Nvidia Lines Up Six Partners for 2.5x Substrate, 15x Optics Build
TL;DR
- Nvidia advanced $1.1 billion cash to KYEC for new U.S. testing facilities, with high-volume production targeted April 2027 and Singapore coming online February 2027.
- TSMC's absence from the six-partner list confirms packaging and substrate, not wafer fabrication, are now the binding constraints in AI accelerator production.
- The Amkor 300% packaging expansion extends a prior $1.5 billion July 2026 agreement, meaning Nvidia has been pre-positioning supply chain capacity for at least six months.
Nvidia posted a list of six supplier commitments this week, each attached to a capacity multiple. In the release, the company says substrate agreements with IBIDEN, Samsung Electro-Mechanics and Unimicron Technology Corporation will 'in aggregate, drive a 2.5x increase in capacity over the next three years.' Amkor plans to 'increase its advanced packaging capacity by more than 300% by the end of 2028.' SPIL is targeting a 15x expansion of optical-engine capacity for co-packaged optics by mid-2027, 'with aims for another doubling by the end of 2028.' Total testing capacity is set to double from 2026 to 2027.
Behind the multiples sit two cash commitments. KYEC gets a $1.1 billion capacity reservation advance from Nvidia to build new U.S. testing facilities, with production targeted for April 2027. IBIDEN gets a multibillion-dollar advance to triple its Ono Plant's substrate capacity by Q1 2028.
'We are in the midst of the largest infrastructure buildout in human history,' CEO Jensen Huang stated. 'Demand is extraordinary. Capacity is the constraint.'
The geography is deliberate: substrates from Japan, South Korea and Taiwan; advanced packaging in South Korea; testing in Singapore (February 2027) and the United States; co-packaged optics out of Taiwan. Unimicron's KF substrate line ramps in Q2 2027 and reaches full capacity by mid-2028. Samsung Electro-Mechanics is expanding in Sejong and Busan.
The release names the capacity multiples and two of the dollar figures, and stops there. It does not disclose Nvidia's total advance-payment exposure across the six partners, nor the per-GPU volume these ramps are sized against. The announcement sits a day behind yesterday's DSX POD rack-scale launch on our chips tracker; the inputs and outputs of the Vera Rubin ramp, a day apart.
What others are reporting
-
Semiecosystem Read →
Industry trade analysis by Mark LaPedus; identifies TSMC exclusion as the tell that packaging has displaced wafers as the constraint, and flags Nvidia's capex-financing role as a structural shift.
When the buyer with the best balance sheet in the chain starts financing its suppliers' capex, it tells you the bottleneck is real
-
TechTimes Read →
Surfaces the $279B aggregate supply commitment figure and explains the physics case for co-packaged optics: 70% power reduction and millimeter vs. 15-30 cm electrical paths.
"Capacity is the constraint," said NVIDIA founder and CEO Jensen Huang in his GTC Berlin statement.
-
Unite.AI Read →
Provides the earliest confirmed high-volume ramp date (Singapore testing, February 2027) and per-segment capacity percentages, grounding the announcement in production milestones.
"We are in the midst of the largest infrastructure buildout in human history. Demand is extraordinary. Capacity is the constraint."
-
Securities.io Read →
Links the Amkor deal back to a July 2026 $1.5B prior agreement, establishing that these are not new relationships but a deepening of commitments already on Nvidia's books.
"We are in the midst of the largest infrastructure buildout in human history. Demand is extraordinary. Capacity is the constraint."
Originally reported by nvidianews.nvidia.com
Read the original article →Original headline: Nvidia Pulls Six Partners Into 2.5x Substrate, 300% Packaging and 15x Optics Capacity Build