Nvidia Invests in Exa’s $85M Round to Boost AI-Powered Web Search for Developers

Nvidia Leads $85M Funding Round for AI Search Startup Exa Labs

Exa Labs Inc., a startup building a search engine tailored for artificial intelligence models to browse the web efficiently, has secured $85 million in a Series B funding round. Benchmark led the round, joined by Nvidia Corp., Y Combinator, and Lightspeed. With this investment, Exa’s valuation has reached $700 million.

Why Exa’s Search Engine Matters for AI

Many enterprise AI applications rely on retrieving accurate and relevant data from the web. Exa’s search engine aims to do this faster and more effectively than existing options like Google Search. The company claims its system can handle over 100 queries per second with latency under 450 milliseconds. This speed is crucial because AI tools often need to perform multiple searches within a single request, making low latency a necessity.

Developers can customize what data Exa extracts in response to queries. For instance, one application might only need webpage titles and URLs, while another might require full webpage content. Exa also offers summarization of the extracted text before passing it to AI models, reducing processing time and data overload.

Additionally, developers can apply custom algorithms to filter or refine the collected data automatically, enabling more precise results and task automation.

Developer-Friendly Access and Tools

Exa makes its search engine accessible via an API that integrates into applications with minimal coding effort. Thousands of organizations have adopted its platform so far.

Alongside the search engine, Exa offers Websets, a data retrieval tool that supports more complex queries and verifies results using AI agents. This tool is particularly useful for large-scale tasks such as collecting data on thousands of e-commerce listings simultaneously.

Technical Backbone: Customized Database for Fast Embedding Searches

Exa’s software runs on a custom database optimized to store embeddings—mathematical representations AI models use to store and retrieve information. The database compresses these embeddings to save storage and organizes them into datasets containing up to 100,000 items each.

When a query is executed, Exa searches only the relevant dataset instead of the entire database, improving speed considerably.

Future Plans and Use of Funds

The new funding will be used to expand Exa’s hardware infrastructure significantly. The company plans to increase its graphics card cluster fivefold, enhancing the internal research capabilities that currently run on 144 Nvidia H200 GPUs and 3,456 CPUs.

Exa also intends to scale its indexing and processing systems to cover a larger portion of web data. To support this growth, the company will hire more executives, developers, and sales staff.

This investment signals strong confidence in Exa’s approach to improving AI data retrieval, which could have wide-ranging impacts on enterprise AI applications.

For developers and IT professionals interested in AI tools and data retrieval technologies, further exploration into AI search APIs and embedding databases can provide valuable insights. You can explore relevant courses and resources on Complete AI Training.