Exa says its index serves 80 billion pages, targets Google scale in 2027

Why it matters

If Bryk's figures hold, Exa is becoming a rare independent web-index operator. Its larger test is converting crawl scale into fresher, cheaper and more accurate agent search.

Exa CEO and co-founder Will Bryk (@WilliamBryk) said in an August 3rd post on X that Exa now serves 80 billion pages and tracks 1.4 trillion URLs. Bryk expects Exa to reach what he calls "Google-scale" in early 2027. The figures are Exa's own, and Bryk acknowledged that search providers make index sizes difficult to compare. (x.com)

Bryk has been working toward that target since before AI agents became a market category. He studied computer science and physics at Harvard, conducted machine-learning research and led the university's robotics club before working as an engineer at Cresta. Bryk and co-founder Jeffrey Wang, his Harvard roommate, built an early search engine in their dorm. Wang later worked on data and web infrastructure at Plaid. The pair entered Y Combinator's Summer 2021 batch with a plan to rebuild web search around semantic retrieval. (ycombinator.com)

The URL inventory has nearly tripled since May

The new figure of 1.4 trillion tracked URLs is nearly three times the "over 500 billion" URLs Exa said its crawlers tracked when it announced its Series C on May 20th. Bryk's post distinguishes that larger discovery pool from the 80 billion pages Exa says it can serve. That distinction matters: a crawler can discover or monitor URLs without retaining every page in the active corpus available to search customers. Exa's about page separately advertises infrastructure for more than 500 billion webpages. Exa has not published a technical reconciliation of its tracked, indexed and served metrics. (exa.ai)

Bryk estimated Google's index at roughly 1 trillion pages, Bing's at 500 billion and Yandex's at 200 billion. Those exact figures are Bryk's estimates rather than standardized disclosures from the search providers. Google says its index covers "hundreds of billions" of webpages and other digital content. Brave's Search API says its independent index contains more than 30 billion pages, while a newer Brave glossary places it above 40 billion. Differences in how providers count duplicates, stale pages, documents, discovered URLs and pages eligible for retrieval limit any direct ranking by size. (google.com)

Exa is spending to own the search stack

Exa's expansion follows a $250 million Series C led by Andreessen Horowitz at a $2.2 billion valuation in May. Exa said the financing would fund larger retrieval models and enough infrastructure to process hundreds of thousands of searches per second. That round followed an $85 million Series B led by Benchmark at a $700 million valuation in September 2025 and $22 million in earlier seed and Series A financing. Exa has announced $357 million in funding across those rounds. (exa.ai)

The infrastructure push supports a product line aimed at software developers and AI agents rather than a consumer search box. Exa says more than 400,000 developers and 5,000 companies use its search products, naming Cursor, Cognition, HubSpot, OpenRouter and Monday.com among its customers. Those adoption figures are company-reported. (exa.ai)

Exa has also been turning the underlying index into more specialized retrieval products. In April, Exa became a grounding provider and agent marketplace partner for Google Cloud's Gemini enterprise products. It released Exa Agent in June for research, list-building and entity enrichment, then added a dedicated index of roughly 350 million academic publications on July 23rd. (exa.ai)

A larger corpus gives Exa more control over coverage, freshness and ranking than search APIs that repackage results from an incumbent provider. It also creates a substantial infrastructure burden. Crawling more URLs does not by itself establish better search quality: Exa still has to keep useful pages fresh, discard spam and duplicates, rank long-tail results and deliver them within the latency and cost limits of agent workflows. Bryk's 80 billion-page claim shows how aggressively Exa is spending to close the scale gap. Whether that inventory produces better answers will determine if Exa becomes core agent infrastructure or another search API competing at the edge of Google and Bing.