Nvidia announced plans with six suppliers at GTC Berlin on October 21, 2026, to expand four parts of the AI-accelerator supply chain: substrates, advanced packaging, semiconductor testing and optical engines. The company’s targets run through 2028 and describe planned capacity growth, with production milestones expected in 2027 and 2028. Nvidia’s announcement says the agreements are intended to support growth and strengthen supply-chain resilience.
What the six suppliers handle
The agreements cover different manufacturing stages, rather than a single pool of finished accelerators. IBIDEN, Samsung Electro-Mechanics and Unimicron Technology Corporation are tied to substrates. Amkor Technology is expanding advanced packaging, King Yuan Electronics (KYEC) is involved in semiconductor testing, and Siliconware Precision Industries (SPIL) is expanding optical engines for co-packaged optics.
A substrate, an interposer wafer, a testing operation and an optical engine each have their own capacity measure. A target for one stage therefore does not give a count of finished AI accelerators.
Capacity targets and milestones
Nvidia’s figures are plans for distinct stages, with different baselines and timelines. The company announced the following targets:
| Supply-chain stage | Supplier(s) | Announced target | Location and timing |
| Substrates | IBIDEN, Samsung Electro-Mechanics and Unimicron Technology Corporation | Nvidia targets 2.5× substrate capacity for its products over the next three years. | Japan, South Korea and Taiwan |
| IBIDEN substrates | IBIDEN | The Ono Plant targets three times its current capacity. | Japan; additional capacity is expected to begin ramping in Q1 2028. |
| Unimicron substrates | Unimicron Technology Corporation | The KF expansion is expected to begin adding capacity in Q2 2027 and reach full capacity by mid-2028. | Taiwan |
| Advanced packaging | Amkor Technology | Nvidia and Amkor target an increase of more than 300% in interposer-wafer capacity. | South Korea; by the end of 2028 |
| Semiconductor testing | KYEC | Nvidia expects total testing capacity to double from calendar 2026 to calendar 2027. | The United States, Taiwan and Singapore; high-volume production is targeted for February 2027 in Singapore and April 2027 at new U.S. facilities. |
| Optical engines for co-packaged optics | SPIL | SPIL targets 15× capacity by mid-2027, with a further doubling aimed for the end of 2028. | Taiwan |
The multipliers cannot be combined into one overall production figure: they refer to separate stages, from substrate supply to testing and optical engines. The announcement also identifies Samsung Electro-Mechanics as a substrate partner without assigning it a separate capacity multiplier in the targets listed here.
What Nvidia’s $1.1 billion KYEC advance supports
Nvidia describes the $1.1 billion advance to KYEC as a capacity reservation supporting new testing facilities and equipment in the United States. The U.S. facilities are targeted to begin high-volume production in April 2027. KYEC’s expanded Singapore testing operations are expected to begin high-volume production in February 2027.
