Nvidia announced a series of multi-billion-dollar agreements with manufacturing partners on October 21 at GTC Berlin. The deals are meant to expand production capacity for parts of its AI accelerator supply chain, where Nvidia says demand is running ahead of available capacity. The company says the agreements will reserve production for Nvidia and add resilience and backup options as it builds out its AI systems.
The commitments and production plans extend through 2028. They cover substrates, advanced packaging, semiconductor testing and optical engines, which are used in co-packaged optics. Nvidia CEO Jensen Huang said the aim is to help suppliers scale production around the world. The company did not disclose individual deal values in the details provided.
What capacity is Nvidia reserving?
For substrates, Nvidia signed agreements with Ibiden, Unimicron and Samsung Electro-Mechanics. Tom’s Hardware describes them as three of the world’s five largest substrate manufacturers. Nvidia says it plans to secure 2.5 times as much dedicated substrate capacity over the next three years. The announcement does not give a starting capacity figure, so the target cannot be translated into a specific number of units.
Nvidia also signed an agreement with SPIL, or Siliconware Precision Industries, to expand manufacturing capacity for optical engines used in co-packaged optics. The company says that capacity is set to grow 15-fold by mid-2027. The announcement groups this work with other supply-chain expansion plans, but the figures disclosed here are specific to substrates and SPIL’s optical engines.
Why the supply chain matters
These are commitments to build and reserve manufacturing capacity, not a disclosed count of AI accelerators that Nvidia can ship. The deals address several stages in making its systems, rather than only the processors themselves. Nvidia says bottlenecks are limiting current and future AI infrastructure expansion, but it has not given a timeline for how much the agreements will ease those constraints.
Why it matters
For organizations planning AI infrastructure, the deals could make more component capacity available to Nvidia, but the announcement does not say how many accelerators will result or when the broader supply constraints will ease. For everyday buyers, no direct change to chip availability or prices is established.
