Nvidia Partners Plan to Raise Substrate Capacity 2.5-Fold

Nvidia and its supply-chain partners announced multiyear capacity expansions for AI infrastructure, including a target to raise substrate capacity 2.5-fold.

At GTC Berlin, Nvidia highlighted partner use cases for its Hyperion autonomous-driving platform and presented an AI telecommunications infrastructure platform. Agile Robots announced Agile ONE SN, a humanoid reference platform built on Nvidia technology. The initiatives span autonomous driving, robotics and telecommunications infrastructure.

The substrate expansion involves Ibiden, Samsung Electro-Mechanics and Unimicron Technology Corporation. King Yuan Electronics (KYEC) is adding U.S. capacity. Siliconware Precision Industries (SPIL) plans a substantial increase in co-packaged optics capacity.

Amkor’s CoWoS-R capacity is set to triple. The announced expansions cover several parts of the AI infrastructure supply chain.

Nvidia and its partners also introduced DSX POD, a containerized design for shipping fully assembled, multirack Vera Rubin systems to sites for power-on. At a prepared site, the design aims to reduce deployment time from several months to two weeks. DSX POD is expected to be available in early 2027.