There might be few takers for Pinecone’s new serverless vector database, dubbed Pinecone Serverless, analysts believe.

“Why set up and administer a separate database—even one with the advantages of serverless scalability—if you can get the same functionality from the database you are already using and in which you are already managing your data?”, said Doug Henschen, principal analyst at Constellation Research.  

Other than mainstream vector databases, such as Milvus, Weaviate, and Chroma, vector embedding and search features have either already been added or are coming soon to database service providers, including MongoDB, Couchbase, Snowflake, and Google BigQuery, among others.

“The addition of vector embeddings and search make it harder for fledgling, vector-only databases to develop a big market,” Henschen said.

Vector databases and vector search, according to experts, are two technologies that developers use to convert unstructured information into vectors, now more commonly called embeddings.

These embeddings, in turn, make storing, searching, and comparing the information easier, faster, and significantly more scalable for large datasets. 

The scalability advantage of vector search has also helped it win favor among developers who are building applications based on generative AI as more data you can feed to a large language model (LLM), as and when required, the more accurate responses the model can generate, in turn making the top layer application more efficient.

However, the principal analyst said that he was not convinced that vector databases, such as Pinecone, with more bells and whistles eyeing developers and data scientists working on AI would force enterprises to pay for an additional database service that’s only used for development of AI-based applications.

Flat IT budgets could add to Pinecone’s worries

Moreover, the launch of Pinecone Serverless comes at a time when IT budgets of enterprises continue to remain flat.

“While there is a lot of interest in generative AI, budgets are not yet spiking accordingly,” said Tony Baer, principal analyst at dbInsight.

“The flat budgets can be attributed to the immaturity of the field; choices of everything from tooling to foundation models to runtime services are just in their infancy, and aside from copilots and natural language query, enterprises are still on the learning curve for identifying winning use cases,” Baer added.

Along with feeding demand for generative AI, Pinecone expects the new serverless database to help enterprises reduce cost and the need to manage infrastructure.

The cost reduction is made possible by separating reads, writes, and storage, the company said, adding that the database aims to reduce latency by adopting an architecture, under which vector clustering sits on top of blob storage.

The database, according to the company, comes with new indexing and retrieval algorithms to enable fast and memory-efficient vector search from blob storage without sacrificing retrieval quality.

The new vector indexing, according to Baer, gives Pinecone an advantage over other vector and operational databases. Pinecone supports almost a dozen index types, the analyst said.

The serverless atrribute of the database, too, Baer says, is the need of the hour.

“The nature of retrieval augmented generation (RAG) workloads is that they will have the characteristics of any query-driven workload (think analytics), which are spikey in nature. Without serverless, customers must provision “just-in-case” capacity that is likely to often sit silent,” Baer explained.

A secondary reason for Pinecone taking the serverless route is to help ease developer complexity as it eliminates the need to provision servers.