Databricks announced Thursday that a new Coatue-led funding round has valued the company at $188 billion, its fourth up-round in 18 months and a 40% jump from the $134 billion mark it set in February. The company did not disclose the size of the raise, but the round is reported at roughly $3 billion. Databricks said the money is not in hand yet and that the deal will close later this summer.
It is unusual for a company to publicize a round before the wire hits, but demand for the allocation was heavy enough that keeping the number quiet served no purpose. The valuation caps a rebranding stretch in which Databricks has repositioned from a big-data analytics vendor into an AI infrastructure provider — a shift that has systematically re-rated the company each time it goes to market.
The pace is worth laying out. In December 2024, Databricks raised $10 billion at $62 billion, a record at the time. Nine months later, in September 2025, it added $1 billion at $100 billion. Five months after that, in February 2026, it closed a $5 billion Series L at $134 billion. Five months on, it is at $188 billion. Each round has stepped the valuation up sharply enough that one observer joked on social media about "turning on alerts for when we get a Series AA."
Key facts
- 01Databricks announced a new funding round on July 17 at a $188 billion valuation, led by Coatue, with the raise reported at roughly $3 billion.
- 02The mark is up 40% from the $134 billion valuation set in February, when the company closed a $5 billion Series L.
- 03In September 2025, Databricks raised $1 billion at $100 billion; in December 2024, it raised $10 billion at $62 billion.
- 04CEO Ali Ghodsi manages AI costs across 3,000 software engineers and has publicly endorsed Z.ai's GLM 5.2 over proprietary models from Anthropic and OpenAI for coding.
- 05The round has not yet closed; Databricks said the money isn't in its hands and expects the deal to finalize later this summer.
Founded in 2013, Databricks first scaled during the big-data era, selling enterprises software to store and query cloud-resident datasets at speed. That installed base of governed enterprise data turned out to be the asset that mattered when generative AI arrived: companies wanted AI products that inherited the same security and access controls as the systems already holding their data. Databricks has since shipped Lakebase, a database built for AI agents; Unity, an AI gateway; and a meta-harness called Omnigent that coordinates multiple agents.
The company has also become one of the more visible enterprise adopters of Chinese open-weight models, a cost-control move that has emerged as one of the defining infrastructure trends of 2026. Databricks has been a particular champion of Z.ai's GLM 5.2 for coding workloads. Last week CEO Ali Ghodsi published internal benchmarking results comparing model options against the actual tasks his 3,000 software engineers run.
The finding: open-weight models can now match proprietary offerings on the hardest coding tasks at lower total cost. That is a direct swipe at the pricing tier occupied by Anthropic's Claude Code and OpenAI's Codex, and it is being made not by a research shop but by one of the largest AI-native enterprise buyers.
The benchmark also surfaced a less obvious variable. Databricks found that the harness — the agentic tool that wraps a model and manages its context — moved total cost as much as the choice of model itself. The open-source harness Pi ranked among the cheapest options without sacrificing output quality.
That framing matters because it reshapes how enterprise buyers should think about their AI stack. If harness efficiency rivals model selection in cost impact, the industry's obsession with the frontier model leaderboard captures only half the economics. It is also a convenient story for Databricks, whose platform is designed to sit above the model layer and route work.
The counterweight is that Databricks has not disclosed revenue with its recent rounds, and a $188 billion private mark implies revenue and growth expectations that will eventually need to hold up in an IPO filing or a public comparable set. The AI-halo effect is real — Jersey Mike's mentioned AI 22 times in its S-1 — but it is also indiscriminate, and private valuations set in a hot tape can compress fast when public multiples move. Coatue and the syndicate are underwriting a specific thesis about where enterprise AI spend consolidates.
For the AI infrastructure market, the read is that capital continues to flow toward the layer that owns enterprise data and workflow, not just the layer that trains models. Databricks is arguing, with its purchasing decisions as much as its pitch deck, that model providers are becoming substitutable inputs and that the durable margin sits in the orchestration and governance layer above them. If that thesis holds, the next repricing event in AI is not a frontier lab's Series H — it is the platform companies that decide which models the frontier labs get to sell into.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
