News

Major Cloud Providers, Including Amazon and Google, Compete with In-House AI Semiconductors

Wednesday, December 3, 2025 at 09:26 AM

Amazon and Google are actively utilizing their own proprietary AI semiconductor products as a strategy to compete and acquire new customers in the cloud market.

Context

Major cloud providers Amazon and Google are accelerating their in-house AI semiconductor development to reduce dependency on Nvidia and capture the booming AI workload market. At its re:Invent conference in December 2024, Amazon announced the general availability of its Trainium2 chips, with the next-generation Trainium3 slated for late 2025. This follows Google's May 2024 announcement of its sixth-generation Trillium TPU, which is now becoming available to cloud customers. The strategic goal is to offer better performance at a lower cost. Amazon claims its Trainium chips can provide up to 50% lower training costs compared to equivalent GPUs. Meanwhile, Google states its Trillium TPU delivers a 4.7x increase in peak compute performance per chip and is over 67% more energy-efficient than its predecessor. High-profile adoption, such as Anthropic committing to AWS Trainium and Amazon itself deploying over 80,000 custom chips for its internal services, signals a significant market shift away from a single-supplier hardware ecosystem.

Related Companies

Google
Google
GOOGL
US
Amazon
Amazon
AMZN
US