Snowflake is a NYSE-listed (NYSE: SNOW) data-cloud and data-warehouse company that has increasingly positioned itself as an enterprise AI-data platform. It does not build foundation models; instead it offers the data layer and AI tooling (Snowflake AI / Cortex AI) on which enterprises run AI deployments. In 2026 it figured in two AI-market developments: a five-year, $6 billion agreement to buy access to AWS's in-house Graviton CPUs for agentic-AI workloads, and the competition among data-platform vendors over a standardized "semantic data layer" for enterprise AI agents.
| Field | Value |
|---|---|
| Type | Data-cloud / data-warehouse company; enterprise AI-data platform |
| Founded | 2012 (San Mateo, California) |
| Founders | Benoît Dageville, Thierry Cruanes, Marcin Żukowski |
| HQ | Bozeman, Montana (legal HQ); San Mateo, California (operational) |
| Listing | NYSE: SNOW (IPO September 2020) |
Snowflake's September 2020 IPO was, at the time, the largest software IPO on record, at a valuation of roughly $33 billion.
Snapshot
Major announcements (2026)
| Date | Event | Source |
|---|---|---|
| 2026-05-28 | Lowe's confirmed publicly that a "semantic" data layer is improving the reliability of its AI agents, becoming the latest enterprise to weigh in as Microsoft, Databricks, and SAP compete over the standardized business-definition layer that governs how enterprise AI agents reason over corporate data. (Source: theinformation.com) | The Information |
| 2026-05-27 | Five-year, $6B deal to buy access to AWS Graviton CPUs for agentic-AI workloads. Snowflake shares rose as much as 35% after hours. AWS described Snowflake as joining Apple and Meta among AWS's largest CPU customers and attributed sustained CPU demand to AI agents that require many processors to orchestrate tasks. (Source: wsj.com) | WSJ |
Quantitative coverage is incomplete: this page does not yet carry revenue, customer-count, or market-cap rows. A foundational ingest of the most recent 10-K and the Q1 FY27 earnings release would be needed to add those.
Position in the AI ecosystem
Snowflake is a data infrastructure company rather than an AI lab, and it has positioned itself as the data substrate underneath enterprise AI deployments. Snowflake AI / Cortex AI is its AI product line, layered on top of the data-warehouse core. Whether Snowflake's data gravity is sufficient to make it the default agent substrate for enterprises, or whether competitors such as Databricks (Lakehouse plus Mosaic AI) and Microsoft (Fabric plus Copilot Studio) prevail, is an open competitive question.
The company runs on AWS, Microsoft Azure, and Google Cloud. The May 27, 2026 AWS deal deepens, without making exclusive, the AWS relationship; it does not change Snowflake's multi-cloud posture, but it indicates that AWS's custom silicon (Graviton CPUs, alongside its Trainium and Inferentia accelerators) is being adopted for agentic workloads at scale.
AWS Graviton CPU deal
On May 27, 2026, Snowflake signed a five-year, $6 billion agreement to buy access to AWS's in-house Graviton CPUs for agentic-AI workloads (Source: wsj.com). Snowflake shares rose as much as 35% in after-hours trading. AWS described Snowflake as joining Apple and Meta among its largest CPU customers, and attributed the sustained CPU demand to AI agents, rather than training or inference, on the basis that each agent orchestrates many parallel tool calls and thus needs many processors. AWS's framing presents the contract as anchored on agent-orchestration economics rather than model-serving economics.
This deal connects to the AI infrastructure capex arc on the CPU side, as a counterweight to the GPU-dominated capex story, and to the agentic-economy arc at the compute-substrate layer.
Semantic data layer competition
A "semantic data layer" is a layer of standardized business-definition metadata that determines how reliably enterprise AI agents reason over corporate data. Snowflake, Microsoft, Databricks, and SAP are competing to control this layer, which sits between enterprise customers and the LLMs in the agentic stack. On May 28, 2026, Lowe's publicly confirmed that a semantic data layer is improving the reliability of its AI agents, one of the first large-enterprise data points in this competition (Source: theinformation.com).
The Lowe's reference reflects Snowflake's positioning of semantic models over its data lake and warehouse as a differentiator for agent reliability, locating the differentiator not in the LLM or the agent framework but in the semantic layer that supplies the agent with shared business-term definitions. Competing approaches include Databricks' MosaicML / Unity Catalog stack and Microsoft Fabric's OneLake / Semantic Models.
Relationships
- deploys-in: Retail — AI Deployment (Lowe's customer reference); broadly across Fortune 500
- partners-with: Amazon (AWS Graviton CPU $6B deal); also major Microsoft Azure and Google Cloud partner
- competes-with: Databricks, Microsoft (Fabric / Copilot), SAP, Google BigQuery (in different sub-markets)
- related: Agentic Economy, AI Infrastructure Capex, Agentic AI, Enterprise Ai, Inference Economics and Token Pricing, Amazon (cloud partner), Microsoft, Databricks
- instance-of: AI Infrastructure Capex (the CPU-side capex story)
Sources
- Wall Street Journal — "Amazon Strikes $6 Billion Deal With Snowflake for Agentic Computing Chips," 2026-05-27 (Source: wsj.com)
- The Information — "Lowe's Says 'Semantic' Data is Boosting Its AI Agents," 2026-05-28 (Source: theinformation.com)