Databricks Annualized Revenue Surges Past 80% to $6.9 Billion as AI Agent Costs Squeeze Margins

Databricks announced Monday that its annualized revenue has grown more than 80% year over year to $6.9 billion, up from $5.4 billion in the fiscal fourth quarter, even as the rapid adoption of AI agents is driving up costs and compressing the company's margins.

The figures were disclosed to analysts at the company's Data and AI Summit in San Francisco on Tuesday.

CEO Ali Ghodsi attributed the margin pressure directly to the mechanics of how clients are now deploying AI. "It's the consumption-based business model, agentic AI coming," Ghodsi told CNBC in an interview at the conference. "The agents are generating way more queries. We have all these agents, the agents and agent platform we have also generates revenue, so it just increases the consumption of everything all around."

Ghodsi declined to specify Databricks' current gross margin but confirmed it will move lower. As AI products grow in popularity, the company must spend more on the underlying models that power them.

Databricks now generates $1.7 billion in annual revenue from AI products alone, up from $1.4 billion as of February.

The San Francisco-based company carries a private market valuation of $134 billion, placing it ahead of publicly traded rival Snowflake, which has a market cap of roughly $83 billion. Snowflake's annualized revenue stands at approximately $5.6 billion based on its most recent quarterly results.

Despite its scale, Databricks has remained on the sidelines of the public markets. OpenAI and Anthropic have each filed for confidential offerings, while SpaceX last week completed what was described as the largest debut on record, topping a $2 trillion market cap on its first day of trading.

Ghodsi also addressed a broader shift in how enterprises are managing AI spending. Companies have moved away from what he characterized as "tokenmaxxing" — encouraging workers to maximize token usage — and toward what he called "value-maxxing," optimizing efficiency while still leveraging AI capabilities.

Large enterprises, he said, want access to frontier models for complex tasks but are increasingly turning to open-source alternatives for routine work. "They are interested in that, but not for everything, right? And for the mundane tasks, they absolutely want to curb the cost and use simple open-source models," Ghodsi said, highlighting Anthropic's Mythos as an example of a frontier model in demand.

Chinese AI models are also drawing strong interest among Databricks customers. "The customers are really demanding the choice," Ghodsi said.

The company is expanding into new verticals to sustain its growth trajectory. Databricks announced its entry into the cybersecurity market in March with the introduction of its Lakewatch software. On Tuesday, the company said it would acquire Panther, a security startup valued at $1.4 billion in 2021, and unveiled CustomerLake, a new software product aimed at managing marketing data.

With margins under pressure and a public offering still absent from the roadmap, Databricks faces a test of whether its consumption-driven model can sustain investor confidence at its current valuation as the cost structure of agentic AI continues to evolve.

Disclaimer