In more good news for Amazon, Snowflake signs $6B deal with AWS for AI CPU chips
Original reporting by TechCrunch
Cloud data storage behemoth Snowflake has just inked a landmark $6 billion, five-year agreement with Amazon Web Services, a deal that rivals Snowflake’s total historical sales through the AWS Marketplace. This massive contract signals a pivotal moment, driven largely by the insatiable demand for artificial intelligence. Snowflake customers are rapidly increasing their spending on AWS to power their AI initiatives, particularly through Snowflake’s own Cortex AI tools, which leverage vast enterprise datasets for advanced analytics and automation.
The Graviton Factor
At the heart of this expansion lies Snowflake’s increased access to AWS’s homegrown ARM-based Graviton CPU chips. As AI applications transition from intensive training to widespread daily usage and automation via agents, CPU utilization skyrockets. While GPUs excel at training, CPUs manage much of the operational heavy lifting, making cost-efficient solutions like Graviton highly attractive. This strategic move highlights AWS’s broader push to deploy its proprietary chips, offering “better price-performance” and securing multi-billion-dollar commitments from major players like Meta. It’s a clear indication that cloud providers are vying for a larger slice of the burgeoning AI compute market, challenging established chip giants even as they collectively benefit from AI's unprecedented growth.
The $6 billion agreement between Snowflake and Amazon Web Services is far more than a routine cloud contract; it is a stark illustration of how rapidly artificial intelligence is reshaping the foundational economics of enterprise technology. Driven by Snowflake’s accelerating customer spend on AI services, particularly for its Cortex AI tools running on AWS, the deal underscores the immediate, tangible impact of AI on cloud infrastructure demand. More specifically, Snowflake’s interest in AWS’s Graviton chips highlights a strategic evolution in AI compute, moving beyond an exclusive focus on GPUs for training, towards robust, cost-effective CPUs crucial for deploying AI agents and handling the vast array of daily AI tasks.
The new chip frontier
This multi-billion-dollar commitment signals a broader trend where major cloud providers are aggressively investing in their own custom silicon. AWS, Google Cloud, and Microsoft Azure are developing chips like Graviton, Maia, and their tensor processing units not merely to cut costs, but to offer optimized, high-performance alternatives to traditional chip giants like Nvidia. While Nvidia remains a dominant force, these cloud-native chips represent a calculated effort to diversify the supply chain, control pricing, and deliver tailored solutions for the escalating compute needs of AI. Such deals, including AWS’s recent agreement with Meta, mark a pivotal moment in the competitive landscape, challenging established chip vendors and signaling a future where cloud providers increasingly dictate hardware innovation. Ultimately, as AI continues its pervasive integration into business operations, cloud infrastructure providers are unequivocally positioned to be primary beneficiaries, securing their critical role in powering the next generation of intelligent applications.