· Deployment & production· 日本語で読む
NVIDIA Announces Expansion of Manufacturing Capacity Across AI Infrastructure Supply Chain
nvidianews.nvidia.com reported on Nov. 17, 2026, “NVIDIA Announces Expansion of Manufacturing Capacity Across AI Infrastructure Supply Chain.”
What happened
nvidianews.nvidia.com reported on Nov. 17, 2026, “NVIDIA Announces Expansion of Manufacturing Capacity Across AI Infrastructure Supply Chain.” We have not been able to verify the primary source, so please refer to the original for details.
Why it matters
NVIDIA’s announcement describes expanding AI infrastructure supply across substrates, advanced packaging, wafers, testing and system production. Because capacity expansion plans at Kinsus, Amkor and SPIL span multiple years, assessing AI accelerator supply requires looking not only at chip manufacturing but also at the timing and capacity of back-end processes.
Analysis
A notable aspect of NVIDIA’s announcement is that manufacturing capacity expansion is spread across substrates, packaging, wafers, testing and system production, rather than being a request to increase output at a single chip factory. In terms of scale, Kinsus’ substrate capacity is set to rise to more than 3.5 times its current level, the contract with Amkor is worth about $3 billion, and the expansion for SPIL extends through mid-2029. The scale and timeline vary by process. No overall amount was disclosed, so these figures cannot be treated as a single total investment. Structurally, the implication is that the supply of computing semiconductors is being built up in parallel with preparations for back-end processes such as substrates, advanced packaging and testing. The announcement also said that volume shipments of Vera Rubin NVL72 and Vera CPU would begin, using TSMC’s 5nm and 3nm processes and CoWoS. Product shipment plans are thus tied to multi-year efforts to build capacity for components and back-end processes across the same supply chain. However, the announcement alone does not show how each company’s capacity expansion will correspond to actual product volumes or utilization rates. In connection with previous coverage, an article titled Leaderobot.com publishes commentary on NVIDIA and embodied AI was published on Nov. 20, 2026. What can be confirmed is only that the article published commentary on NVIDIA and embodied AI; its specific arguments and figures cannot be added as facts about this matter. The official announcement in this case concerns the semiconductors and manufacturing capacity that support AI infrastructure, not supply volumes or allocation by application for embodied AI. This distinction is important to avoid conflating manufacturing expansion for computing infrastructure with implementation or deployment results in robotics. From a supply-chain perspective, Kinsus’ substrates, Amkor’s and SPIL’s packaging processes, TSMC’s manufacturing technologies and Foxconn’s global production network each serve different stages. NVIDIA links this collaboration to expanding supply of its platforms, but the announcement does not specify the allocation of each process to finished products or volumes by customer. Going forward, it will be important to track progress on capacity expansions planned for 2027 and beyond, when SPIL’s contracted capacity becomes operational and what results it delivers, and how plans for substrates, packaging and testing are allocated by product. Planned capacity and actual production volumes should be tracked separately.
Sources
Translated from the Japanese original: /news/nvidia-ai-infura-kyoukyuu-ami-no-seizou-nouryoku-kakudai-wo-475d81. The Japanese version is authoritative if they differ.