Google: Google's Arm Data Center CPUs Reach Milestone Before Predicted 2025 Cloud Deployment

Following the successful introduction of Arm server processors by Amazon Web Service, Google Cloud is preparing two Arm CPUs for its cloud service. Before launching the cloud service in 2025, the firm made a significant stride and will put them into trial operation this year. The cloud servers will compete with Graviton chips from Amazon Web Services, a multibillion-dollar market that might account for as much as 10% of Elastic Compute Cloud revenue in 2022. Two Arm server CPUs are being developed by Google's server chip design team, headed by longtime Intel engineering executive Uri Frank. One is based on preexisting designs from Marvell Technology, which creates its own Arm processors and is code-named, Maple. Taiwan Semiconductor Manufacturing Co. has received Maple designs for test manufacturing. 

The second quarter will see the delivery of the Cypress CPU, which is based on an internal design created by a team in Israel. Cypress and Maple both use 5nm manufacturing nodes as their foundation. By early 2025, it may make its cloud debut in mass manufacturing, which is anticipated for the second half of 2024. Google has also created its own AI chip series for YouTube, the TPU, and the Argos semiconductor for video encoding.