NVIDIA is putting money behind the manufacturing steps needed to deliver more AI hardware, including a $1.1 billion advance for new US chip-testing facilities. Its October 21, 2026 announcement at GTC Berlin brings six suppliers into a coordinated capacity push, with major production targets stretching into 2028 rather than delivering an immediate increase across the board.
The program covers four manufacturing areas: substrates, advanced packaging, semiconductor testing and optical engines. NVIDIA says the work will support its accelerated computing and AI infrastructure platform while adding resilience and redundancy to its global supply chain. The partners are Amkor Technology, IBIDEN, King Yuan Electronics (KYEC), Samsung Electro-Mechanics, Siliconware Precision Industries (SPIL) and Unimicron Technology Corporation.
Reservation payments back a three-year substrate target
NVIDIA aims to raise substrate capacity for its products to 2.5 times current levels over the next three years. Agreements with three suppliers contribute to that goal, with more agreements underway. The arrangements combine capacity from an existing supplier, a new entrant to NVIDIA’s supply chain and a Taiwan manufacturing expansion.
- IBIDEN: An agreement signed and announced earlier this year includes a multibillion-dollar capacity reservation advance from NVIDIA for the Ono Plant in Japan. The expansion targets three times the plant’s current capacity, with additional capacity expected to begin ramping in the first quarter of 2028.
- Samsung Electro-Mechanics: The company is joining NVIDIA’s substrate supply chain with immediate capacity from existing South Korean operations. The collaboration also supports upgrades and new factories in Sejong and Busan. NVIDIA projects that its allocation will represent a significant portion of the supplier’s annual package substrate capacity by 2028.
- Unimicron: NVIDIA is supporting the KF substrate manufacturing expansion in Taiwan through a capacity reservation agreement. Additional capacity is expected to start coming online in the second quarter of 2027, with full capacity targeted for mid-2028.
Testing gets an earlier production schedule
The testing buildout has earlier milestones. NVIDIA expects its total semiconductor testing capacity to double between calendar 2026 and 2027, supported by work in the United States, Taiwan and Singapore. The US portion includes the $1.1 billion capacity reservation advance to KYEC, which will support new facilities and equipment.
High-volume production at the new US testing operations is targeted for April 2027. Expanded testing operations in Singapore are expected to begin high-volume production in February 2027. Those dates describe planned production starts; the company’s doubling target covers its total testing capacity across the three countries.
Demand is extraordinary. Capacity is the constraint.
Jensen Huang, NVIDIA founder and CEO
Packaging and optical connections follow separate targets
For advanced packaging, NVIDIA and Amkor are targeting an increase of more than 300% in interposer wafer capacity in South Korea by the end of 2028. That is the specific capacity measure behind the packaging goal. Alongside Amkor’s development of new South Korean manufacturing facilities, the expansion is intended to support higher production volumes for NVIDIA’s current and next-generation platforms.
SPIL’s portion addresses connections within large computing infrastructure. The companies plan to increase Taiwan manufacturing capacity for optical engines used in co-packaged optics, which NVIDIA links to growing connectivity requirements. Their target is a 15-fold increase by mid-2027, followed by another doubling by the end of 2028.
These are supplier-specific manufacturing targets, not a forecast that NVIDIA’s finished-system output will rise by the same multiples. NVIDIA identifies reliance on third-party manufacturers and the timing of capacity expansion among the risks to its plans. The announcement commits support to several production stages, but the stated increases remain forward-looking goals.
Reader comments
Newest comments first. Replies stay oldest first.