Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More
When Edo Liberty was completing his Ph.D. in Computer Science at Yale on random projections, he could have hardly known that a decade later it would be a fundamental component of modern AI.
Liberty is the co-founder and CEO of vector database pioneer Pinecone, which has raised over $138 million including a $100 million round in 2023. As it turns out, random projections, which was his thesis topic, is a cornerstone of modern vector search, even as new innovations and use cases for vector databases proliferate. In 2024, vector database technology is no longer a niche or an outlier, but is a required component to enable Retrieval Augmented Generation (RAG) use cases with generative AI.
When Pinecone was founded in 2019, vector database technology was not widespread. That’s no longer the case as nearly every major database vendor including Oracle, MongoDB, DataStax and even Google Cloud all provide vector database capabilities.
Pinecone today is continuing to differentiate itself against other vector database technologies in several ways. Today the company announced the general availability of its Pinecone serverless database offering on all three major cloud vendors including AWS, Microsoft Azure and Google Cloud. In addition to the general availability, Pinecone is integrating a series of new features that expand the capabilities and practical utility of its vector database platform technology.
“We grew as a company from a tiny handful of people building a product that nobody has heard of, to being probably the hottest database category in the world,” Liberty told VentureBeat.
How the Pinecone serverless vector database works
Pinecone first previewed the serverless version of its vector database in January. The service first became generally available on AWS and with today’s announcement is now also available on Google Cloud and Microsoft Azure.
The basic promise of serverless is that organizations get an optimized, managed approach where cost is based on usage. Liberty emphasized that the benefit is ease of use, by removing the complexity of infrastructure service management.
“First of all, you as a customer have zero interaction with any concept of compute, you don’t choose node sizes or CPUs,” Liberty said. “You interact with reads and writes and storage in terms of capacity.”
The other key benefit of the serverless approach is scalability. Liberty said that the user shouldn’t care if they are starting an application that has five thousand or five billion vectors.
“You create an index and you start using the service,” he said.
New features expand Pinecone’s serverless vector database
With the general availability of the Pinecone serverless vector database across the three cloud vendors also comes a series of new features.
One of the new features is bulk import of data into Pinecone.
“That means that now if you have a large amount of data on one cloud, you can move to the other, or if you just have it somewhere else, you can create a huge index very easily and very cheaply,” Liberty said.
Pinecone is now also adding Role-Based Access Control (RBAC) to its serverless vector database offering. RBAC is a feature that is commonly associated with security, but that’s not the primary benefit for Pinecone’s users. Liberty said that the new RBAC feature will be a big help with data governance overall, providing access control functionality.
“When you build with a piece of infrastructure you want to be able to control who has rights to do what, in terms of reads and who can write, who can delete, role-based access control gives you that right,” Liberty said.
Alongside the database update, Pinecone is also debuting a new software development kit (SDK). The new SDK aims to make it easier for developers to integrate Pinecone into an application workflow, specifically for dot net applications.
Why Pinecone isn’t worried about vector database competition
With the proliferation of vector database support capabilities across multiple vendors, Liberty remains confident that his firm has solid differentiation.
In his view, database vendors that have multi-model approaches where the vector is just another data type are not able to outperform Pinecone. Liberty emphasized that vector has always been Pinecone’s focus and provides a strong competitive advantage.
“From day one, we have an outstanding developer experience, then once you get started, you start building, we are by far the most scalable, efficient, performing, cost-effective piece of software out there for vector search,” Liberty said. “We are very focused on production and enterprise readiness.”