Brand logo
HomeAutomationSupply ChainSmart ManufacturingLogisticsEnergyGeneral
  1. Home
  2. /
  3. general
  4. /
  5. nvidia-suppliers-scale-global-ai-manufacturing

World Industrial Weekly Guide: From Ono Plant to US Test Lines, How NVIDIA Is Rewriting Supplier Capacity

A reservation signed in a boardroom reaches the shop floor years later, as tool moves, cleanroom staffing and shipment schedules. That lag is the right lens for the manufacturing plan NVIDIA assembled with its supply chain partners: substrate output is set to grow 2.5x, semiconductor testing throughput is set to double from 2026 to 2027, and nearly every milestone in between already carries a date.

The trigger is demand for accelerated computing and a full-stack AI factory platform rather than one product line. By working with Amkor Technology, IBIDEN, King Yuan Electronics (KYEC), Samsung Electro-Mechanics, Siliconware Precision Industries (SPIL) and Unimicron Technology Corporation, NVIDIA is pushing capability outward into substrates, advanced packaging, testing and optical engines at new and existing suppliers — a move aimed at resiliency and redundancy across four process layers that rarely fail alone.

Four capacity lines and the numbers behind each

Capacity layerTarget stated in the planWhere it landsDated milestone
Package substratesgrow 2.5x over the next three yearsJapan, South Korea, Taiwanadditional capacity begins ramping in the first quarter of 2028
Advanced packaging (interposer wafers)greater than 300% increaseSouth Koreaby end of 2028
Semiconductor testingtotal capacity to double from calendar 2026 to 2027US, Taiwan, SingaporeApril 2027 and February 2027 for high-volume production
Optical engines for co-packaged optics15x, then another doublingTaiwanmid-2027, then the end of 2028

Read together, the four lines describe a chain being widened at every handoff, not simply one bottleneck being relieved.

Case file: three substrate sites, three different ramp curves

IBIDEN and the Ono Plant

Under the agreement signed and announced earlier this year by IBIDEN, NVIDIA is extending a multibillion-dollar capacity reservation advance behind the Ono Plant expansion in Japan. That site is pointed at capacity reaching 3x current levels, with additional capacity expected to begin ramping in the first quarter of 2028.

Samsung Electro-Mechanics: capacity now, two build-outs later

South Korea's existing operations contribute immediate capacity into the substrate supply chain, while facility upgrades and new factories in Sejong and Busan follow. By 2028, production allocated to NVIDIA is projected to represent a significant portion of Samsung Electro-Mechanics' total annual package substrate production capacity.

Unimicron Technology Corporation: a timed ramp in Taiwan

A capacity reservation agreement backs Unimicron in its KF substrate manufacturing expansion in Taiwan. That expansion is expected to begin adding capacity in the second quarter of 2027 and to reach full capacity by mid-2028 — a two-step curve buyers can plan around.

Packaging, test and optics: the layers that decide throughput

Amkor Technology and interposer wafer supply

NVIDIA and Amkor intend to expand advanced packaging capacity in South Korea, targeting a greater than 300% increase in interposer wafer capacity by the end of 2028. The planned expansion, alongside Amkor's development of new manufacturing facilities there, supports higher production volumes for current and next-generation platforms.

Testing: the shortest path from reservation to volume

Total semiconductor testing capacity is expected to double from calendar 2026 to 2027, with buildouts in the U.S., Taiwan and Singapore providing support. The $1.1 billion capacity reservation advance to King Yuan Electronics (KYEC) funds new U.S. testing facilities and equipment, with high-volume production targeted for April 2027. Expanded testing operations in Singapore are expected to reach high-volume production in February 2027. The effort also strengthens US manufacturing.

SPIL and co-packaged optics

Siliconware Precision Industries (SPIL) is working with NVIDIA to expand capacity in Taiwan for optical engines used in co-packaged optics, addressing the connectivity requirements of large-scale computing infrastructure. The target is an increase of 15x by mid-2027, with aims for another doubling by the end of 2028.

Four misconceptions that distort the read

  • Treating a capacity reservation as finished output. Money moves long before wafers do: IBIDEN's additional capacity is expected to begin ramping only in the first quarter of 2028, and Unimicron's expansion is scheduled to reach full capacity by mid-2028.
  • Assuming "double" describes the whole test market. The figure covers NVIDIA's total semiconductor testing capacity moving from calendar 2026 to 2027.
  • Reducing the plan to one technology. Substrates, interposer-level packaging, test and optics carry separate targets — 2.5x, greater than 300%, a doubling and 15x — and each relieves a different constraint.
  • Mapping the build-out to a single country. Named sites span Japan, South Korea, Taiwan, Singapore and the US, so qualification, freight and compliance teams are tracking several environments at once.

What industrial operators should do next

  1. Anchor internal assumptions to dated ramps rather than announcement dates: February 2027 in Singapore, April 2027 in the U.S., mid-2027 for optical engines, then the end of 2028 for packaging.
  2. Identify which layer binds your own program. Interposer wafer capacity is the slowest to widen, with the greater than 300% target landing only by the end of 2028.
  3. Read reservation advances as balance-sheet signals. A multibillion-dollar commitment to one plant tells competing customers something about how that capacity will be prioritized.
  4. Re-score supplier redundancy. Test capacity spread across the US, Taiwan and Singapore is a template for how buyers should think about second sources in any long-lead component.
  5. Watch allocation disclosures. By 2028, the share of Samsung Electro-Mechanics' annual package substrate capacity allocated to NVIDIA is projected to be significant — the kind of disclosure that should reshape a substrate bid list.

FAQ

Is this a greenfield fab announcement?

No. It is a capacity-reservation program layered onto supplier sites: output from existing operations plus facility upgrades and new factories, ramping on staggered dates instead of one opening.

Which line grows fastest in relative terms?

Optical engines for co-packaged optics, at 15x by mid-2027, with aims for another doubling by the end of 2028.

Where does the added capacity sit geographically?

Substrates at IBIDEN's Ono Plant in Japan, at Samsung Electro-Mechanics sites in Sejong and Busan, and on Unimicron's KF expansion in Taiwan; advanced packaging in South Korea; testing in the U.S., Taiwan and Singapore.

What is the longest-dated commitment?

The substrate line: 2.5x over the next three years, with Unimicron reaching full capacity by mid-2028 and IBIDEN's new capacity ramping from the first quarter of 2028.

Why does this matter outside semiconductors?

Because the plan consumes equipment, cleanroom space and test throughput — inputs that feed availability, pricing and lead times for anyone deploying automation, plant electrification or edge compute hardware.

What to watch next

The most useful element here is not a percentage; it is a calendar. Each target now has a site, a partner and a quarter attached, which means slippage will surface as moved dates rather than softened language. Tracking the substrate, packaging, test and optics programs through 2028 is the practical way to see when reservation language becomes delivery performance.

Brand logo

Industrial Weekly

In-Depth Weekly Insights on Manufacturing & Supply Chain

AutomationSupply ChainSmart ManufacturingLogisticsEnergyGeneralGreen CarbonWorld

© 2026 Industrial Weekly.

Cookie Notice

We use cookies to improve your experience and analyze website traffic. Read our Privacy Policy to learn more.