WRITTEN BY

Will Bryk

PUBLISHED ON

Sep 3, 2025

We’re thrilled to announce that Exa has raised an $85m Series B to be the search engine for AI.

The round was led by Benchmark at a $700m valuation, with participation from Lightspeed, Nvidia, and YCombinator. Peter Fenton from Benchmark will join our board. Peter‘s a legendary investor who’s taken 7 companies to IPO and who passionately shares our vision.

Exa already serves web search to thousands of companies, from AI startups like Cursor to top private equity and consulting firms. We’ve built a state-of-the-art search engine from scratch that is both the fastest and highest quality for AI applications.

That was all stage one. We now enter stage two.

Origins of Exa

We started Exa in 2021 – far before AI needed search. We believed back then that the world needed a better search engine than Google and that we could build it.

So we bought a GPU cluster, built a large-scale indexing system, and tried (many) new techniques for searching the web.

Our goal was a search engine that gives users full control over the web in ways Google can’t - e.g. “give me all ML engineers in NYC who have a blog, and sort by years of experience”.

We launched our first search engine in November 2022. Two weeks later ChatGPT came out. We quickly started getting requests for API access to our search. People were building the first generation of AI apps, and those AI apps needed information from the web.

That’s when we realized – AIs need web search! In fact, we realized AIs would soon search the web FAR more than humans do.

Luckily, Exa was a perfect fit to be the search engine for these AIs.

What’s a search engine for AI?

Like humans, AIs don’t have all the world’s information in their heads. AIs need to search the web to get the most up-to-date and comprehensive information, whether news, code, research papers, company data, etc.

But AIs are very different creatures from humans. They require a new type of search engine. While traditional search engines were designed decades ago for humans, Exa is the only search engine designed for AIs to use. Here are some reasons why:

High Quality Knowledge

AIs just want the highest quality knowledge, not SEO or ads. Garbage in, garbage out.

Exa’s ranking algorithm is optimized purely for high quality knowledge. We have no ads and thus have no perverse incentives other than the highest quality search possible.

Full Content

AIs need more than just urls and titles. They want to consume as much information as possible about each result.

Exa provides full page content for each result for the AI to process all the necessary information on the page.

Fast

AIs need faster search than humans, because they often use multiple tool calls within a single request, search being just one. The latency adds up.

Exa built the fastest search API in the world at sub 450ms for exactly this reason.

High-Compute

Sometimes AIs actually don’t care about latency at all, and just want the most comprehensive search. This is particularly true for asynchronous applications.

Exa built a high compute search product called Websets, which is by far the most comprehensive search engine in the world. You can get massive lists of people, companies or really anything you want.

Customizable

Each AI application has a specific use case and therefore would be better off if the search was customized to the application.

Exa was built with customization in mind. We can do things like exclude 1000s of domains, or get 100s of results, or create a custom classifier to run on every search.

Zero Data Retention

Query data from enterprises is often very sensitive. Enterprises need a search API that has Zero Data Retention (ZDR), meaning the queries never get stored anywhere. Many search APIs actually wrap Google and therefore cannot offer true ZDR.

Exa built a search engine from scratch, so we can and do offer ZDR for all our product endpoints.

Most AI companies are starting to integrate search like Exa — from Cursor searching for technical documentation to Notion AI searching the news to large financial firms searching financial data within their custom AI workflows. The optimal AI products won’t wrap traditional search engines built for humans, they’ll use a search engine designed for AI.

The next stage

We’ve come far, but everything we said was just stage one of Exa. With this funding, we enter stage two.

Our goal isn’t just to build better search than anything before — our goal is perfect search. We dream of a world where all information is fully organized.

To get there, some immediate next steps are to:
1. Scale up our indexing/processing in order to gather the vast majority of the world’s information
2. Get a 5x bigger GPU cluster so we can research way faster and develop the new techniques that are necessary to truly organize the world’s information far beyond Google.
3. Expand our world-class team to build, sell, and lead Exa into our next stage.

Exa really is just a group of super smart and fun people in a large room working together on really hard problems. If you want to help us build critical societal infrastructure that doesn’t currently exist, come join. We’re hiring across engineering (backend, infra, ML research engineers), GTM, and operations.exa.ai/careers

And if you want to try Exa yourself, we have 2 products: the API and Websets. Try it on our home page here: exa.ai

SEE MORE

Zero Data Retention across Exa Search Products

Enterprises can now do things like private deep research reports

Will Bryk

August 18, 2025

The world's fastest search API

AI systems need faster search than humans. They now have it.

Will Bryk

July 29, 2025

Notion

Bringing real-time web data to Notion Research Mode

The Exa Team

September 2, 2025