From Humanoid Robots to Telecom Trials, 12 Major Announcements at NVIDIA GTC Berlin 2026

NVIDIA's announced agreements with suppliers such as IBIDEN, Samsung Electro-Mechanics, Unimicron Technology, Amkor, and King Yuan Electronics directly address critical manufacturing constraints in the AI infrastructure buildout. These expansions, including a planned 2.5x increase in substrate capacity over three years and a 300% rise in advanced packaging capacity by the end of 2028, aim to alleviate bottlenecks across key semiconductor production stages. This proactive strategy reflects the high demand for AI-related hardware and the recognition that scaling production is as crucial as technological innovation.

The introduction of DSX POD and the DSX AI Factory platform, featuring modular AI factories and infrastructure efficiency designs, points to a strategy focused on rapid deployment and operational cost reduction. Jensen Huang's statement that "An AI factory creates value when it starts producing tokens, not when construction is complete" underscores the drive to minimize time-to-value for AI deployments. Innovations like DSX MaxLPS for power efficiency and a new cooling design for water conservation suggest an awareness of the environmental and economic pressures associated with large-scale AI operations, with estimates of significant water savings.

The successful AI-RAN trials with T-Mobile in the United States and Elisa in Europe demonstrate NVIDIA's push into integrating AI computing directly into telecom infrastructure. Running AI workloads on live commercial networks, alongside voice and mobile data, could enable more intelligent and efficient network management. This development could accelerate the adoption of AI within telecommunications, potentially leading to new services and optimized network performance by combining AI computing with radio access network infrastructure.