Snowflake CEO Sridhar Ramaswamy reports 37% revenue growth and says AI is industrializing software at every layer of the enterprise

Snowflake posted $1.49 billion in revenue for the quarter, up 37% year-on-year, and Sridhar Ramaswamy is bullish on where the growth is coming from. The company's AI products — Coco, a builder tool for data engineers, and Cowork, an enterprise AI assistant — are seeing broad adoption, with Cowork deployments now reaching thousands of users within individual companies.

“We had a pretty amazing quarter — $1,490,000,000, 37% year on year. Our AI products, Coco and CoWork, getting really broad adoption. We are seeing deployments of Cowork go to thousands of users within companies. For the majority of the company, the biggest mindset shift they need is: scale is not just about people any longer. Cocoa, for example, it's a breakout success, but for much of its existence, it never had more than five people.”

Org structure in the AI era

Ramaswamy's clearest internal bet is that headcount and scale are decoupling. For most functions, he argues the leverage now comes from agentic AI rather than adding people, which means the cultural shift required is as significant as the technical one. The exception is customer-facing roles: he expects account executive headcount to keep growing. Engineering, by contrast, is where he sees the most dramatic leverage available.

Coco's trajectory makes the point concretely. The product is a breakout success and never had more than five people during most of its development. Ramaswamy is deliberately giving small teams space to experiment rather than staffing up, on the basis that the magic-in-a-bottle phase requires room to move, not resources.

AI spend and model routing

On internal token budgets, Ramaswamy is relaxed. The company optimizes continuously — default model selection, task graphs that route simpler problems to cheaper models, open-source experimentation — but he doesn't treat token cost as a top priority. Support and SRE teams run through tens of thousands of alerts daily and spend heavily on tokens, which he regards as justified because the leverage is so high.

On model routing as a standalone market, he's cooler. Snowflake has an AI Gateway product, and is adding governed MCP tool access and experimenting with agent trajectory security, but he sees the higher-value prize as getting Coco in front of every data engineer and Cowork in front of large fractions of enterprise workforces — not capturing a slice of token flow.

M&A lens

The acquisition framework is simple: does this accelerate Snowflake's position as the AI and data platform for enterprises? Ramaswamy flags the Natoma acquisition as a clean example of that logic, arguing MCP is increasingly critical for providing real-time context from tools like Slack and email directly inside Snowflake's harness.

The competitive pressure is real — hyperscalers and foundation labs are both pushing into the space — but Ramaswamy's answer is to stay close to where enterprise data already lives and build the workflow layer on top of it.