AI Drove Half of Snowflake’s Quarterly Outperformance
Snowflake’s accelerating product-revenue growth is being driven in part by AI products that prompt customers to do more work on its data platform, CEO Sridhar Ramaswamy told Bloomberg Technology. He said AI accounted for roughly half of the company’s quarterly outperformance, with its free Coco coding agent designed to generate paid consumption across Snowflake rather than subscription revenue. Ramaswamy argues that Snowflake’s advantage over Databricks lies in its governed position around enterprise data, though the company must still show that delayed renewals will lift its weaker-than-expected contracted-revenue measure.
AI is becoming a consumption engine, not just a product feature
Sridhar Ramaswamy said AI contributed roughly half of Snowflake’s outperformance in the latest quarter, as customers used its AI products to do more work on the company’s data platform. Snowflake reported $1.49 billion in product revenue, up 37% year over year, while expanding operating margin by 400 basis points to 15%.
| Measure | Reported result | Year-over-year growth | Displayed estimate |
|---|---|---|---|
| Total revenue | $1.55 billion | 35% | $1.49 billion |
| Product revenue | $1.49 billion | 37% | $1.42 billion |
Ramaswamy described a feedback loop rather than a discrete AI upsell. Customers bring more data to Snowflake, use products such as Coco and Cortex to work with it, and consume more of the platform in the process. Greater use, he said, encourages customers to place more workloads on Snowflake.
About half of the beat, the outperformance, came from AI.
That pattern gave the company confidence to raise full-year guidance to 36%, Ramaswamy said. His argument is that AI is not merely an additional feature set: it can increase the volume and range of work customers conduct on Snowflake, turning product adoption into wider platform consumption.
A free agent is meant to unlock paid workloads
Coco is a coding agent for data that comes with Snowflake. Ramaswamy said he has used it to write software himself, but characterized its core use as accelerating jobs performed in Snowflake. For a data team working on the platform, he said, Coco can make work go five to 10 times faster.
The commercial logic depends on more than that initial productivity claim. Coco is included with Snowflake rather than sold through per-user subscriptions, Ramaswamy said, and operates through a consumption model. Its immediate revenue effect is token use. But he emphasized the activities that may follow: customers bring additional data into Snowflake, optimize and debug pipelines, create interactive tables for dashboards, and build Cortex agents for business users.
It’s the second and third order effect of Coco that’s most interesting for Snowflake.
Snowflake publishes benchmarks that, Ramaswamy said, show Coco outperforming off-the-shelf models and foundation-lab harnesses on Snowflake-specific work. Its advantage, in his account, is familiarity with the platform itself. That makes it easier to sell to existing data teams: the agent is intended to remove friction from work they already do, while expanding the number of tasks they can undertake.
Ramaswamy cited Indeed as an example of the broader deployment he has in mind. The job site has rolled out Coco and Cortex to employees in more than 60 countries and 28 languages, he said. The result is ready access to data, reasoning, and tools for building further on Snowflake—capabilities he said would not have been possible a year or two earlier.
The RPO result is a test of the growth story
The consumption thesis also frames the quarter’s main qualification. Remaining performance obligations, or RPO, came in below Wall Street consensus expectations. Ramaswamy urged investors not to place much weight on the number, saying larger renewals were being pushed out by a quarter into the latter part of the year.
His position was that revenue is the healthier leading indicator and that RPO should increase sharply in the fourth quarter. The tension is straightforward: Snowflake is reporting strong current product consumption and attributing about half of its outperformance to AI, while the contracted-revenue measure investors follow did not meet expectations in the same period.
Ramaswamy’s explanation rests on timing rather than a change in demand. If the delayed renewals land as expected, the RPO measure would catch up with the revenue growth he presented as evidence of the platform’s underlying momentum. Until then, the company is asking the market to evaluate the AI-led consumption loop principally through usage and revenue rather than obligations scheduled for future recognition.
Snowflake says governed enterprise data is the advantage
Sridhar Ramaswamy drew a contrast with Databricks while acknowledging that the market is large. Databricks’ heritage, he said, is serving developers through open code bases and a more sprawling data environment. Snowflake’s strength, by contrast, is a cohesive, easy-to-use, trusted data platform, with greater attention to governance and disaster recovery.
That distinction matters, Ramaswamy argued, because Snowflake is where enterprises have placed the data used to run their businesses. The company’s position in those environments gives it an opportunity to move upstream: not simply providing underlying technology, but helping customers change sales operations or optimize supply chains.
Snowflake is becoming a lot more than a tech provider that’s largely invisible.
Ramaswamy did not argue that technology has ceased to matter. His claim was that AI becomes more valuable when it can operate on data already central to enterprise operations. Snowflake’s governance, disaster-recovery posture, and existing presence around that data, he said, put it in a position to help customers pursue operational outcomes rather than merely consume infrastructure.