Snowflake will spend $6 billion over the next 5 years on Amazon Web Services' in-house Arm-based Graviton CPU chips, the Wall Street Journal reported on May 27.

Graviton, unveiled by AWS in 2018, is used to run data centre servers and AI systems.

Under the deal, Snowflake will become one of AWS' biggest CPU-based computing customers. Major AWS Graviton customers include Meta and Apple. AWS is also supplying customised semiconductors for AI model training and inference.

Snowflake shares jumped about 30 percent in after-hours trading after the contract was announced.

Snowflake launched in 2015 on the AWS platform and has steadily expanded its cooperation. It now has about 14,000 customers.

Snowflake's move reflects rising CPU demand as the AI agent market grows. AI agents, which autonomously perform a range of tasks on behalf of human users, need large CPU resources to coordinate multiple computing tasks in sequence.

As a result, revenue growth at CPU-related companies such as Intel, AMD and Arm is also rising. Some in the industry also say the AI sector is entering a phase in which demand is spreading from GPUs to CPUs and customised semiconductors.