Startups

Databricks locks in $5B at $190B valuation as AI-agent demand pushes ARR past $7B

The San Francisco data-and-AI company capped its second mega-round of 2026 after investor interest hit $15 billion, with CEO Ali Ghodsi pointing to enterprise agent spending and CFO alarm over token costs.

By Giulia Donatello · Startups & funding · 2026-08-15 · 3 min read

Photo: Unsplash / Alexandre Debiève — Close-up of a dense circuit board representing enterprise data infrastructure

Databricks closed a $5 billion strategic round at a $190 billion valuation on August 13, its second mega-round in six months and a reminder that in 2026 the private markets are the venue for enterprise AI infrastructure, not the public ones. The San Francisco data-and-AI company had planned to raise $1 billion. According to CEO Ali Ghodsi, a story in The Information during the company’s user conference triggered inbound calls that swelled to roughly $15 billion of interest before the round was rationed down.

Coatue led. Blackstone, MGX, and T. Rowe Price participated, and Sixth Street Growth, founded by former Goldman Sachs chief investment officer Alan Waxman, joined the cap table for the first time.

The financial backdrop makes the demand legible. Databricks crossed a $7 billion revenue run-rate in Q2 with more than 80% year-over-year growth, and says it’s been cash-flow positive on an adjusted basis over the last 12 months. Its Lakehouse data-warehousing business is running above $1.5 billion, growing more than 100% year over year. Lakebase, the serverless Postgres product pitched at AI agents, is already at a $100 million run rate. The acquisition of Electric, maker of the PGlite Postgres database, is the company’s fourth deal of 2026 after buying AI cybersecurity firm Panther in June and two other startups in March.

Ghodsi’s framing of the moment doubles as elite psychology. Enterprise adoption of AI agents is now large enough that inference bills are showing up on quarterly reviews. “this token maxing has freaked out the CFOs,” he told CNBC, describing the pull toward the Unity AI Gateway multi-model governance product and open-source options including Chinese models. Genie, the company’s AI coworker for business data, sits inside the same enterprise cost story.

On going public, Ghodsi was blunt: “there would be too much distraction in the public market” while OpenAI, Anthropic, and SpaceX compete for listing capital. That sentence is the whole 2026 IPO market in one line. The capital is available, it just isn’t looking for a ticker.

Sources