.
• May 2026 — Users: More than 5,000 businesses and 400,000 developers (company-stated).
• 2025 — Previous valuation: US$700 million after an US$85 million round.

Investors

Exa builds a search engine designed for AI rather than people. Developers and companies use its API to let AI models and agents find and read up-to-date information on the web. In May 2026 it raised a US$250 million Series C led by Andreessen Horowitz at a US$2.2 billion valuation, more than triple its value from the previous autumn. This profile explains what Exa does, why AI needs its own search engine, and how it competes.

Exa's valuation

Exa's Series C of US$250 million, led by Andreessen Horowitz with existing backers Benchmark, Lightspeed Venture Partners and Y Combinator, valued it at US$2.2 billion, according to Bloomberg, and law firm Latham & Watkins advised on the deal. Bloomberg said the valuation was more than triple the US$700 million at which Exa raised US$85 million in late 2025. Secondary-market prices in September 2026 were roughly in line with the round.

What Exa does

Traditional search engines are built to show people a page of links ranked partly by advertising and clicks. AI systems need something different: precise, clean, machine-readable results they can feed into a model's context. Exa crawls and indexes the web itself and uses neural embeddings, numerical representations of meaning, to match queries to pages by what they are about rather than by keywords alone. Its API returns results with full page contents, so an AI agent can read them directly.

Products include a search API, a "contents" API for retrieving page text, an answer endpoint and a research agent that runs many searches to compile reports. Exa also offers specialised indexes, such as for companies, people and research papers, and "Websets", which assemble lists of entities matching complex criteria, useful for lead generation, recruiting and market research.

Customers and use cases

Exa says more than 5,000 businesses and over 400,000 developers use its tools, according to PYMNTS. Customers include AI application companies that need web search inside their chatbots and agents, research and finance teams, and companies building internal AI tools. Exa plans to use the new money to train new models, expand its infrastructure and handle hundreds of thousands of searches per second.

Founders

Will Bryk, the chief executive, and Jeff Wang founded the company in 2021, originally as Metaphor, after studying at Harvard. They began building neural search before ChatGPT made AI assistants mainstream, which gave Exa an early start on the infrastructure AI applications now need.

Why AI needs its own search engine

Large language models are trained on data up to a cutoff date and can make up facts. Connecting them to live search, often called retrieval-augmented generation, makes their answers more current and verifiable. As AI agents take on research and multi-step tasks, they may run dozens of searches for a single request, far more than a person would. That creates demand for search that is fast, cheap at high volume and returns content in a form models can use, without the ads and layout of consumer search pages.

Competitive landscape

Exa competes with search APIs from Microsoft (whose Bing API changed in 2025), Google, Brave and Tavily; with AI search companies such as Perplexity, which offers its own API; with you.com; and with model providers that build search into their own products. Its edge is an index built for AI and specialised datasets. See our guides to AI infrastructure and AI agents.

Risks

• Big Tech competition: Google and Microsoft own the largest web indexes.
• Model providers: OpenAI, Anthropic and Google offer built-in web search.
• Content access: publishers increasingly restrict crawling and seek licensing payments.
• Cost of scale: crawling and indexing the web is expensive.

Questions to ask

• How large is Exa's index, and how often is it refreshed?
• What is its revenue and growth, and how is usage priced?
• How does result quality compare with Google- or Bing-based APIs?
• How does it handle publisher permissions and licensing?

Timeline in brief

• 2021: founded as Metaphor.
• 2024: renamed Exa; raises Series A.
• 2025: US$85 million round at US$700 million.
• May 2026: US$250 million Series C at US$2.2 billion, led by Andreessen Horowitz.

Exa at a glance

Exa is a San Francisco AI search infrastructure company founded in 2021 by Will Bryk and Jeff Wang. Its API lets AI models and agents search and read the web. It raised a US$250 million Series C led by Andreessen Horowitz in May 2026 at a US$2.2 billion valuation, up from US$700 million in 2025, and says more than 5,000 businesses and 400,000 developers use its tools.

How developers use Exa

A developer building an AI assistant can call Exa's search API with a natural-language query, such as recent research on a medical treatment or companies building battery recycling plants in Europe, and receive a list of relevant pages with their full text. The AI model then reads those pages and writes an answer with citations. Agents use Exa repeatedly during a task: searching, reading, refining the query and searching again. Exa prices by the number of searches and the amount of content retrieved, so its revenue grows with the volume of AI usage.

Exa's "Websets" product works differently. A user describes the kind of entities they want, for example Series A climate startups in Germany with a technical founder, and Exa searches, verifies and assembles a list. That makes it useful for sales prospecting, recruiting and investment research, areas where companies already pay for data.

Valuation context

At US$2.2 billion, Exa is valued as a core piece of AI infrastructure. That price reflects the belief that a meaningful share of web searches will soon be made by AI agents rather than people, and that those agents need a dedicated provider. If that shift happens at scale, search infrastructure for AI could become a large market; if model providers keep search in-house, independent providers will have to compete on quality and specialised data.

What to watch

Watch for new model releases, major customer deals, publisher licensing agreements and competition from Big Tech search APIs.

Takeaways

Exa is building the search layer for AI agents. If agents become a main way people find information, the infrastructure they rely on could become very valuable.

Sources

Every figure on this page comes from the public sources below. Where sources disagreed, we used the company's own announcement or a primary news report. Private-company data changes; check the date beside each figure.

1. Bloomberg Government
2. PYMNTS