Snowflake has signed a historic contract with Amazon Web Services worth $6 billion over five years. The agreement focuses on agentic artificial intelligence, a technology that allows machines to act autonomously. The big news is that they will use Amazon's Graviton processors, a cheaper alternative to Nvidia's GPUs.
Custom chips and efficiency: the shift in enterprise AI 🤖
This move marks a clear trend: large companies are looking to reduce costs by using cloud providers' own hardware. AWS's Graviton processors, based on ARM architecture, offer competitive performance at a lower price than Nvidia's expensive GPUs. For the end user, this translates into more accessible AI services in sectors like healthcare or finance, where efficiency is key. The massive adoption of enterprise AI accelerates by eliminating dependence on expensive chips.
Nvidia watches sidelong as Amazon takes the cash 💰
While Nvidia was counting its money with its GPUs, Amazon and Snowflake had a $6 billion dinner without inviting it. Now it turns out that to do artificial intelligence you don't always need the most powerful processors, just the ones that won't bankrupt the company. We'll see if the Graviton chips also work so that agentic AI doesn't ask us for a loan to buy more GPUs.