Nvidia announced expanded manufacturing commitments on November 17, 2026, with substrate and advanced-packaging capacity targets stretching into 2028 and 2029. The package spans Kinsus, Samsung Electro-Mechanics, SPIL, Amkor, TSMC and Foxconn, covering several steps between fabricating a chip and delivering an AI system.
The biggest new commitments are a Kinsus substrate expansion and a multiyear advanced-packaging agreement with SPIL. Samsung Electro-Mechanics has amended an existing agreement. Amkor’s nearly $3 billion in agreements, meanwhile, were announced earlier in 2026 and are being restated, not added again.
That distinction matters. Nvidia’s announcement describes a coordinated manufacturing plan, but it is not six newly signed capacity deals with six newly disclosed investment totals.
What Is New, Expanded or Already Announced?
Nvidia groups the commitments across substrates, advanced packaging, wafer manufacturing and contract manufacturing. Those stages serve different purposes, and their capacity figures cannot be treated as interchangeable.
Here is what the release actually discloses:
| Partner | Status and manufacturing role | Capacity timing | Funding or scale disclosed |
|---|---|---|---|
| Kinsus | New multiyear agreement for advanced ABF package substrates at its Youth campus in Taoyuan, Taiwan | Production expected in the second half of 2027; additional capacity targeted for Q3 2028 | Nvidia-dedicated capacity to expand to more than 3.5 times current levels; no dollar amount |
| Samsung Electro-Mechanics | Amendment to an existing substrate agreement, improving capabilities at existing factories; builds on collaboration in Sejong and Busan, South Korea | No new dated milestone | No amendment value or capacity figure |
| SPIL | New multiyear advanced-packaging expansion in Taiwan | Increased contracted capacity dedicated to Nvidia by mid-2029 | Multibillion-dollar Nvidia funding; exact amount undisclosed |
| Amkor | Previously announced advanced-packaging agreements covering Arizona and South Korea | Agreements announced in July and October 2026 | Nearly $3 billion combined, including prepayments |
| TSMC | Wafer and advanced-packaging production update | Vera Rubin NVL72 and Vera CPUs began volume shipments in Q3 2026 | No new capacity quantity or funding amount specified |
| Foxconn | Deeper collaboration supporting next-generation GPU and networking platforms | No new dated milestone | No new capacity quantity or funding amount specified |
The clearest numerical capacity expansion belongs to Kinsus. SPIL has a funding description and deadline, but no published throughput figure. Samsung’s amendment has neither. That makes a single headline number for the entire package impractical: the announcement mixes capacity ratios, financial commitments and production updates.
Substrates: Kinsus Has a Measurable Expansion
Kinsus plans to expand advanced ABF package-substrate capacity dedicated to Nvidia to more than 3.5 times current levels at its Youth campus. Nvidia expects production to begin in the second half of 2027, with the additional capacity targeted for the third quarter of 2028.
Those are two separate milestones. Starting production in 2027 does not mean the full planned expansion will be available then. The later date is the useful reference for readers tracking the additional capacity.
Package substrates provide the electrical connection between a semiconductor package and its circuit board. As VLSI’s technical breakdown explains, ABF substrates contain the build-up layers used in these connections, and larger accelerator designs can require more substrate area and layers.
That creates an important limit on the headline multiplier. More than 3.5 times the substrate capacity is not a forecast for 3.5 times as many Nvidia GPUs. Nvidia gives no absolute starting capacity, and different packages can consume different amounts of manufacturing capacity. Finished-system output also depends on the other production stages.
Samsung Electro-Mechanics is a less quantifiable part of the announcement. Nvidia says the amendment improves operational capabilities at existing factories and builds on its previously announced multiyear collaboration supporting compute and high-speed networking substrates in Sejong and Busan.
I would treat that as an expansion of an established manufacturing relationship, without assigning it a speculative output increase. The release supplies no basis for calling it a new factory project or estimating how much additional capacity it creates.
SPIL Adds Packaging Capacity; Amkor’s Funding Is Earlier
SPIL’s agreement expands advanced-packaging operations in Taiwan with multibillion-dollar funding from Nvidia. The target is increased contracted capacity dedicated to Nvidia by mid-2029.
The project covers new production lines with turnkey support across four steps:
- Bumping: forming small electrical contacts used to connect semiconductor components.
- Interposer wafer fabrication: producing the intermediate structures that connect components within an advanced package.
- Assembly: integrating those components into the package.
- Testing: checking the completed devices.
The scope matters as much as the funding description. Advanced packaging is a sequence of manufacturing operations, not one machine or one final assembly step. Coordinating capacity across that sequence can help avoid a situation in which one operation expands while another remains constrained.
Nvidia describes the SPIL funding only as multibillion-dollar. It gives neither an exact amount nor a capacity increase measured in wafers, packages or systems. The mid-2029 milestone is therefore more useful for understanding the production plan than for forecasting a specific shipment total.
Amkor’s agreements are separate. Nvidia says they total nearly $3 billion, cover expansion in Arizona and South Korea, and include prepayments to help Amkor scale manufacturing. They were announced in July and October 2026.
Readers tracking AI industry investment should keep those earlier commitments out of any calculation of newly announced November spending. They also should not label the entire amount as Nvidia-owned factory investment: the disclosed structure includes supply agreements and prepayments supporting a manufacturing partner.
My reading is that the financing mechanism is consequential. Prepayments help connect a supplier’s expansion spending with a customer’s expected future production needs. The release does not disclose the full commercial terms, so it cannot tell us how the companies divide demand or utilization risk.
TSMC and Foxconn Connect the Plan to Shipping Systems
Nvidia says Vera Rubin NVL72 and Vera CPUs began volume shipments in the third quarter of 2026, using TSMC’s 5nm, 3nm and CoWoS advanced-packaging technologies.
The process-node references describe semiconductor manufacturing technologies. CoWoS is an advanced-packaging technology, a different part of the production flow. Listing them together underscores that access to wafer fabrication alone is insufficient to deliver the finished products.
Nvidia does not provide a product-by-product allocation of those technologies in this announcement. Nor does it disclose a newly reserved TSMC wafer or CoWoS capacity figure. This part of the release is primarily a production-status update and a statement of continuing collaboration.
Foxconn sits farther along the chain, supporting manufacturing for next-generation GPU and networking platforms through its global production network. Its inclusion broadens the announcement beyond chipmaking: Nvidia also needs manufacturing capacity to turn components into deliverable infrastructure.
The Foxconn update carries no new dated capacity target or financial amount. It should be read as part of the wider production plan, not another quantified investment.
Frequently Asked Questions
3 questions
1When Is Nvidia’s New Capacity Expected?
Kinsus production is expected in the second half of 2027, with additional capacity targeted for Q3 2028. SPIL’s contracted advanced-packaging capacity increase is targeted for mid-2029.
