SurrealDB raises $23m to make its database the memory for AI agents

On 18 February, SurrealDB said it had added $23m (£17m) to its Series A, taking the round to $38m (£28m) and its total funding to $44m (£33m). Chalfen Ventures and Begin Capital joined existing investors FirstMark and Georgian. Mike Chalfen, founder of Chalfen Ventures, has joined the board; Companies House lists him as a director from 26 September 2025. The round was first reported on 17 February, the day the company made SurrealDB 3.0 generally available, a release it titled “the future of AI agent memory”.

SurrealDB was founded by brothers Tobie and Jaime Morgan Hitchcock. Tobie Morgan Hitchcock is co-founder and chief executive; Jaime Morgan Hitchcock is the other co-founder and a director. In January 2023, TechCrunch reported that the brothers, after years of building cloud software, began asking in 2015 whether there might be a platform for building and scaling applications more quickly. Their earlier ventures included an app that let golf courses track how golfers play each hole and Hire Insight, a candidate assessment service. The company, SurrealDB Ltd, was incorporated on 10 September 2021, and each brother holds more than 25% and up to 50% of the shares, according to Companies House. It trades from Huckletree Oxford Circus in London.

The product is a multi-model database written in Rust. Relational, document, graph, vector, search, geospatial, time-series and key-value data sit in one engine with one query language, SurrealQL. The pitch is consolidation: a business running separate databases for transactions, search and vectors can, the company says, run one instead. Case studies on its website name the users. Samsung runs a knowledge graph for audience insights and ad targeting in its ad division. Tencent consolidated nine backend tools into one monitoring platform. Verizon uses SurrealDB behind a generative AI assistant for 10,000 field technicians, and PolyAI connects it to Agent Studio for customer service voice AI.

With the 3.0 release and the extension, SurrealDB is arguing that AI agents, which need to store, recall and connect context across every kind of data, turn its consolidation pitch from a developer convenience into infrastructure that larger buyers will pay for. How much of today’s usage is paid is something the company has not put on the record.

Agent memory as the selling point

The 3.0 announcement, written by Tobie Morgan Hitchcock, says the release moves SurrealDB “beyond storing data” and makes it “the foundation for agent memory and intelligence”. It adds a first stage of native file storage, so images, audio and documents sit beside structured records and can be queried in SurrealQL. It adds Surrealism, a WebAssembly extension system that runs custom logic and model execution inside the database, along with enhanced vector search and indexing for embeddings. The company says it closed more than 150 bugs for the release and that its Go and Java SDKs reached 1.0.

Chalfen’s statement in the funding announcement framed the market the same way. “Every compute era requires a new database paradigm. We are in the AI era, but most ambitious enterprise AI projects stall,” he said. The company’s funding post makes the same case in its own words: most enterprise AI projects stall, it says, because they are built on fragmented data stacks.

The post lists what it says agents need from a database: persistent memory, a unified view of structured and unstructured data, context that stays in sync as systems grow and real-time performance at scale. SurrealDB’s claim is that one engine can supply all four, where an agent project would otherwise pair a vector store with a graph database and an operational one. Its website carries comparison pages against Postgres, MongoDB and Neo4j, the systems it is pitching against.

Adoption figures ahead of revenue

The funding post gives SurrealDB’s traction as of February 2026: 2.3 million downloads, 31,000 GitHub stars and more than 1,000 forks. When it announced the first part of its Series A in June 2024, led by FirstMark and Georgian, it reported more than 25,000 GitHub stars. The commercial product is SurrealDB Cloud, a managed service with a free tier, a single-node tier priced from $0.021 an hour and a dedicated multi-node tier priced on request, according to its pricing page at the start of February.

When SurrealDB announced a $6m (£4.4m) seed round led by FirstMark in January 2023, TechCrunch reported that it was pre-revenue. The company has not published revenue since. Its funding post lists where the new money goes: enterprise-grade reliability and performance, security and governance features, cloud scalability and new regions, production support and deployment tooling, and a larger team. Each item is aimed at the buyer who signs a support contract rather than the developer who downloads the open-source engine.

Existing backers putting in more

FirstMark, which led the seed round, and Georgian, which co-led the Series A in 2024, both took part in the extension, UKTN reported. Tobie Morgan Hitchcock put the round down to usage. “This fresh investment demonstrates a growing level of excitement for our category-defining, developer-friendly database. Chalfen Ventures and Begin Capital have joined this round due to our strong momentum in real-world usage, and clear path to large-scale production,” he said.

The 3.0 release narrows SurrealDB’s message to agents, and the extension pays for the enterprise work behind it. On the public record, SurrealDB has won developers, a run of enterprise users and investors willing to add to their stakes. The $23m buys it time to show that agent memory is something businesses will pay a specialist database company for, rather than a feature they wait for their existing database suppliers to add.