A year ago, I joined Pinecone with a simple thesis: every company sits on data its AI cannot use, and whoever closes that gap wins the next decade of enterprise software. Twelve months later, the thesis has held. And the past six months have been the best stretch in this company's history.
The numbers tell part of the story. Serverless Database retention is running above 130%, the clearest sign that customers keep expanding what they run on Pinecone as their knowledge needs grow. Annual commitments now make up roughly 40% of our revenue, up from roughly 25% three quarters ago. Our committed backlog has grown more than 60% since the start of the year, and revenue is compounding quarter over quarter and accelerating. We have begun a genuine business transformation, adding an enterprise flywheel to an already strong developer-centric business.
We made just as much progress on the product. We went from a single product to a platform: Database, Nexus, and Marketplace, spanning multiple surfaces and personas, serving vectors, knowledge, and work to be done. Gross margins swung by more than 67%, and Serverless gross margins by 96%. We now ship a full portfolio, including full-text search and an enterprise-class BYOC, with more in the pipeline as we execute our "Pinecone Everywhere" strategy.
Numbers are a lagging indicator of something better: trust. This year we began building a real enterprise business, and some of the most demanding enterprises in the world chose Pinecone as the knowledge layer under their AI. One of the world's largest automakers, rebuilding how its people find answers across a global business. One of Latin America's largest e-commerce companies, grounding agents that serve hundreds of millions of shoppers. A dominant design platform, trusted for both enterprise brand governance and everyday creativity. A major US media and financial-data publisher. Multiple Fortune 500 manufacturers landed the same way, brought in by partners who share our commitment to making complex, regulated industries AI-ready. This is what a market settling around a few winners looks like, and we intend to be one of them.
None of this would have been possible without Edo. Seven years ago, he saw what almost no one else did: that vector search would become the substrate of the AI era. He built the category that proved it. A year ago, he handed me stewardship of the company he founded, a trust I do not take lightly. Edo is now stepping back from his day-to-day responsibilities at Pinecone. He remains deeply committed to the company as a shareholder and board member. I am grateful for his vision, his dedication, and his continued partnership as we write the next chapter of the Pinecone story. Thank you, Edo.
A year ago, Pinecone meant one thing: the database that pioneered vector search. Today it is a trusted knowledge platform. Database remains the foundation more than 1M developers and 10,000 customers build on. Nexus is where the founding thesis becomes a product. It is the knowledge engine that turns the data a company already has into knowledge its agents can use: grounded in the company's own sources, current, and cited back to where it came from. Nexus cuts the tokens an agent burns on a task by more than 90%, completes that work up to 30x faster, and shows its provenance on every answer. If the last era of software was about giving people applications, the next is about giving agents knowledge they can trust, and Nexus is how we close that gap. Marketplace then puts that same knowledge to work for the hundred thousand people who never write a line of code. Models are commoditizing, and the only moat that does not erode with the next model release is knowledge: how a company works, what it has learned, who it trusts. We are becoming the trusted AI knowledge platform for every developer, agent, and business team, wherever they sit.
Everything above was built by this team. Thank you for the year, and for the standard you set while doing it.
The past year was the best we have had. I believe the next one will make it look like a warm-up.
Ash
Was this article helpful?