AWS announced on Tuesday that it will integrate key Nvidia technology into future generations of its artificial intelligence chips, a move designed to strengthen its appeal to major AI customers.
Amazon Web Services said its upcoming Trainium4 chip will incorporate Nvidia’s “NVLink Fusion,” a high-speed interconnect technology that links different types of chips. Although AWS did not share a release date, NVLink is widely considered one of Nvidia’s most valuable innovations.
The news was revealed during AWS’s annual cloud conference in Las Vegas, which draws roughly 60,000 attendees. Nvidia has been working to expand adoption of NVLink across the industry, with Intel, Qualcomm, and now AWS joining the initiative.
By adopting the technology, AWS will be able to build larger and faster AI server clusters—critical for training massive AI models that rely on thousands of interconnected machines. As part of the partnership, AWS customers will also gain access to “AI Factories,” dedicated high-performance AI infrastructure hosted within their own data centers.
“Nvidia and AWS are creating the compute fabric for the AI industrial revolution—bringing advanced AI to every company and accelerating the world’s path to intelligence,” Nvidia CEO Jensen Huang said.
Amazon also announced new servers powered by its Trainium3 chips, available immediately. Each server contains 144 chips and delivers more than four times the performance of the previous generation while using 40% less power, according to Dave Brown, AWS vice president of compute and machine learning. While he did not provide specific performance metrics, Brown said AWS intends to compete aggressively on price.
“We need to show customers that we can deliver the performance they want at the right price so they get the best value,” Brown said.
Additionally, Amazon introduced updated versions of its Nova AI models. The Nova 2 family is faster and more capable, offering multimodal responses across text, images, video, and speech. A variant named Sonic can generate speech directly from spoken prompts, with what AWS CEO Matt Garman described as “human-like” output.
Despite facing strong competition from OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini, Amazon reported a 20% revenue increase in AWS last quarter, driven heavily by cloud and AI infrastructure demand.
At the conference, Amazon also debuted Nova Forge, a tool that allows organizations to build custom AI models using their own internal data. “It lets you produce a model that deeply understands your information, without losing the core knowledge it was originally trained on,” Garman said.
Source: Reuters Edited by Bernie
