Snowflake Secures $6B AWS Deal for Custom Graviton AI Chips

At a glance:

• Snowflake signs a $6 billion five-year contract with AWS for Graviton AI chips
• Deal underscores AI-driven growth in cloud spending, with Snowflake doubling 2025 AWS revenue to $2B
• AWS's custom chips challenge Nvidia's dominance in AI infrastructure

The $6B Snowflake-AWS Partnership

The $6 billion agreement between Snowflake and Amazon Web Services (AWS) represents one of the largest commitments to custom AI hardware in recent years. Signed on Wednesday, the deal spans five years and focuses on AWS's home-grown Graviton ARM-based CPU chips. This contract is particularly significant because Snowflake has historically relied on AWS as its primary cloud provider, though it now also operates on Microsoft Azure and Google Cloud. For context, Snowflake has generated $7 billion in total revenue from AWS Marketplace since its 2012 founding—a figure this new deal nearly matches in a single contract.

Snowflake's growth on AWS has accelerated sharply in 2025. The company reported $2 billion in AWS revenue for the calendar year alone, doubling from previous periods. This surge aligns with broader trends in AI adoption, where enterprises are increasingly deploying AI tools for data management and analytics. Snowflake's AI platform, Cortex AI, plays a central role in this expansion. Launched a couple of years ago, Cortex AI enables features like natural language queries, automated report generation, and data-driven insights—tools that resonate with enterprises storing vast amounts of data in Snowflake's cloud-native architecture.

The partnership's financial terms reflect AWS's strategy to leverage its custom chip ecosystem. Graviton chips, designed specifically for AI workloads, offer cost advantages over third-party alternatives like Nvidia's GPUs. Amazon CEO Andy Jassy has emphasized that AWS's Graviton chips provide "better price-performance" than Nvidia's offerings, a claim that has attracted customers seeking to optimize AI compute costs. This pricing strategy is critical as AI workloads shift from training phases (where GPUs dominate) to operational phases (where CPUs handle tasks like agent-driven automation).

Graviton Chips: A Strategic Move for AWS

AWS's push for Graviton chips is part of a broader effort to dominate the AI infrastructure market. The company has been deploying these custom chips at scale, with recent deals including a multi-billion-dollar contract to supply Meta with millions of Graviton processors for its AI compute needs. This move is particularly impactful because Meta had previously committed $10 billion to Google Cloud, making AWS's Graviton deal a strategic win. For AWS, Graviton chips reduce reliance on Nvidia's expensive GPUs while maintaining competitive performance. Amazon's cost-conscious approach allows it to pass savings to customers, making Graviton an attractive option for price-sensitive enterprises.

The success of Graviton chips also highlights the growing importance of AI-specific hardware in cloud computing. As AI models become more complex and deployment-heavy, the demand for efficient processing units has surged. AWS's investment in Graviton reflects its recognition that AI workloads require tailored hardware solutions. Unlike general-purpose CPUs, Graviton is optimized for AI tasks, enabling faster inference and lower latency—key factors for applications like real-time analytics and automated workflows.

Nvidia's Response and the AI Chip Arms Race

While AWS's Graviton deal is a boon for Amazon, it has not gone unnoticed by Nvidia. The company's CEO, Jensen Huang, has reaffirmed Nvidia's confidence in its market position, stating that its new AI-specific CPU, Vera, represents a "brand new $200 billion market." Nvidia has already sold $20 billion worth of Vera chips, underscoring its dominance in AI hardware. However, AWS's Graviton deal and similar initiatives from Google and Microsoft signal a shifting landscape. Google has developed its own AI chips for years, and Microsoft launched its Maia AI chip in January 2025, further fragmenting the market.

Nvidia's reliance on GPUs for AI training and reasoning remains strong, but the rise of custom CPUs like Graviton and Maia threatens to erode its market share in certain segments. Huang has acknowledged the challenge, noting that cloud providers are increasingly deploying their own chips to meet AI demand. This trend is exacerbated by the high cost of Nvidia's GPUs, which can be prohibitively expensive for large-scale deployments. As a result, companies like AWS are incentivized to develop alternative solutions that balance performance and cost.

Implications for the AI and Cloud Industries

The Snowflake-AWS deal exemplifies how AI is driving consolidation and innovation in the cloud sector. Custom chips are becoming a critical differentiator for cloud providers, allowing them to offer competitive pricing while maintaining high performance. This shift is not limited to AWS; Google and Microsoft are also investing heavily in their own AI hardware, creating a fragmented but dynamic market. For enterprises, this means more choices but also increased complexity in selecting the right infrastructure.

The broader implication is that AI's growth is fueling a new era of hardware innovation. As AI moves from experimental stages to mainstream applications, the demand for specialized processing units will continue to rise. Cloud providers are at the forefront of this transformation, leveraging their scale to develop chips that cater to specific AI workloads. This trend could accelerate the adoption of AI across industries, from healthcare to finance, by making it more accessible and cost-effective.

What's Next for AI Hardware

Looking ahead, the competition for AI hardware dominance is likely to intensify. Companies will continue to invest in custom chips tailored to specific AI tasks, whether it's inference, training, or agent-driven operations. The success of Graviton and similar chips will depend on their ability to match or exceed the performance of established solutions like Nvidia's GPUs. Additionally, regulatory and ethical considerations around AI hardware may emerge, particularly as governments scrutinize the environmental and economic impacts of large-scale AI deployments.

The Snowflake-AWS partnership also sets a precedent for future deals. As more enterprises prioritize AI integration, cloud providers with robust custom hardware ecosystems will have a competitive edge. This could lead to further consolidation in the cloud market, where companies with strong AI infrastructure will attract more clients. However, the rise of multiple AI chip providers may also foster innovation, as each company seeks to carve out a unique niche in the market.

Conclusion

The $6 billion Snowflake-AWS deal underscores the transformative role of AI in shaping cloud computing. By investing in custom chips like Graviton, AWS is positioning itself as a leader in the AI hardware race. While Nvidia remains a formidable player, the growing adoption of alternative solutions highlights the evolving nature of AI infrastructure. For Snowflake, this partnership not only secures its position as a major AWS customer but also reinforces its role in enabling AI-driven data analytics. As AI continues to permeate every aspect of business and technology, the interplay between software, hardware, and cloud services will remain a critical area of focus.