Snowflake committed $6 billion to Amazon Web Services over five years on May 27 — its largest infrastructure commitment to date — under a multi-year strategic collaboration agreement aimed squarely at running enterprise agentic AI workloads on AWS custom silicon. The disclosure landed alongside a Q1 FY2027 earnings beat that sent shares up roughly 37% in after-hours trading, from a $174.60 close toward $238.81.
The commitment more than doubles Snowflake's prior AWS spend pledge. The figure has stepped up from $1.2 billion at the company's 2020 IPO to $2.5 billion in 2023 to $6 billion now — a useful proxy for how compute consumption scales as data-warehouse workloads add inference and agent orchestration on top.
Graviton CPUs plus GPUs, not just accelerators
The deal centers on AWS Graviton — Amazon's custom Arm-based CPUs — combined with GPUs for AI. Under the agreement, Snowflake will run and train its generative-AI models and services using a mix of AWS GPUs and Graviton cores, and deepen integration between Snowflake Cortex AI and AWS services. Cortex already powers text-to-SQL, summarization, sentiment analysis, and entity extraction inside the Snowflake environment; the collaboration is pitched as reducing friction in wiring customer data to AI services built on AWS.
The spend feeds a fast-growing AWS business. Amazon's custom-chip line — Trainium and Graviton combined — now runs at more than $10 billion in annual revenue and is growing at triple-digit percentages year over year. Snowflake's lifetime AWS Marketplace sales have crossed $7 billion, with more than $2 billion booked in 2025 alone.
The agentic framing — and an MCP acquisition
"We are moving into the era of the agentic enterprise, where AI systems don't just answer questions, but help organizations reason over trusted data, coordinate workflows, and drive real business outcomes," said Snowflake CEO Sridhar Ramaswamy. AWS CEO Matt Garman said Snowflake's "deepened commitment to run on Graviton delivers the world-class performance, flexibility, and cost savings customers need to run data warehousing and AI workloads at scale."
Snowflake also signed a definitive agreement to acquire Natoma, an enterprise Model Context Protocol (MCP) platform, to make it easier to securely connect AI agents to the tools enterprises already use — a direct signal that MCP is becoming table stakes for production agent deployments.
What changes for builders
The headline numbers reinforced the story: product revenue rose 34% to $1.33 billion, and Snowflake raised full-year product revenue guidance to $5.84 billion (31% growth) from $5.66 billion, with operating margin guidance up to 13.5%. For teams building on Snowflake, the AWS lock-in deepens — Graviton-backed pricing and tighter Cortex/Bedrock-adjacent integration should lower the cost of co-locating data and inference, but it further concentrates the agentic stack on a single hyperscaler. It also extends the pattern of large, multi-year AWS compute commitments seen from Anthropic, OpenAI, and Meta, with Snowflake now anchoring the data-and-agents layer.
