December 2, 2025

Executive Summary

Amazon announced that future Trainium4 chips will incorporate Nvidia’s NVLink Fusion technology to create high‑bandwidth links between processors. The company also introduced Trainium3 servers offering four times more compute power and 40 % less energy use, as well as AI Factories that bundle hardware, software and networking for corporate customers. These moves position AWS to compete with Nvidia while partnering with it, creating opportunities for builders of AI hardware and cloud services.

Full Article

In Las Vegas, Amazon Web Services took the wraps off its next‑generation AI hardware strategy. Future Trainium4 chips will use Nvidia’s NVLink Fusion to connect different processors at high speed, helping AI models train faster and more efficiently. This partnership illustrates how even cloud giants that develop their own chips still rely on Nvidia’s interconnect technology.

AWS also launched new Trainium3 servers that deliver four times the computing power of their predecessors while consuming 40 % less energy. The servers, available immediately, are the backbone of AWS’s new “AI Factories” – purpose‑built clusters that combine compute, storage and networking with software tools to speed up deployment of generative AI models. In addition, AWS unveiled a Nova multimodal model that can respond with text, images and speech, and a Nova Forge service to help customers customize models.

For entrepreneurs, the message is clear: the AI hardware arms race isn’t slowing down. Startups building chip design tools, cooling systems or energy‑efficient datacenter infrastructure should anticipate demand from cloud providers. At the same time, Amazon’s embrace of Nvidia’s technology suggests partnerships will be key to scaling. Integrators who package open‑source models with specialized hardware and training expertise could carve out niches as the market splinters into vertical AI platforms.

Subscribe to our newsletter to stay up to date with the latest breakthroughs in AI, Business, Technology, and Mindset.