NVIDIA Targets 15x CPO Optical Engine Capacity at SPIL
1 day ago / Read about 7 minute
Source:Converge Digest
NVIDIA targets a 15x increase in SPIL CPO optical-engine capacity by mid-2027 and advances $1.1B to KYEC for U.S. semiconductor testing expansion.

NVIDIA announced manufacturing capacity agreements with semiconductor supply chain partners, including a planned 15-fold increase in co-packaged optics (CPO) optical-engine capacity at Siliconware Precision Industries (SPIL) by mid-2027. The initiative addresses substrates, advanced packaging, semiconductor testing and optical interconnects supporting expanding AI infrastructure.

SPIL plans to increase CPO optical-engine capacity in Taiwan by 15x by mid-2027, followed by another doubling by the end of 2028. Amkor plans to expand interposer wafer capacity in South Korea by more than 300% by the end of 2028. NVIDIA expects total semiconductor testing capacity to double between 2026 and 2027, including a $1.1 billion capacity reservation advance to KYEC supporting U.S. facilities and equipment, with high-volume production targeted for April 2027.

NVIDIA is also targeting a 2.5x increase in substrate capacity over three years. Agreements include a multibillion-dollar capacity reservation advance supporting IBIDEN’s Ono Plant in Japan, where capacity is expected to reach three times current levels, with additional production ramping in the first quarter of 2028. Samsung Electro-Mechanics is joining the substrate supply chain, while Unimicron’s Taiwan expansion is expected to contribute capacity beginning in the second quarter of 2027.

NVIDIA Supply Chain Capacity Targets
SPIL / CPO15x by mid-2027; another 2x by end-2028
Amkor / interposers>300% increase by end-2028
KYEC / U.S. test$1.1B capacity reservation advance; April 2027 production target
Substrates2.5x aggregate target over three years

Analysis

The SPIL expansion establishes a quantified manufacturing target for CPO optical engines. CPO places optical engines close to switching silicon, shortening high-speed electrical connections. Scaling the architecture requires optical assembly, packaging, testing, integration and adequate production yield; the disclosed capacity multipliers do not specify absolute unit volumes or achieved yields.

The agreements show how optics, advanced packaging, substrates and testing must expand together for next-generation AI infrastructure. NVIDIA’s capacity reservation advances seek to secure future manufacturing resources. These are supplier capacity targets rather than confirmed shipment volumes or production commitments allocated exclusively to Vera Rubin.

NVIDIA | Company Profile
HeadquartersSanta Clara, California
CEOJensen Huang
TechnologyAccelerated computing, AI networking, GPUs, CPO, advanced packaging
Websitenvidia.com

Source: NVIDIA Newsroom, October 21, 2026.