Updates: Snowflake 3Q25 - Rethinking The AI Stack And SNOW's Role (Pt.2)

Summary

• SNOW's adoption of Iceberg and Polaris Catalog mitigates revenue loss fears, positioning it as a leader in open data formats and governance.
• Investor sentiment has shifted positively, driven by SNOW's strategic moves and Ramaswamy's leadership, despite initial concerns about his appointment.
• SNOW's valuation remains attractive compared to peers like NOW, with potential for higher FCF margins and significant growth from new product launches.
• Caution is advised due to potential near-term cost pressures from scaling S&M headcount and acquiring top-tier AI talent.

Product Fundamentals

Iceberg & Storage War

In addition to two acquisitions, SNOW has made significant changes to its product fundamentals. The most notable of these relates to addressing potential revenue losses caused by Iceberg. Previously, investors were cautioned about a possible 11% revenue hit due to customers migrating from SNOW’s proprietary data format to the open Iceberg format, which could be controlled by Databricks following its costly multi-billion dollar acquisition of Tabular, the open-core company behind Iceberg. However, in our last update, we believed this concern was overstated, and we still maintain that view for several reasons:

• SNOW remains the leading CDW engine: Regardless of the data format in use, SNOW continues to be the most attractive engine for compute operations.
• Standardizing on Iceberg benefits SNOW: If Iceberg becomes the standardized format, it’s a win for SNOW, as its superior performance-to-cost ratio will allow it to query a broader range of data without requiring customers to ingest it into SNOW’s proprietary format.
• There’s no truly “independent” storage and catalog: Ultimately, data must be optimized for a specific compute engine. This engine becomes the "first-class citizen," offering the best compute efficiency because the catalog and data storage are optimized for it. While other engines can access the data, they won’t achieve the same level of compute efficiency.
• SNOW-managed Iceberg is the preferred option: Even if customers choose Iceberg, they are likely to opt for SNOW-managed Iceberg. SNOW provides the best and most mature data governance, security, and data-sharing ecosystem. Additionally, to allow SNOW to write to the data, it must be managed by SNOW. Currently, SNOW doesn’t support write operations for third-party-managed Iceberg catalogs, and even if it did, it would be far less efficient.

After observing two quarters of minimal churn in storage revenue, with most customers adopting Iceberg opting for SNOW-managed Iceberg, SNOW's management has reassured investors that they no longer expect to lose the anticipated 11% of storage revenue. Even if that revenue does decline, it will be offset by the growth from new major products that are set to go GA. This marks a complete shift in investor sentiment.

Suddenly, the FUD (fear, uncertainty, doubt) that had been clouding SNOW's prospects has dissipated. Investors now view SNOW as firmly back in BoB status, where concerns over storage revenue losses are off the table, and SNOW’s control over its proprietary format remains intact. This reversal of expectations leads investors to embrace management’s narrative that, in fact, embracing Iceberg will drive even more revenue for SNOW. The reasoning is that SNOW’s compute engine can now query 100x more data, and as more open-format-focused customers adopt SNOW, the company will also benefit from monetizing its open-core business in the catalog (i.e., for metadata management), particularly with the Apache Polaris project.

Ironically, despite this shift in investor perception, SNOW’s fundamental position hasn’t changed significantly. Instead, investor sentiment has swung back and forth quickly and excessively, driven by the evolving narrative and management’s reassurances.

Polaris

SNOW has contributed the Polaris Catalog to the Apache Foundation, significantly reducing the risk that Databricks could leverage Tabular's Iceberg control to impose its own catalog standard on the Iceberg format. If SNOW had not made Polaris open source, Databricks could have capitalized on Tabular’s involvement with Apache Iceberg to create its own proprietary catalog, tightly integrated with Iceberg. This would have allowed Databricks to position its catalog as the default or preferred option for Iceberg users, effectively steering them toward its ecosystem. By controlling a proprietary catalog, Databricks could have imposed vendor lock-in, making it difficult for users to adopt other platforms while leveraging Iceberg's format. Open-sourcing Polaris prevents this by ensuring a widely accessible, neutral catalog alternative.