Business•

Nvidia Puts More Than $1 Billion Into Kinsus to Lift Its Dedicated Chip Substrate Capacity 3.5 Times

Nvidia will invest more than $1 billion in Taiwan's Kinsus to expand ABF substrate output at its Taoyuan campus, with capacity reserved for Nvidia rising over 3.5 times by 2028.

Written By :MINRK
Nvidia Puts More Than $1 Billion Into Kinsus to Lift Its Dedicated Chip Substrate Capacity 3.5 Times

Photo by Vishnu Mohanan on Unsplash

1. Nvidia Kinsus investment tops $1 billion

Nvidia announced on November 17, 2026 that it will invest more than $1 billion in Kinsus Interconnect Technology, a Taiwan-based supplier, to expand production of advanced ABF package substrates. Under the Nvidia Kinsus agreement, capacity dedicated to Nvidia is expected to rise by more than 3.5 times, with full output reached by the third quarter of 2028.

The money is intended to enlarge Kinsus's manufacturing base at its Youth campus in Taoyuan, in northern Taiwan. Operations tied to the expansion are projected to begin in the second half of 2027, and the new lines are then expected to ramp until they reach full capacity in the third quarter of the following year.

Nvidia's stated purpose is to strengthen its artificial intelligence platform by locking in a dedicated future supply of substrates. The commitment is one element of a wider supply chain effort that also involves Amkor, Foxconn and Samsung Electro-Mechanics, companies that handle other stages of turning a finished silicon wafer into a working AI system.

2. What does Kinsus do?

Kinsus Interconnect Technology makes package substrates, the small multilayer boards on which semiconductor chips are mounted. The company is based in Taiwan and supplies these components to chip designers and to the firms that assemble chips into finished packages.

A substrate is easy to overlook because it is hidden beneath the chip, but no advanced processor works without one. A silicon die has thousands of microscopic electrical contacts that are far too fine and too closely spaced to be soldered directly to a circuit board. The substrate sits between the two. It takes the dense pattern of connections on the underside of the chip and fans them out, layer by layer, into the wider pattern that a motherboard can accept, while also delivering power and carrying heat away.

The agreement with Nvidia concerns ABF substrates specifically. ABF stands for Ajinomoto Build-up Film, an insulating material used to build up the many thin wiring layers inside high-end substrates. ABF-based substrates are the standard choice for high-performance processors, including the graphics processors and accelerators used for artificial intelligence, because they support very fine wiring and large numbers of layers.

Kinsus's Youth campus in Taoyuan is where the additional capacity will be installed. Taoyuan is one of Taiwan's main centers for electronics manufacturing, with a dense cluster of circuit board, substrate and component makers, which makes it a natural place for an expansion of this kind.

3. Why substrates matter for AI chips

Substrate supply has become a strategic issue because AI processors place unusually heavy demands on it. The chips at the heart of data center systems are physically large, and modern packages often combine a processor with several memory components on a single base. A bigger package needs a bigger substrate, with more layers and tighter tolerances, and each of those requirements consumes more manufacturing capacity than a conventional chip does.

Yields add to the pressure. Large, complex substrates are harder to make without defects, so a factory produces fewer good units from the same equipment than it would for smaller designs. The effective capacity available for the most demanding products is therefore lower than a simple count of production lines would suggest.

New capacity is also slow to arrive. Substrate plants require clean rooms, specialized plating and lithography equipment and a long qualification process before a customer will accept their output. The schedule attached to the Kinsus expansion illustrates the point: production is not expected to begin until the second half of 2027, and full capacity is not due until the third quarter of 2028. A chip designer that wants substrates in 2028 has to commit money well in advance.

Chip designers such as Nvidia do not own factories. Nvidia designs its processors and relies on outside partners for wafer fabrication, packaging, substrates, testing and system assembly. A shortage at any one of those stages limits how many finished products can be shipped, regardless of how much capacity exists at the others. Substrates have been such a bottleneck in the past, and securing them is one way of making sure that output is not capped by a comparatively inexpensive component.

4. How the deal is structured

The defining feature of the arrangement is that the capacity is dedicated. Nvidia is not simply placing a large order; it is funding an expansion in return for output reserved for its own products. Capacity devoted to Nvidia at Kinsus is expected to grow to more than 3.5 times its current level once the project is complete.

For the supplier, the benefit is obvious. Building a substrate plant is expensive, and the risk for any manufacturer is that demand fades before the investment has paid for itself. A commitment of more than $1 billion from the customer that will consume the output removes much of that risk and allows the expansion to go ahead on a scale that might otherwise be hard to justify.

For Nvidia, the benefit is certainty. Dedicated lines cannot be reallocated to a competitor when supply is tight, and a guaranteed volume of substrates makes it easier to plan production of future chip generations years ahead. The trade-off is that capital is tied up in a supplier's facilities long before any revenue is earned from the products those facilities will support.

The timetable sets clear milestones. The first is the start of operations in the latter half of 2027. The second is the point at which the new lines are running at full capacity, targeted for the third quarter of 2028. Between those two dates, output is expected to climb as equipment is installed and qualified.

5. Part of a wider supply chain push

The Kinsus commitment belongs to a broader strategy, and Nvidia has named Amkor, Foxconn and Samsung Electro-Mechanics among the other partners involved. Each occupies a different link in the chain. Amkor is one of the largest independent providers of chip packaging and testing. Foxconn is a contract manufacturer that assembles servers and other electronic systems. Samsung Electro-Mechanics is a South Korean component maker whose products include package substrates.

Taken together, the list shows that Nvidia is addressing several potential constraints at once instead of concentrating on a single one. Substrates from Kinsus and Samsung Electro-Mechanics, packaging from Amkor and system assembly from Foxconn all have to be available in matching quantities for finished AI hardware to reach customers.

The approach mirrors a wider shift in how the largest buyers of semiconductors deal with their suppliers. For most of the industry's history, chip designers placed orders and left capacity decisions to the manufacturers. As demand for AI hardware has outrun the ability of the supply chain to respond, the biggest customers have increasingly chosen to put their own capital behind suppliers' expansion plans in exchange for guaranteed access. The Kinsus investment follows that model, applied to a component that sits several steps away from the chip itself.

It also reflects a preference for more than one source. Relying on a single supplier for a critical component exposes a company to disruption from accidents, natural disasters or commercial disputes. Spreading commitments across several suppliers in different countries reduces that exposure, even if it means managing a more complicated set of relationships.

6. What the commitment says about AI demand

A company does not reserve capacity for 2028 unless it expects to need it. The scale and duration of the investment indicate that Nvidia anticipates continued growth in demand for AI hardware over the next several years, well beyond the products it is selling today.

That expectation is relevant to a debate that has accompanied the industry's rapid expansion: whether spending on AI infrastructure can be sustained, or whether it resembles a bubble that will eventually deflate. A long-dated capital commitment from the largest supplier of AI processors is evidence of how that supplier sees the outlook, and some observers regard such investments as a stabilizing signal that may ease concern about a sudden downturn.

It is not proof, however. Suppliers and customers in the semiconductor industry have misjudged demand before, and capacity built for one forecast can end up underused if conditions change. The investment shows what Nvidia is planning for. Whether that plan proves correct depends on the purchasing decisions of the cloud providers, companies and governments that buy its systems.

7. What to watch next

The most direct test will be progress at the Taoyuan site. The start of operations in the second half of 2027 and the achievement of full capacity by the third quarter of 2028 are the two dates against which the project can be measured. The timing and scale of the expansion will show whether Nvidia can obtain the substrates it needs to meet the demand it expects.

Nvidia's quarterly earnings and strategic updates will provide a running commentary in the meantime. Any change in the company's growth projections or in the way it describes its market position would affect how the Kinsus investment is judged, since the case for the expansion rests on demand continuing to rise.

Further partnerships are the third thing to monitor. The agreements with Amkor, Foxconn and Samsung Electro-Mechanics suggest that the Kinsus deal is one step in a continuing program, and additional arrangements with suppliers or with other large technology companies would indicate how far Nvidia intends to go in securing its supply chain ahead of the next generations of AI hardware.

Related Articles

NEWSLETTERS

Don't miss another story.

Subscribe to the MINRK Newsletter today.

Tick the newsletters you want. No selection? You'll get MINRK Headlines.

By signing up, you will receive emails about MINRK products and you agree to our terms of use and privacy policy.

MINRK Daybook

Market analysis for crypto traders and investors.

EVERY WEEKDAY

MINRK for Advisors

Defining crypto, digital assets and the future of finance for financial advisors.

EVERY THURSDAY

MINRK Protocol

Exploring the tech behind crypto one block at a time.

WEEKLY

MINRK Long & Short

A must read for institutions. Insights, news and analysis delivered weekly.

EVERY WEDNESDAY

MINRK Headlines

The biggest crypto news and ideas of the day.

EVERY WEEKDAY

MINRK Policy

Examining the intersection of cryptocurrency and government.

WEEKLY

MINRK Research

Join thousands of readers who rely on MINRK for data-driven insights on the latest digital asset trends.

MONTHLY