Databricks Hits $6.9 Billion Run Rate as AI Agents Overtake Humans in Software Usage

Databricks has reached a rare stratosphere in enterprise software, reporting an annualized revenue run rate of $6.9 billion with staggering 80% year-over-year growth. The data infrastructure giant's latest financials cement its position as one of the fastest-growing software companies in history, outpacing the traditional deceleration curves that typically hit businesses operating at this massive scale.[1][7]

But beneath the top-line revenue explosion lies a fundamental shift in how enterprise software is being consumed. The primary driver of this hyper-growth is no longer just human data scientists buying more seats; it is a "swarm" of autonomous AI agents executing tasks, querying databases, and running analytics pipelines around the clock.[1]

"Focus on the new customer, which is the agent," Databricks CEO Ali Ghodsi recently noted, highlighting that the era of software built exclusively for human operators is rapidly closing. According to a recent industry report, AI agents now create 80% of all new databases and 97% of test and development environments on the Databricks platform. Just two years ago, those figures were near zero.[2][5]

This transition from human "copilots" to autonomous agents is fundamentally rewiring corporate IT. Agents interact with infrastructure differently than humans do—they spin up thousands of micro-databases to experiment, test hypotheses, and frequently make mistakes that require instant rollbacks.[5]

To accommodate this non-human user base, Databricks has rolled out specialized infrastructure, including Lakebase, a serverless Postgres database engineered specifically for the rapid, high-volume branching and rewinding that AI agents require. The strategic pivot is paying off: the company's AI-specific product line is now generating an annual revenue run rate exceeding $1.7 billion.[2][3][6]

However, the agentic boom comes with a steep learning curve for enterprise budgets. While Databricks' sales are surging, the massive compute requirements of autonomous agents are pressuring margins. Because agents can run continuously and spawn sub-agents to solve complex problems, they consume vast amounts of processing power and API tokens.[1][3]

Databricks co-founder Patrick Wendell recently revealed that some enterprise clients have seen their AI token costs spike from zero to tens of millions of dollars in a single month due to unchecked agent activity. AI compute has quickly become a top-three corporate expense for these early adopters, trailing only payroll and general IT.[3]

In response, the company has introduced the Unity AI Gateway, a governance tool designed to monitor individual agent sessions, cap spending, and shift corporate behavior from "token maxing to value maxing." The goal is to provide the guardrails necessary for companies to deploy agents safely without bankrupting their IT departments.[3]

The financial markets are aggressively pricing in this infrastructure shift. Following a $5 billion funding round in February that valued Databricks at $134 billion, the company is reportedly in discussions for a new round that could push its valuation as high as $175 billion.[4][6]

With gross margins historically hovering around 80% and the company generating positive free cash flow, Databricks is widely considered one of the most anticipated IPO candidates of the decade. However, leadership has signaled they may bypass a public listing in 2026, opting to stay private while they build out the foundational plumbing for the agent economy.[3][4][6][7]

The broader takeaway extends far beyond a single company's balance sheet. As AI moves from generating text to taking autonomous action, the most valuable companies of the next decade may not be the ones building the smartest models, but the ones providing the data, storage, and guardrails for the agents those models power.[2][5]