Snowflake's $6B AWS Deal Signals a Shift to Custom AI Silicon
The $6 Billion AI Infrastructure Bet
Cloud data storage giant Snowflake has signed a new $6 billion five-year agreement with Amazon Web Services, announced on Wednesday. This contract is significant because it represents nearly the total revenue Snowflake has generated from AWS Marketplace since its founding in 2012. Snowflake reports that its customers are accelerating their spending on AWS, doubling to $2 billion in 2025 alone.
- Deal Value: $6 billion over five years
- Market Context: Close to all historical revenue from AWS Marketplace
- Annual Spend: $2 billion in 2025
Graviton vs. Nvidia: The Price-Performance Play
The driving force behind this massive contract is the shift from AI training to daily usage and automation via agents, which requires significant CPU power. Amazon CEO Andy Jassy recently boasted that AWS's homegrown chips offer "better price-performance" than Nvidia's offerings. By utilizing Graviton, AWS passes these savings to customers, making it a more affordable option for large-scale deployments.
Cloud Giants Erode Nvidia's Monopoly
This deal is part of a broader trend where cloud providers are deploying custom silicon to capture the AI value chain. Last month, AWS signed a deal to provide millions of Graviton chips to Meta. Meanwhile, Google and Microsoft have been developing their own AI chips for years. This competitive pressure is forcing Nvidia to innovate rapidly, with CEO Jensen Huang recently launching a new $200 billion market opportunity for his company's Vera chip.
The Future of AI Compute is Custom Silicon
As AI models become more complex and ubiquitous, the reliance on standard hardware is diminishing. The success of the Snowflake-AWS deal suggests that enterprises are increasingly willing to migrate away from Nvidia-centric architectures to leverage the cost-efficiency and performance of cloud-native chips.