Ramaswamy just delivered a blowout first quarter for Snowflake, which this week reported a beat across the board. The results helped vault its shares up 36% and extended five-day gains past 50%. Shares also surged after the 14-year-old company said it would pay Amazon $6 billion during the next five years for the tech giant’s popular Graviton chips, reflecting strong demand Snowflake is seeing for its services.

The positive results were much needed for Snowflake following a stock slump that has decimated many software-as-a-service businesses due to investor fears about AI replacing traditional software vendors. Snowflake is among a pack of companies anchoring themselves after launching major AI initiatives that incorporate agentic technology with the data the company handles. The strong Q1 results (revenue grew 33% year-over-year, the fastest pace in two years) validated the consumption-based pricing model the company has long had, Ramaswamy said, and showed that traditional software can transition to AI compute.  

“It’s important to understand that all software companies are not the same,” Ramaswamy told Fortune on Friday, days before Snowflake is set to host its tech summit in San Francisco. 

The difference for Snowflake, Ramaswamy said, is that it has priced its products by consumption from the getgo. “We recognize revenue only when a customer actually uses Snowflake’s capabilities,” he said. “We have to show value to make money.”

Software pricing is among the top issues vendors like Snowflake have had to figure out since the advent of agentic AI, which has put pressure on the the industry’s traditional enterprise seat-based pricing model. Ramaswamy predicted that companies reliant on seat-based income will scramble to justify their premiums as employees use AI to accomplish an immense amount of work. 

Ramaswamy became Snowflake’s chief executive in 2024, as the AI boom was taking off. Snowflake’s bet has been that the foundational “infrastructure layer” that supports and runs its user-facing products, along with its consumption model, places the company well for the long run.

About two-and-a-half years ago, Snowflake began a broad effort to put AI into its platform. It eventually developed Cortex Code, its coding agent, and Snowflake Intelligence, an agentic application. It said in its most recent earnings that Cortex Code is in use across more than 7,100 accounts, and that accounts using Snowflake Intelligence more than doubled quarter-over-quarter.

Now, the next step is what Ramaswamy calls the control plane, which he describes as a “cockpit of work” where users, instead of only querying data, are orchestrating tasks across different applications. 

“I liken it to the new browser,” Ramaswamy said of the control plane. 

Snowflake relies heavily on Amazon and is doubling down on the cloud provider because of the quality of its chip performance, Ramaswamy said. Amazon is Snowflake’s largest partner, accounting for over 70% of how it operates its business. 

Snowflake and other large software vendors have been making major pushes to prove their long-term sustainability in the age of AI. Salesforce Chief Executive Marc Benioff said Wednesday that the company has “returned record levels to our investors,” referencing its largest-ever accelerated share repurchase of $25 billion in one quarter, which happened as the company has shown some positive results for its AI product Agentforce but is still looking for more growth to rally investors.

While Salesforce and others struggle to dispel fears from the so-called SaaSpocalypse, the sentiment has become more positive over time for the most entrenched players.

Like Benioff, Ramaswamy remains optimistic, even as major labs like Anthropic test highly autonomous systems such as the startup’s much-hyped Mythos model. Ramaswamy declined to say if Snowflake had early access to Mythos, but he argued that responsible companies should be able to leverage such powerful technologies to create and run automated security scans on the software they ship.

“You have to figure out how to harness the awesome power of these coding agents and put them to work in a responsible way,” he said. “I’m also very paranoid about making sure that I actually know what it’s doing and give permissions to it.”

Ramaswamy also said that he sees a shift away from the hundreds of different “off-the-shelf” SaaS applications toward a future that may involve far fewer major applications and more bespoke, small-scale applications.

“There will be major applications that folks will continue to buy, but there will definitely be a consolidation,” he said.