Snowflake to Burn $6B on AWS Graviton CPUs and AI Accelerators

Why It Matters

By anchoring its AI stack to AWS’s low‑cost, high‑density Graviton processors, Snowflake aims to lower compute expenses while expanding its AI‑driven revenue streams, a key growth lever for the data‑warehousing market.

Key Takeaways

• Snowflake commits $6 B to AWS Graviton CPUs and AI accelerators.
• Investment spreads over five years, averaging $1.2 B annually.
• Move targets faster AI on governed data, boosting revenue potential.
• AWS marketplace sales from Snowflake topped $7 B, $2 B in 2025.
• Shares jumped >30% after the announcement, indicating investor confidence.

Pulse Analysis

Snowflake’s strategic pivot to Amazon’s Graviton Arm processors reflects a broader industry shift toward specialized CPUs for AI workloads. While GPUs dominate model training, the surrounding data‑preparation steps—SQL queries, Python scripts, and data transformations—still rely heavily on CPU performance. Graviton’s 192‑core V3 design, paired with high‑bandwidth memory, promises lower latency and better price‑performance than traditional x86 chips, making it an attractive foundation for Snowflake’s Cortex AI platform, which converts natural language into actionable queries.

The $6 billion spend, roughly $1.2 billion per year, signals Snowflake’s confidence that AI‑enabled services will drive substantial incremental revenue. By hosting both GPU‑based model training and Graviton‑powered inference within AWS, Snowflake can offer customers a seamless, governed environment for generative‑AI applications, from sentiment analysis to automated reporting. The move also leverages Snowflake’s $7 billion lifetime AWS Marketplace sales—$2 billion recorded in 2025—demonstrating a proven revenue pipeline that can justify the hefty infrastructure outlay. Investor reaction was immediate, with shares rallying over 30% after hours, underscoring market optimism about the AI‑centric growth trajectory.

Snowflake is not alone in betting on Graviton. Meta disclosed plans to deploy tens of millions of Graviton 5 cores, positioning the chip as a bridge to future Arm‑based AI silicon. This convergence of cloud providers and AI‑focused enterprises accelerates the demand for high‑density, low‑cost compute, pressuring rivals like Google Cloud and Microsoft Azure to expand their own custom silicon offerings. For Snowflake, the partnership deepens its reliance on AWS, but it also locks in a scalable, cost‑effective platform that could become a differentiator as AI integration becomes a baseline expectation for data‑warehousing customers.