- Vector database startups like Pinecone have received term sheets from top-tier VCs, Insider has learned.
- VCs like Andreessen Horowitz and Index Ventures are offering valuations ranging up to $700 million.
- These large rounds continue a trend of hype around startups in the buzzy generative AI ecosystem.
Within the span of just a few days in the past month, three startups — Chroma, Weaviate, and Pinecone — received term sheets and valuations ranging as high as $700 million, highlighting new investor interest in an AI category titled vector databases, Insider has learned.
Similar to the investor excitement around generative AI, VCs are now flocking to another area within the buzzy AI ecosystem. Vector databases help users search across complex, unstructured data like text, images, and videos using their actual content instead of human-generated tags or labels.
They accomplish this through vector embeddings, or representations of unstructured data that describe them as numeric values in hundreds or thousands of different dimensions, allowing them to be more easily grouped together based on similarities.
“It basically provides a database or long-term memory around all the calls and responses of unstructured data,” Essence VC managing partner Tim Chen said. “So therefore it allows you to be more flexible with how you want to call LLMs.” LLMs refers to large language models like OpenAI’s GPT-4.
Vector databases have grown in popularity as developers look for more efficient and accurate ways to build applications on top of LLMs.
For instance, one common use case for this technology is searching through data for similar items, or similarity search. If a developer is building a chatbot application on top of OpenAI’s GPT-4, they may not want to call the AI model every time a question is asked, if similar questions have been asked by users in the past. Instead, a vector database can store and quickly find past responses to similar questions by understanding the meaning behind the statements without having to call the base model.
Because the vector database space is so intertwined with LLMs, it’s benefitted greatly from recent hype around generative AI and LLMs.
“The LLM tax is real,” one VC said, referring to the high valuations for buzzy vector database startups like Chroma, Weaviate, and Pinecone.
Open-source embedding database Chroma is in talks to raise a funding round at a $75 million post-money valuation, according to six people with knowledge of the financing. Chroma declined to comment.
Some venture capital firms are trying to get ahead of competitors by offering term sheets outside of formal fundraising rounds, like Index Ventures, which preempted a funding round for open-source vector database Weaviate at a post-money valuation of at least $200 million, according to multiple people with knowledge of the financing. Weaviate did not respond to a request for comment, and Index Ventures declined to comment.
And when that isn’t an option, bidding wars ensue. This was the case for vector database Pinecone, according to one source, which eventually saw Andreessen Horowitz win out as the round’s lead investor with a post-money valuation of at least $700 million, according to multiple people with knowledge of the financing. Pinecone said as a company policy, they do not comment on rumors or speculation. Andreessen Horowitz did not respond to a request for comment.
These people spoke with Insider on condition of anonymity, as they were not authorized to speak publicly on these funding rounds.
These startups’ eye-boggling rounds continue a trend of hefty rounds and valuations for startups in the AI ecosystem, like Adept’s $350 million Series B round and Jasper’s $125 million Series A, both driven by hype and higher costs of talent and compute for these companies, VCs told Insider.
Have a tip on a startup funding round? Contact reporter Stephanie Palazzolo at spalazzolo@insider.com or through encrypted messaging at 979-599-8091.
Read the full article here
