Databricks closed a $5 billion funding round at a $190 billion valuation on Thursday, a 42% mark-up from the $134 billion price tag it carried six months ago after its previous $5 billion raise. The company said revenue crossed a $7 billion run rate in the second quarter, growing more than 80% year-over-year. Coatue, Blackstone, MGX, T. Rowe Price and Sixth Street Growth led the round.
The pace of the re-rating is the story. Two $5 billion rounds inside six months, plus $2 billion in new debt capacity taken alongside the earlier deal, puts Databricks in a small club of private companies pulling in frontier-lab-scale capital without going public. The valuation now sits ahead of Snowflake's public market cap, inverting a rivalry that used to run the other direction.
CEO Ali Ghodsi told CNBC that customer demand is "crazy," pointing to enterprise adoption of AI agents as the driver. He singled out three products: the Lakebase database unit, the Genie business agent, and the AI Gateway tool that helps customers control model usage and costs.
Key facts
- 01Databricks closed a $5B funding round at a $190B valuation, up from $134B six months earlier.
- 02Revenue run rate crossed $7B in Q2 with 80% year-over-year growth.
- 03Lakebase, the company's database for AI agents, already passed a $100M revenue run rate.
- 04Lakehouse, the data warehousing product, surpassed a $1.5B run rate.
- 05Coatue, Blackstone, MGX, T. Rowe Price and Sixth Street Growth led the round.
The product numbers back the framing. Lakebase, positioned as a database purpose-built for AI agents, has already crossed a $100 million revenue run rate. Lakehouse, the older data warehousing product that competes head-on with Snowflake, is now at a $1.5 billion run rate. Together they anchor a portfolio that has moved well beyond Databricks' original analytics roots since the company was founded in 2013.
Ghodsi also flagged a shift in how CFOs are thinking about model spend. As token costs pile up across enterprise AI deployments, buyers are getting more aggressive about routing workloads through gateways and cheaper models, including open-weight Chinese options they previously avoided.
The attitude a year or two ago, Ghodsi said, was that enterprises needed frontier proprietary models and could ignore Chinese ones. That has flipped as spend has scaled. The AI Gateway sits directly in the path of that behavior change, letting customers swap models under the hood based on cost and performance.
“What has happened is that this token maxing has freaked out the CFOs.”— Ali Ghodsi, Databricks CEO
The IPO question came up on CNBC, and Ghodsi's answer was the same one a growing number of late-stage AI companies are giving: not yet. He said Databricks wants to be a public company but sees "too much distraction" in current public markets. Private capital, meanwhile, is showing up in nine and ten-figure checks with no earnings-call overhead attached.
That puts Databricks in the same holding pattern as several peers. SpaceX went public earlier this year and has traded with volatility since. Anthropic and OpenAI have both confidentially filed to go public and could debut as soon as this year, but neither has priced. Databricks landed at No. 3 on CNBC's 2026 Disruptor 50 list, behind Anthropic at the top.
The counterweight to the numbers is concentration risk on the demand side. Databricks' growth is riding an enterprise agent boom that is still early, still expensive on a per-token basis, and still producing ROI figures that CFOs are actively scrutinizing. Ghodsi's own "token maxing has freaked out the CFOs" line cuts both ways: it drives demand for the AI Gateway, but it also signals that some agent budgets could contract if the payoff doesn't land.
The bigger read is that infrastructure companies serving the AI build-out are being valued at multiples that used to be reserved for the model labs themselves. A $190 billion price on $7 billion of revenue is a roughly 27x multiple, richer than most public software comps and predicated on the view that agent workloads will keep compounding. Databricks is betting the private markets will fund that thesis for another year or two before it has to defend the number to public shareholders.
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.