Summary: On July 17, 2026, Databricks announced a strategic round with existing investor Coatue at a $188 billion valuation. No dollar amount was announced, though The Wall Street Journal reported close to $3 billion and closing was expected later in the summer. The company says new capital will accelerate Unity AI Gateway, Genie, and Lakebase, deepening its shift into multi-model agent platforms.
A term sheet is now signed. Databricks said on July 17, 2026 that Coatue, already on its investor list, is behind a strategic round that prices the company at $188 billion. No dollar figure came with the announcement. The Wall Street Journal, though, reports a number close to $3 billion. Closing is expected later in the summer, so the cash has yet to change hands. Founded in 2013 around the Spark big-data analytics engine, the company has climbed one more valuation step in the space of eighteen months — reinforcing the role TechCrunch assigns it, AI’s favorite second act.
Four Valuation Rungs in Eighteen Months
One way to read this company is through its own list of rounds, which works almost as a year-by-year record of enterprise appetite for AI. The December 2024 raise set what was then a record on both counts: $10 billion raised, $62 billion valuation. Two rounds followed during 2025, carrying the valuation past $100 billion and on to $134 billion. The new term sheet writes $188 billion onto the same line. An unnamed venture capitalist spoke to TechCrunch: the deal is solid, this person said, and the queue for a piece of it ran so long that modesty about the price served no purpose. Online, the joke is about the alphabet of round names — one poster wrote, “Turning on alerts for when we get a Series AA.” Look at the ordering, too: news of the term sheet arrived before any money did, a sequence that holds only when the line for allocation beats caution. Enterprise budgets, meanwhile, have travelled the same road as the product — out of on-premise databases and into the lakehouse first, then off dashboards and toward agents.
A Data Lake Business Turns Into an Agent Platform
Under the valuation sits a product portfolio rebuilt around AI. Three products get the new capital, the press release states: Unity AI Gateway, the point where enterprises govern multi-model access and control cost; Genie, an AI coworker whose job is turning business data into answers and into action; and Lakebase, a serverless Postgres database designed for AI agents. Omnigent, described as a “meta-harness” for coordinating fleets of agents across tools, was demoed separately by CEO Ali Ghodsi. Customers lend the narrative its weight: over 20,000 organizations, of which roughly 70% sit in the Fortune 500, among them adidas, AT&T, Bayer, Block, Mastercard, Rivian, and Unilever. Read across the portfolio and one assumption repeats — the enterprise in mind runs many models and many agents at the same time, not a single one of each. The release says later acquisitions in AI and more research on AI come alongside product work, a signal that the data-and-agents layer still has a long way to consolidate.
Tokenmaxxing Gives Way to Valuemaxxing
The round’s meaning comes through best in Ghodsi’s own words: “Enterprises are moving from tokenmaxxing to valuemaxxing.” The buyer, put another way, is not shopping for the priciest model but for the best result each dollar buys, whoever supplies it. A benchmark of coding agents published recently by 3,000 software engineers inside Databricks — run against the multi-million-line codebase the company maintains — points the same direction. Open-weight models now take on the hardest coding work, GLM 5.2 from Z.ai more than any other, and finish it for less total spend than the proprietary models from Anthropic or OpenAI. Less expected was this: the harness wrapped around the model, Codex or Claude Code for instance, moved costs by about as much as the model pick did, while the open-source Pi harness came out among the strongest for context management. The write-up closes without hedging: picking a model settles only part of the question.
What This Means for Buyers and the Market
Three things stand out. One: putting a signature on a term sheet does not mean the capital has closed, and much could still shift before the summer close, yet the queue to participate says something about speed — how fast the AI platform layer is condensing into a few firms. Two: the primary battleground chosen here is multi-model governance, which lands squarely on the fear enterprises carry of being locked to one model vendor; what the rising valuation prices is the platform role, not a single model. Three: noise travels with frenzy, and the sandwich chain Jersey Mike’s put AI in its IPO filings 22 times. Telling real AI content from halo grows costlier each day — a warning for any business whose valuation rests on an AI story, and for the backers who fund them.