Amazon Web Services has signed cloud data outfit Snowflake as its latest chip customer, as AI agents keep chewing through compute.
According to the Wall Street Journal Snowflake plans to spend $6 billion across five years for access to Amazon’s Graviton chips inside AWS data centres.
Graviton, launched by AWS in 2018, is a central processing unit, which is the general-purpose silicon brain behind plenty of modern computing. That means everything from phones and laptops to car computers, data centre servers and AI systems.
The deal makes Snowflake one of AWS’s biggest customers for CPU-based computing, which is handy for Amazon’s homegrown silicon ambitions.
Other large Graviton buyers include Meta Platforms and the Fruity Cargo Cult Apple, although Job’s Mob was only mentioned as part of AWS’s existing customer list. AWS sells custom chips for training and running AI models, but CPUs are getting more attention as agentic AI workloads multiply.
Snowflake shares jumped as much as 35 per cent after the market closed, because investors apparently still like hearing “AI” and “chips” in the same sentence. Snowflake was founded on AWS’s platform in 2015 and has deepened its relationship with AWS ever since.
The company now has nearly 14,000 customers, including rewards and advertising startup Fetch and AI analytics platform Hex.
The rise in AI bots has boosted demand for CPUs because agents need stacks of processors to arrange computing tasks in the right order.
Chip firms selling CPUs, including Intel, AMD and Arm have seen shares rise and sales improve in recent months.
Chipzilla has benefited from the renewed CPU interest, after years of being treated like the bloke still selling fax machines at a cloud conference. AMD and Arm have been getting the same warm breeze from agentic AI, as the market starts to look beyond graphics processors alone.
The wider message is that AI infrastructure is not just about Nvidia GPUs, however much Wall Street’s AI bingo machine may pretend otherwise.
