Start with Data, Prove Value, Then Scale Is Snowflake’s Counter to the AI Hype

“A lot of folks went out and bought GPU capacity or model licenses without thinking about where that’s going to create value,” said Sridhar Ramaswamy, CEO of Snowflake. His perspective on enterprise AI investments has reflected a broader product philosophy: start with data, build small, prove value, and expand from there.

Ramaswamy pointed to a lightweight chat interface for Snowflake’s internal sales enablement content as an example. “It didn’t cost a lot of money to build,” he said, “but it’s getting a lot of use. That told us we were onto something worth growing.”

That same principle underpins the launch of the Cortex LLM Playground in January 2025. The no-code tool was designed for Snowflake customers to experiment with large language model (LLM) prompts through a simple chat interface. Users can run side-by-side comparisons of model outputs and directly integrate generated SQL code into their workflows. The product is deliberately lightweight and focused on usability which is a contrast to high-cost, large-scale deployments.

“We made $3.5 billion last year, growing at 30%,” Ramaswamy said. “And if we fulfill our mission to truly help enterprises mobilize data, we should be growing at faster than that clip for the next 10-ish years.” He compared the opportunity to his time at Google Search, where revenue scaled from $1.5 billion to nearly $100 billion by the time he left.

Data readiness is central to Snowflake’s strategy. “AI is only as good as its data” is a common refrain, but Ramaswamy emphasized its operational implications. At Snowflake, over 100 SaaS applications are in use. Without unifying those systems, he said, even something as basic as a dashboard becomes difficult. “And if you can’t run a dashboard,” he added, “you definitely can’t build a useful AI application.”

External tools like ChatGPT or Gemini cannot access internal enterprise systems unless those systems are integrated. This limitation underscores the role of Snowflake’s platform in making organizational data accessible across tools and workflows. “That’s why data readiness isn’t just a technical project,” Ramaswamy said. “It’s the foundation of whether your AI investments will even work.”

Snowflake’s approach also includes open data formats and accessible storage. “One is a wholesale embrace of open formats,” he explained. “Of course, one way to look at it is storage revenue that we used to get, now we won’t get because it’s sitting in open storage formats on cloud storage.”

“But a different way to look at it,” he continued, “is most large enterprises have hundreds, sometimes thousand times as much data sitting in cloud storage as they do inside Snowflake. And all of a sudden our amazing compute engine can now be used for data engineering, can now be used for data ingestion.”

The shift to open formats includes support for Apache Iceberg and Snowflake’s introduction of Apache Polaris, an open catalog format that facilitates dataset discovery.

Migration complexity has also been a focus. Advanced SQL and complex stored procedures often stall data warehouse transitions. “That’s why we are excited to announce the public preview of the SnowConvert Migration Assistant,” the company stated. This new AI-powered feature, integrated into Snowflake’s Visual Studio extension, leverages Cortex AI to help teams resolve migration errors, warnings, and issues (EWIs). By offering AI-driven suggestions and explanations for flagged code directly within SnowConvert outputs, the assistant reduces the manual effort typically required during migrations.

“Step one is making information easier to access,” Ramaswamy said. “Step two is letting models decide what to pull. Step three is chaining those components together. That’s where the orchestration begins.”

However, he cautioned against skipping foundational steps. “That would be a costly mistake.” Enterprise adoption of AI depends on these building blocks, centralized data, basic interoperability, and focused use cases that demonstrate value before scale.

Mohamed Zouari, General Manager for the Middle East, Turkey, and Africa at Snowflake, sees additional momentum. “In 2025, AI observability will move from a niche topic to a critical enterprise requirement,” he said, highlighting the need for trust, refinement, and scalability in AI systems. He also pointed to the growing importance of agentic systems, applications that make autonomous decisions within human-defined boundaries and semantic layers that allow enterprises to extract insights without reinventing data structures per application. “These advancements align with Snowflake’s mission to empower businesses with actionable intelligence,” Zouari said.

Ramaswamy views Snowflake’s dual structure as a deal-oriented sales model and a usage-driven consumption model which is essential to delivering real value. “There’s also the art of driving consumption with use cases and creating value,” he said. “Snowflake is always the yin and yang of consumption and deals.”

Drawing on lessons from his time at Google, Ramaswamy advocates for Boolean efficiency metrics which track how many team members exceed a defined baseline over simplistic averages. “There are some techniques that transfer over and other new things that I’ve had to learn,” he said, reflecting on Snowflake’s sales evolution. “But that’s life and that’s fun.”

“In a world where AI is thriving, Snowflake will thrive,” Ramaswamy said. “Because we are the layer underneath that powers this data access.”

That position, he argued, only becomes more important as AI systems evolve. “The AI promise begins and sometimes ends with what you feed it.”