The integration will appear first in Trainium4, the next-generation custom AI accelerator AWS is developing. NVLink Fusion enables high-speed interconnects between different types of processors, allowing the construction of larger, more efficient AI computing clusters essential for training complex models.
Nvidia CEO Jensen Huang described the collaboration in a prepared statement: "Together, Nvidia and AWS are creating the compute fabric for the AI industrial revolution - bringing advanced AI to every company, in every country, and accelerating the world's path to intelligence."
The partnership will also give select AWS customers access to dedicated "AI Factories"—high-performance infrastructure built with combined Nvidia and AWS technology and deployed inside customers' own data centers for maximum speed and control.
On the same day, AWS launched new server instances powered by its current Trainium3 chip. Each instance contains 144 Trainium3 processors, delivering over four times the performance of the previous generation while consuming 40 percent less energy.
Dave Brown, AWS vice president for compute and machine learning services, emphasized the focus on cost-effectiveness: "We've got to prove to them that we have a product that gives them the performance that they need and get a right price point so they get that price-performance benefit. That means that they can say, 'Hey, yeah, that's the chip I want to go and use.'"
AWS also introduced updated Nova foundation models. The new Nova 2 series offers faster response times and multimodal capabilities, accepting and generating text, images, video, and speech. A specialized version called Nova Sonic supports real-time voice-to-voice interaction.
To help enterprises build customized AI solutions, AWS unveiled Nova Forge, a service that enables companies to fine-tune models using their proprietary data while preserving the broad knowledge of the base model. CEO Matt Garman noted during his keynote: "This allows you to produce a model that deeply understands your information, all without forgetting the core information that the thing has been trained on."
The announcements come as AWS continues to expand its share of the fast-growing market for AI cloud infrastructure. In its most recent quarter, AWS recorded 20 percent year-on-year revenue growth, driven largely by demand for computing resources supporting artificial intelligence workloads.
The week-long Las Vegas conference, attended by approximately 60,000 developers and technology professionals, serves as AWS’s primary platform to showcase new services and strengthen partnerships across the technology sector.
