Exa secures $17M from Lightspeed, Nvidia, YC to develop AI's Google

Innovative Startup Aims to Provide AI with a Tailored Search Engine

A new trend is emerging among startups looking to innovate in the search engine space. Unlike the many startups focusing on replacing Google for human users, one company stands out with a unique approach: creating a “Google for AI.” Exa is addressing the need for AI platforms, rather than humans, to have a specialized search engine capable of retrieving accurate information efficiently. With AI systems increasingly becoming integral in both corporate and consumer settings, the requirement for these platforms to fetch reliable data has become paramount.

Exa, building on this idea, focuses on a tool that allows AI models to execute web searches uniquely suited for them. Instead of relying on traditional user input methods, Exa’s founders invested in a million dollars worth of GPUs and developed a machine learning model. Utilizing a vector database and embeddings instead of classic transformer-based LLMs, their model understands links rather than words and sentences. This innovative approach enables their search engine to predict the next link in a sequence, using web links as a dataset for training.

A Novel Approach to Web Search

In traditional language models, predicting the next word in a sentence is the standard. Exa’s model, on the other hand, predicts the next link, promising a novel search algorithm. This approach aims to avoid the pitfalls of SEO spam and AI-generated content that flood conventional search engines. Recently, Exa announced it raised $17 million in a Series A funding led by Lightspeed and included participation from NVentures and Y Combinator. The total funding now amounts to $22 million.

Founder's Vision and Evolution

Founded by Will Bryk and Jeff Wang, Exa was conceptualized about a year before the launch of ChatGPT. Initially, the company's goal was to leverage AI for enhanced search capabilities. However, with the advent of ChatGPT, AI companies began expressing interest in an API version of Exa’s search engine, signaling a pivot towards catering primarily to AI applications. Situated in San Francisco’s Cerebral Valley, Exa has become part of a burgeoning scene of AI startups.

Use Cases and Market Adoption

Exa’s search engine is now utilized for various purposes, including AI chatbots retrieving internet data to answer user questions and companies curating training datasets. Databricks, a notable customer, employs Exa’s technology to gather extensive training sets for its models. Since launching the API version approximately a year ago, Exa has experienced significant traction among developers. The company offers a free tier for limited use and multiple fee-based tiers, serving thousands of developers.

Despite Exa’s growing influence, the founders maintain their focus on enhancing AI-centric search capabilities rather than directly challenging Google's dominance. However, the rise of AI may well place specialized search engines like Exa’s in a position to impact the traditional search market significantly.