SNOW — Snowflake
TL;DR
Snowflake is the canonical example of “owning the meter” — its query-compute consumption model made revenue a direct derivative of cloud infrastructure growth. The AI-era question: can SNOW get into the token path the way it got into the query path? The bear case (Ridire Research, Aug 2025 short): FY26 product revenue guide $4.325B, NRR 124% vs. prior highs, EV/S ~17x — priced for perfection with NRR softening. The bull case: Snowflake Cortex (AI/LLM layer on Snowflake data) and Snowpark position it in the path of AI inference workloads on enterprise data.12
Business
Cloud data warehousing and analytics platform. Revenue model: consumption-based, charged on “credits” per query or compute operation — metered and aligned directly with data workload growth. Key products: Snowflake Data Cloud (core warehouse), Snowpark (ML/app development inside Snowflake), Snowflake Cortex (AI/LLM capabilities on Snowflake data), and Marketplace (data sharing). Listed on NYSE.2
Thesis
- Consumption model coupled to enterprise data growth. Snowflake’s revenue scales automatically with enterprise data workloads — no new seats required. Clouded Judgement: “Snowflake monetizes query compute — every new query, every new dataset, every new workload meant more revenue without having to sell a single new seat.” In the AI era, if enterprise AI workloads run on Snowflake data, inference tokens consuming Snowflake data translate to Snowflake credits.2
- Cortex and Snowpark — getting into the token path. Snowflake Cortex adds LLM/AI capabilities natively inside Snowflake, allowing customers to run inference on their data without extracting it. If enterprises run AI agents on Snowflake data via Cortex, Snowflake captures both storage and inference credits — expanding TAM from query compute to AI compute.2
- Enterprise data gravity. Petabytes of enterprise data already in Snowflake creates high switching costs. AI agents need data context; Snowflake’s position as the data store gives it pole position to run agents where the data lives.
Risks
- NRR softening signals deceleration. Ridire Research (Aug 2025): NRR at 124% vs. prior highs — expansion within existing accounts slowing. At EV/S ~17x, the multiple prices in a re-acceleration that isn’t yet visible in NRR trends.1
- Databricks competitive pressure. Databricks competes directly on data + AI workloads with stronger ML/training capabilities and open-source ecosystem. If enterprises consolidate on Databricks Lakehouse for both data engineering and AI, Snowflake loses workloads.
- Token path may bypass Snowflake. If AI inference happens primarily inside model providers or hyperscalers, and Snowflake is just a data source rather than an inference executor, it may not capture token-path value the way it captured query-compute value. The primitive shift from compute to tokens doesn’t automatically favor Snowflake.2
- Hyperscaler cannibalization. BigQuery, Redshift, and Synapse all offer competitive warehousing with deeper integration into cloud-native AI services.
Recent catalysts
- 2026-03-06 — Clouded Judgement: Snowflake used as canonical “own the meter” cloud-era success; question posed whether SNOW can repeat in the AI/token era via Cortex.2
- 2025-08-08 — Ridire Research: SHORT SNOW — FY26 product revenue guide $4.325B, NRR 124% (declining from peaks), EV/S ~17x; thesis “priced for perfection.”1
Second-order reads
- 2026-03-06 — Clouded Judgement, Get in the Token Path — if SNOW gets into the AI token path via Cortex, premium multiple is justified; if it stays as a data substrate, compresses toward data warehouse multiples → Cortex traction is the key variable.
Valuation & positioning
Ridire Research (Aug 2025) flagged EV/S ~17x with NRR 124% as “priced for perfection.” Clouded Judgement’s framework suggests SNOW needs to prove it’s in the AI token path to sustain premium multiples. The Cortex strategy is the key observable.12
Sources
Related
CRM — Salesforce; both competing to be enterprise AI data layer; CRM owns customer data, SNOW owns query compute