While there’s no shortage of startups looking to replace Google with AI-powered search (we’re looking at you, Perplexity), a startup called Exa has a different idea: Google for AI.
People aren’t the ones who desperately need a new kind of search engine, Exa’s founders believe. Rather, as AI increasingly takes hold in corporate and consumer life, it is the AI platforms themselves that must regularly go online to search for information and return bona fide answers, not hallucinations. And they can’t just type their requests on their keyboards.
Exa is building a tool that allows AI models to perform a sort of web search, but with an AI flavor.
The co-founders bought a million dollars worth of GPUs (which were easier to get in those days) and, using a vector database and embeddings (rather than a classic transformer-based LLM), they started building a machine learning model trained to understand naturally links and not words and sentences.
“Transformers usually predict the next word. We’re training our search engine to predict the next link,” says Bryk. “So people share links on the web; We use this data as the dataset for our model that we train. And we train the model to predict the next connection. So this is a new search algorithm.
So just as LLM would complete a sentence by providing the most likely next word, Exa’s system does so with the most likely link (or ten), but possibly minus the SEO spam and (ironically) the AI-generated traffic jam friend any regular search engine These days.
On Monday, the startup announced it had raised a new $17 million Series A round led by Lightspeed’s Guru Chahal, with participation from Nvidia’s venture arm NVentures and Y Combinator, it told TechCrunch exclusively. Exa has already raised a total of $22 million, including the previous seed of $5 million. (Exa was in summer, YC cohort 2021)
“It’s a very ambitious vision,” says Chahal. “What Google is for humans, they’re building for AI.”
The team was founded about a year before ChatGPT launched by two best friends who met their freshman year at Harvard: CEO Will Bryk (now 27) and co-founder Jeff Wang (26).
“We launched before ChatGPT. Our original goal as a company was not to serve AI at all. It was: How do you use AI to create better search?” Wang said.
After ChatGPT took the tech world by storm, AI companies started asking Exa for an API version of their search engine that they could incorporate into their models. Exa is based in San Francisco, part of the cozy Cerebral Valley AI startup suite. In fact, as TechCrunch previously reported, Wang’s tweet went viral when he was looking for other companies that wanted to get involved in ordering office nap capsules, and the response was overwhelming. (The work-nap-repeat culture is alive and well in this part of the tech industry.)
With AI companies already its primary customers, use cases for Exa’s search engine range from an AI chatbot that searches the web for information while answering customer questions to companies looking to prepare data for training.
Databricks, for example, is Exa’s main customer, using it to find large training sets for its own model training initiatives, the founders say.
The API version of the product was released about a year ago. “Since then, it’s gotten a crazy amount of traction,” says Wang. Today, Exa says it serves thousands of developers — although it’s worth noting that Exa has a free tier that allows anyone to try out its search engine in a limited way. It also has multiple tiered fee levels. The founders wouldn’t disclose revenue, other than to say they have it and it’s growing. (Interestingly, in addition to running its own GPU cluster, Exa hosts its product on AWS rather than the AI-centric Google Cloud.)
The team isn’t particularly focused on being the search startup that turns Google upside down. Although, if AI becomes everything the tech industry thinks it will be, AI bot search engines could be the unexpected threat to search hegemony.
