a16z co-founder Ben Horowitz: AI agents now initiate more queries than humans
Andreessen Horowitz co-founder Ben Horowitz published a lengthy article titled Databricks & the Future of the Firm around October 27, 2026, discussing how data and AI platform company Databricks, through its Lakehouse and Lakebase architecture, provides enterprise system records for AI agents and may change companies’ daily work and organizational structures. The article cites company data saying that Databricks’ annual recurring revenue run-rate has surpassed $7 billion, up more than 80% year over year; its Lakehouse data warehouse business has a run-rate of more than $1.5 billion, up more than 100% year over year; and its Lakebase database for AI agents has a run-rate of more than $100 million. The company completed a $5 billion financing round in August 2026, valuing it at $190 billion.
Horowitz said in the article that three years ago every interaction on the Databricks platform was completed by humans, whereas now the number of queries initiated by agents has surpassed that of humans, and agents start more than 17 million databases every day. This statement is consistent with the direction of the company’s earlier report: Databricks’ 2026 State of AI Agents report, based on anonymized telemetry data from more than 20,000 organizations, said that AI agents created about 80% of new databases and about 97% of test and development environments. The company acquired Neon, a serverless Postgres company, for about $1 billion in 2025 and integrated its technology into the Lakebase product, which is designed for agents to create, branch, and destroy databases quickly.
Databricks was founded in 2013 by UC Berkeley researchers including Ali Ghodsi and Matei Zaharia. It was initially based on the open-source Apache Spark technology and later launched the Lakehouse architecture, which combines the low-cost storage of data lakes with the reliability of data warehouses. The company’s products include the natural-language interface Genie and the agent development platform Agent Bricks. CEO Ali Ghodsi has publicly said that enterprises need not only AI that can talk, but agents that can remember context in business operations, give accurate answers, and carry out work without going over budget. The company remains private, and its financial data are self-disclosed run-rate metrics.
Horowitz’s analysis positions Databricks as infrastructure that supports a “system of record for agents,” arguing that as the number of agents far exceeds the number of human employees, companies need unified records of state, permissions, and evaluation so front-line staff can act directly without layered reporting. The article is an investment firm’s view and industry observation, not an official company announcement; the specific figures on agent queries and database starts currently mainly appear in the article’s citations.
Original sources:
Title: Databricks & the Future of the Firm Publisher: Ben Horowitz (a16z) Publication date: 2026-10-27 URL: https://www.a16z.news/p/databricks-and-the-future-of-the Nature of source: investment firm analysis article Verification status: main points and cited data from the article were read in this session, verification date 2026-10-27 Supported content: revenue run-rate, valuation, Lakebase/Lakehouse figures, descriptions of agent queries and database starts, historical technology path
Title: Databricks Grows >80% YoY, Surpasses $7B Revenue Run-Rate... Publisher: Databricks / Sixth Street and other reports Publication date: 2026-08-13 URL: https://sixthstreet.com/newsroom/databricks-grows-80-yoy-surpasses-7b-revenue-run-rate-scales-lakebase-genie-and-unity-ai-gateway Nature of source: company news release and media report Verification status: financing and revenue disclosure content read in this session, verification date 2026-10-27 Supported content: $7 billion run-rate, 80% growth, $1.5 billion Lakehouse, $100 million Lakebase, $190 billion valuation
ABAB AI Insight
Databricks’ financial disclosures in August 2026 showed that its annual recurring revenue run-rate had surpassed $7 billion and was still growing rapidly, while it completed $5 billion in financing at a $190 billion valuation. On the product side, both the traditional Lakehouse business and the newly launched Lakebase contributed significant run-rate figures, with the latter reaching more than $100 million about a year after the Neon acquisition. These figures are company-reported run-rate metrics rather than audited GAAP revenue, but they are consistent with the direction of multiple independent media and investor reports.
In terms of technical path, the company moved from early Spark distributed computing to Lakehouse for unified analytics and storage, and then into the operational database (OLTP) space through the Neon acquisition, forming a storage-and-compute-separated architecture tailored to AI agents. Lakebase supports rapid branching and on-demand scaling, matching the usage pattern of agents frequently creating temporary databases. The company’s report shows that agents have come to dominate new database creation, which corroborates Horowitz’s description that agent queries exceed human queries and that millions of databases are started every day, but the specific figure of 17 million currently mainly appears in that article.
Compared with peers, Databricks’ valuation and growth rate are higher than those of some publicly traded data platform companies (such as Snowflake in revenue scale and growth), but it remains a private company with limited financial transparency. Horowitz’s analysis emphasizes that when the number of agents far exceeds humans, enterprises need a “system of record” to centralize state and permissions while decentralizing execution, similar to the logic by which historical databases decentralized decision-making. This view is an industry observation, not a confirmed result of organizational change.
What this article changes is the depth of public discussion of Databricks’ role in AI agent infrastructure; what it does not change is the fact that the company remains private, its revenue data are self-reported run-rate figures, and large-scale enterprise deployment of agents still faces governance and trust challenges. The next points to watch are the actual scale of customer adoption of Lakebase, the share of revenue tied to agents, and whether the company proceeds with an IPO or further financial disclosures.
ABAB News · Cognitive Law
In August 2026, Databricks disclosed an annualized revenue run-rate of more than $7 billion and a valuation of $190 billion.
The company launched Lakebase through the acquisition of Neon and reported that AI agents have created most new databases.
Ben Horowitz’s analysis positions Databricks as infrastructure supporting enterprise agent system records.