On May 27, 2026, Snowflake and AWS announced a five-year, $6 billion infrastructure agreement centered on expanded use of AWS’s homegrown, ARM-based Graviton processors and its AI infrastructure. The same day, Snowflake reported first-quarter earnings, raised its full-year forecast, and watched its shares surge 36 percent in extended trading. The detail worth pausing on: the headline chip is not a GPU. As AI workloads shift from model training toward everyday inference and agent automation, cloud-native CPUs have quietly become a battleground of their own.
What $6 Billion Actually Buys
Per TechCrunch, the agreement goes well beyond raw Graviton capacity. It bundles deeper product integrations for generative and agentic AI, expanded go-to-market through the AWS Marketplace, and workload migration services meant to move businesses “from experimenting with AI projects to using them routinely.”
The numbers show how deep the relationship already runs. Snowflake has sold roughly $7 billion of services through the AWS Marketplace since its 2012 founding — meaning this single contract nearly matches fourteen years of marketplace revenue. Customer spending through AWS doubled in 2025 to $2 billion, driven by AI adoption. Snowflake’s Cortex AI stack (natural-language database queries, summary reports) has been on the market for a couple of years, and on the earnings call management singled out Cortex Code and Snowpark as growth engines for the quarter.
Why the CPU Is the Headline, Not the GPU
GPUs own training and large-scale reasoning. But once AI becomes daily software — agents orchestrating API calls, data preparation, scheduling, state management — most of that work lands on CPUs, and the volume scales with the number of agents running. Amazon CEO Andy Jassy claimed last month that the company’s homegrown chips deliver better price-performance than Nvidia’s, and AWS separately signed a deal to supply millions of Graviton chips to Meta — notable because Meta had signed a $10 billion Google Cloud deal just months earlier. Microsoft, for its part, launched its own Maia AI chip in January.
Nvidia CEO Jensen Huang pushed back last week, calling Vera, the company’s new AI-specific CPU, a “brand new” $200 billion market with $20 billion already sold after a record quarter (we analyzed Vera in depth). Read those moves together and the picture is unusual: cloud giants are using homegrown CPUs to claim more of the AI infrastructure stack, while the incumbent GPU leader attacks the cloud vendors’ CPU turf. The CPU, for once, is where the two sides collide.
The Earnings Report Was the Real Catalyst
For markets, the deal was the appetizer; the numbers were the main course. First-quarter revenue came in at $1.39 billion, beating the $1.32 billion analyst estimate. Snowflake raised its FY2027 product revenue forecast from $5.66 billion to $5.84 billion, and guided second-quarter product revenue to $1.415–1.42 billion against an LSEG consensus of $1.37 billion. “Based on a combination of strength in our core data platform business and meaningful uplift from AI capabilities,” CEO Sridhar Ramaswamy said in explaining the raise. D.A. Davidson analyst Gil Luria argued the AWS agreement “adds another element to the growth path” for Snowflake and positions it to play “an even bigger role in their customers’ transition to AI” while aligning it more closely with its largest partner.
What It Means for Developers and Platform Teams
Three practical takeaways. First, cloud-native ARM CPUs are now a default tier for AI-adjacent workloads, so cross-architecture compatibility (amd64/arm64) belongs in your CI pipeline today, not after the first migration incident. Second, major data platforms are locking in multi-year infrastructure commitments: if you build on Snowflake, expect to drift closer to AWS’s AI services over the next five years, with exit costs rising accordingly — factor that into vendor strategy now. Third, when you model the cost of agent workloads, don’t count GPUs alone. The unit economics of features like Cortex Code depend heavily on CPU price-performance, which is exactly the bet embedded in this deal. AI’s compute story is broader than the GPU shortage narrative, and the cheapest capacity wins the workloads nobody glamorizes.
Sources
- In more good news for Amazon, Snowflake signs $6B deal with AWS for AI CPU chips — TechCrunch
- Snowflake boosts forecast, signs $6 billion AWS deal as enterprise AI adoption grows — Reuters
- Snowflake boosts forecast, signs US$6 billion AWS deal — The Standard
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.