When Pinecone launched a vector database aimed at data scientists in 2021, it was probably ahead of its time. But as the use cases began to take shape last year, the company began pushing AI-driven semantic search. With the rise of LLMs in the public consciousness, companies are beginning to see the value of vector databases even more.
Investors apparently agree. Today, the company announced a $100 million Series B investment on a $750 million post valuation. These kinds of numbers have been hard to come by in a conservative investment environment, but the company is growing fast and investors saw an opportunity to grab a market leader, says Pinecone CEO and founder Edo Liberty.
“We are clearly the creators of this category and the leaders in it. When we came out with this, with the vector database category, nobody knew what the hell we’re talking about. Now of course, this is a well-formed market and the category has different players, and so incumbents and clouds and so on, and we’re clearly ahead. And so it’s very easy to bet on the leader of a category that is already formed,” Liberty told TechCrunch.
That first-to-market advantage is helping them grow from a handful of customers last year to 1,500 today, and Liberty says the growth rate is more like a consumer tool than a highly technical database. The company is seeing interest from businesses of all sizes, including technology companies like Shopify, Gong and Zapier.
“This is like consumer-based adoption of B2B deep tech stuff. I’ve never seen anything like this. And so you have to accelerate the building of those capacities, and that’s very expensive and very hard,” he said.
He says interest in LLMs is driving interest in the vector database, but it’s a different proposition. While both take large amounts of data and let you search on them, with an LLM the data is baked into the model, and therefore less flexible, while the vector database is built for semantic search, but has the flexibility of a database.
“This whole knowledge management scheme ends up being a lot more flexible, a lot more efficient, a lot easier to operate [with a vector database],” he says. He points to GDPR compliance as an example. If you have to delete a record, it’s fairly trivial to do in a database, but it’s much more difficult to remove bad data from a model because of the way it’s structured.
Peter Levine, who is leading today’s investment for Andreessen Horowitz and will be joining the Pinecone board, sees the vector database as a key piece of the AI stack. “We believe that this vector database and especially Pinecone has the opportunity to be really a fundamental component in the new AI data stack. And so we really feel very strongly that putting the resources behind the company is going to help it achieve Edo’s end vision,” Levine told TechCrunch.
What’s more, Levine sees the vector database working together with LLMs to act as a source of truth, one that could presumably cut down on the hallucination problem we’ve seen with LLMs. “Well, they work together. I mean, think of the LLM as an application almost that sits on top of this database, and what the database will do is it will hold information and feed it into LLM for more precise answers for long-term storage of results,” he said. Liberty similarly thinks of this as the database acting as the long-term memory for the LLM.
With a $100 million runway, the company will be hiring. It’s up to around 100 employees today, and Liberty expects to get to perhaps 150 or 200 by the end of year.
The vector database space has been heating up since Pinecone launched a few years ago with players Qdrant, Zilliz and Chroma, all raising funds recently.
Today’s investment was led by Andreessen Horowitz with participation from ICONIQ Growth and previous investors Menlo Ventures and Wing Venture Capital. The company has now raised $138 million including a $28 million Series A last year and a $10 million seed investment in 2021.
Pinecone drops $100M investment on $750M valuation, as vector database demand grows by Ron Miller originally published on TechCrunch