TRENDING

TECHNOLOGY NEWS FROM AROUND THE WORLD

→
Nvidia gets more supplier capacity to meet AI demand
Supply Chain DiveIN-DEPTH NEWS

Nvidia gets more supplier capacity to meet AI demand

NVIDIA is expanding long-term substrate and advanced packaging commitments with suppliers. The November 17 announcement sets targets through 2029, not completed output.

· Published: ·Supply Chain Dive
Reading progress0%

What was announced on November 17?

NVIDIA said on November 17, 2026 that it was expanding long-term commitments with suppliers across several inputs needed for AI infrastructure. The announcement identifies package substrates from Kinsus and Samsung Electro-Mechanics and advanced packaging at SPIL, alongside work involving TSMC, Amkor and Foxconn. Supply Chain Dive examined the supply implications on November 20; Reuters separately reported on financing for the Taiwanese partners. The existing news record keeps the November 20 source publication date, while the event itself is dated November 17. The figures describe company plans and commitments, not completed factories or verified output.

A larger substrate commitment in Taiwan

Kinsus plans to increase advanced ABF substrate capacity dedicated to NVIDIA at its Youth campus in Taoyuan to more than 3.5 times the current level, according to NVIDIA. Production is expected to start in the second half of 2027, with additional capacity targeted for the third quarter of 2028. Reuters reported that Kinsus said it had secured more than $1 billion in NVIDIA investment. A package substrate connects a chip package to a circuit board. A shortage at this step can constrain shipments even when the silicon itself is available. The timetable remains a target, not proof that the expansion is already operating.

Packaging and South Korean facilities

Samsung Electro-Mechanics amended an existing substrate agreement to improve capabilities at its factories; NVIDIA linked the change to earlier collaboration in Sejong and Busan. The new announcement does not specify a separate output figure for that amendment. In Taiwan, NVIDIA and SPIL plan a multiyear advanced-packaging expansion supported by multibillion-dollar NVIDIA funding, with additional dedicated capacity targeted by mid-2029. NVIDIA describes lines covering interposer fabrication, assembly and testing. Reuters says it had previously reported the financing for expanding SPIL’s CoWoS capacity. These are linked but distinct stages of the manufacturing chain.

Why the production chain matters

An AI accelerator is not finished when a silicon wafer leaves a fabrication plant. Components must be connected in a package, mounted on a substrate and tested before integration into a larger system. Delays at any one stage can limit the value of investment elsewhere. NVIDIA says volume shipments of its Vera Rubin NVL72 and Vera CPUs began in the third quarter, using TSMC’s 5-nanometer and 3-nanometer processes and CoWoS packaging. That is a company account of the production setup; the announcement does not provide an independent performance benchmark or establish the cost of an individual system.

Earlier commitments and what remains uncertain

The announcement also refers to Amkor agreements disclosed in July and October, totaling nearly $3 billion and supporting packaging capacity in Arizona and South Korea. Those earlier agreements should not be counted as entirely new November 17 deals. NVIDIA likewise lists October capacity plans across several suppliers, each with its own location and timetable. A wider supplier network may reduce exposure to individual disruptions, but construction, equipment availability, manufacturing yields and geopolitical conditions can affect outcomes. A financing commitment does not establish that supply shortages have ended. The meaningful follow-up will be whether the facilities enter service on schedule and how much usable product reaches customers.

Photo: Brecht Corbeel / Unsplash ↗

Sources and verification

Be the first reader ✎ Write a comment

Comments

Reader question: What is your view or technical question about this story?

No comments yet. Be the first to contribute.

Write a comment

Your first and last name will appear publicly alongside your comment.

Perspectives on this story

Discussion examples prepared by the editor.

Kendall Morgan

In “Nvidia gets more supplier capacity to meet AI demand”, I would like more detail about the gap between peak compute and real applications. Which source supports the point?

Editor's perspective

To address this, separate observations from assumptions about the gap between peak compute and real applications. A measurable trial would help.

Parker Singh

What gives me pause here is how restrictions might affect product availability. A comparison that includes counterexamples would be more convincing than one success story.

Sage Bennett

It would be useful to compare “Nvidia gets more supplier capacity to meet AI demand” with earlier work. Are the methods and data comparable when it comes to compatibility across drivers and software ecosystems?