Nvidia Signs Multi-Billion Deals to Expand AI Chip Supply

Nvidia signed multi-billion-dollar, long-term contracts with major supply-chain partners at the GTC event in Berlin to expand manufacturing capacity for AI chips and address production bottlenecks; the programs run through 2028.

The agreements cover substrate production, advanced packaging, outsourced semiconductor assembly and test services, and integrated optical engines.

Jensen Huang said, "We are in the midst of the greatest infrastructure expansion in human history. Demand is extraordinary and production capacity is our bottleneck. Our supply-chain partners are the best manufacturers in the world. I am grateful that they are expanding capacity worldwide to scale production and help us deliver AI supercomputing systems for this new industrial revolution."

The commercial understandings split into two core areas: hardware interconnects and OSAT services for assembly and testing.

For the physical substrate layer, Nvidia agreed with Ibiden, Unimicron and Samsung Electro-Mechanics to reach a dedicated capacity 2.5 times current levels within the next three years.

In optics, Nvidia reached an agreement with SPIL to increase production capacity for integrated optical engines 15-fold by mid-2027.

Chip packaging will expand through a collaboration with Amkor and facility growth in South Korea, targeting an increase in intermediate wafer capacity of more than 300 percent by the end of 2028.

For final testing, Nvidia allocated $1.1 billion to reserve capacity at King Yuan (KYEC) to double chip test capacity by 2027.

The company says the measures are intended to guarantee sustainable capacity and throughput for its AI accelerators through 2028.