Deb Shoquist’s remit at NVIDIA spans the manufacturing and logistics network behind its artificial-intelligence products, from foundry operations and supplier management to production planning and quality control. The company says she joined in 2007 as senior vice-president of operations and became executive vice-president two years later. Her role has grown in importance as NVIDIA has expanded from designing chips to delivering complete data-centre systems.
That shift has al...
Continue Reading This Article
Enjoy this article as well as all of our content, including reports, news, tips and more.
By registering or signing into your SRM Today account, you agree to SRM Today's Terms of Use and consent to the processing of your personal information as described in our Privacy Policy.
Some of those partnerships have long histories. NVIDIA’s account traces its relationship with Taiwan Semiconductor Manufacturing Company, or TSMC, to 1998. Chip assembly and testing firms Siliconware Precision Industries, known as SPIL, and King Yuan Electronics have also worked with the company for more than two decades. NVIDIA’s wider manufacturing network includes companies such as Foxconn and Wistron, which build and assemble systems.
The scale and precision required are illustrated by an Amphenol cable cartridge used in NVIDIA equipment. The assembly contains about 1,500 small cable legs and is reported to require a defect rate below one part per million. Amphenol developed more than 100 automated production lines in Vietnam to meet the volume and accuracy demands. These figures and the account of the supplier’s production work are reported descriptions, rather than independently audited performance data.
NVIDIA’s own disclosures offer broader context for the network behind such products. Its fiscal 2025 sustainability report says the company has expanded supplier relationships to improve resilience and secure manufacturing capacity in line with customer demand. The supply chain is concentrated mainly in Asia-Pacific. TSMC and Samsung make semiconductor wafers, while contract manufacturers including Hon Hai Precision Industry, the parent company of Foxconn, and Wistron support assembly and other production stages.
The report also describes a formal system for managing suppliers. Strategic partners may be reviewed quarterly or twice a year, and performance can influence how business is allocated. NVIDIA says it assesses suppliers by product category and conducts annual risk reviews that consider factors including geography and industry. Its stated expectations extend to responsible sourcing, environmental and social performance, security and compliance.
As AI systems have become larger, the company has had to build capacity in areas where technical expertise exists but production may not yet be sufficient. NVIDIA has supported optical suppliers Coherent and Lumentum as they work towards higher-volume output. It is also developing co-packaged optics, which brings optical components closer to processors in order to meet the demands of advanced systems.
A recent NVIDIA announcement points to further expansion plans. The company said it is working with partners including Amkor, SPIL, King Yuan Electronics and Unimicron on additional packaging, testing and optical capacity. It has set a target of increasing interposer-wafer capacity in South Korea by more than 300 per cent by the end of 2028, and said capacity for optical engines used in co-packaged optics is planned to rise fifteen-fold by mid-2027, with a further doubling by the end of 2028. These are company targets, not completed production results, and the announcement notes that such projections are subject to risks and uncertainties.
The manufacturing challenge is evident in the design of rack-scale AI systems. NVIDIA says its Vera Rubin NVL72 rack contains 1.3 million individual components and nearly 1,300 chips. In a technical account, the company describes how redesigning compute trays cut assembly time from nearly two hours to five minutes. The figures relate to that system, not necessarily to every NVIDIA rack, but they demonstrate the engineering and production coordination required to move from chip design to large-scale infrastructure.
NVIDIA’s Taiwan ecosystem also illustrates the breadth of that co-operation. The company says more than one million MGX rack components are brought together across 25 factory sites in the region, which has more than 500 NVIDIA ecosystem partners. Its examples include suppliers and manufacturers using simulation, robotics and digital-twin tools to plan and manage production. Those descriptions come from NVIDIA and its partners, but they underline that the relationship extends beyond buying components: suppliers also contribute to the processes used to make and deploy the systems.
For Shoquist’s operations organisation, the challenge is therefore not simply to secure more parts. It is to help a dispersed group of manufacturers develop new capabilities, maintain quality and increase output as product designs evolve. The company’s model relies on sustained engagement with partners, backed by formal reviews and capacity planning, to turn technically ambitious designs into equipment that can be produced and installed at scale.
Source: Noah Wire Services



