Snowflake's $6 Billion AWS Investment Sparks In-House CPU Battle in AI Computing War

On May 27, cloud data storage leader Snowflake announced a five-year strategic partnership worth 6 billion USD with AWS (Amazon Web Services). This deal goes beyond a long-term infrastructure agreement, representing a deep integration around AI computing architecture between the two tech giants.

One, Behind the Staggering Numbers: A Contract Worth 14 Years of Past Revenue

The 6 billion USD deal is striking. Since its founding in 2012, Snowflake has sold roughly 7 billion USD in cloud services through AWS Marketplace. That means this new contract equals nearly 85% of all the revenue Snowflake has ever generated on AWS over the past 14 years.

This surge is fueled by rising enterprise spending on AI: In 2025 alone, Snowflake customers doubled their cloud consumption on AWS, reaching 2 billion USD.

Two, Core Drivers: AI Shifts from "Training" to "Automation"

The engine behind this growth is Snowflake's key AI tool, Cortex AI. As a central hub for enterprise data, Snowflake lets users query databases or generate analytical reports using natural language through Cortex.

But as AI applications move beyond simple model training into everyday use and AI-driven automation, computing demands have shifted fundamentally:

GPU handles training and inference

CPU manages large-scale agent logic and supporting tasks

As enterprise AI scales, CPU workloads grow exponentially, driving demand for high-performance, cost-efficient processors.

Three, Strategic Shift: How Graviton Chips Are Reshaping Cloud Competition

A key part of this deal is that Snowflake will gain broader access to AWS's custom ARM-based Graviton processors.

Strong Price-to-Performance: Amazon CEO Andy Jassy noted that their in-house chips deliver better value than general-purpose alternatives. By deploying Graviton chips at scale, AWS cuts its own costs while offering competitive pricing to attract major clients like Snowflake.

Securing AI Computing Footing: AWS had already supplied millions of Graviton chips to Meta. The choices of Meta and Snowflake show that cloud giants are breaking into NVIDIA's dominated computing market with custom CPUs.

Four, Industry Signals: The "CPU War" Between NVIDIA and Cloud Giants

This trend puts real pressure on GPU leader NVIDIA. Although CEO Jensen Huang recently introduced a dedicated AI CPU called "Vera" and claimed 20 billion USD in orders to defend its position, cloud providers are already pushing back:

Google has long developed its own AI chips (TPUs).

Microsoft launched its Maia AI chip earlier this year.

AWS is rapidly winning large cloud customers with Graviton through competitive pricing.

Industry Deep Analysis

The 6 billion USD alliance between Snowflake and AWS signals where competition in the AI era is heading: Data processing is no longer just about who has the most powerful GPUs, but also about who can handle large-scale reasoning and automation at lower cost.

For Snowflake, teaming up with AWS Graviton helps enterprises cut daily AI operating costs. For AWS, landing a major client like Snowflake accelerates the maturity of its custom chip ecosystem. The outcome of this battle will determine who becomes the computing foundation for the next wave of AI.