This Is How the World’s Most Advanced AI Infrastructure Supply Chain Invents the Future
NVIDIA operations lead Deb Shoquist, honored as a Woman of Innovation, oversees the global supply chain that turns AI designs into reliable systems. Her team selects suppliers with technical fit, financial strength, and willingness to do unprecedented work, relying on long-term partners such as TSMC and Amphenol. As NVIDIA moves from chips to data center platforms, it invests in optical component supply chains and precision manufacturing in Vietnam to scale new technologies.
Key Takeaways
- NVIDIA’s AI infrastructure depends on operations that turn advanced designs into reliable physical products through a global manufacturing and supplier network.
- The company’s success relies heavily on long-term partnerships with manufacturers such as TSMC, Foxconn, Amkor, Coherent, and Lumentum, built on trust and collaboration at the factory level.
- As NVIDIA expanded from chips to full data center systems, it had to invest in and develop new supply chains, especially for emerging technologies like optical components and co-packaged optics.
- Precision manufacturing is central to NVIDIA’s scaling, as shown by partners building highly complex, high-quality components such as Amphenol’s cable cartridge in Vietnam.
- Managing scarce components, especially wafers and chips, requires weekly allocation decisions and close coordination with manufacturing partners to maintain output and quality.
NVIDIA operations lead Deb Shoquist has paused her routine supplier visits to accept the 2026 Dr. Lisa Su Woman of Innovation Award. Her role is to oversee the supply chain and manufacturing behind the company’s AI platforms, turning engineering designs into physical systems that reach customers around the world.

That summary understates the work. Shoquist, her team and a global network of manufacturing partners help NVIDIA introduce leading technologies year after year. The result is not just faster production, but the ability to build products that did not exist before.
From Blueprints to Reliable Systems
At its simplest, Shoquist describes operations as the function that moves a design into the physical world. A design team produces the technical plan; her team turns that plan into a dependable, high-performance supply chain.
She compares the work to matchmaking. A good supplier must fit NVIDIA’s production ramp style, have the technology needed for precise designs, and be financially strong enough to invest in the volume NVIDIA requires. It also must be willing to attempt work that has not been done before.
Those three qualities are essential. If a potential partner only meets two of them, NVIDIA may invest in the company to help it become a full partner.
A Network Built on Long-Term Partnerships
Much of the team’s success rests on deep relationships with major manufacturers, including TSMC, Foxconn, Wistron, SPIL, King Yuan Electronics, Unimicron, Lumentum, Coherent and Amkor.
TSMC has worked with NVIDIA since 1998. SPIL and King Yuan Electronics were assembling and testing NVIDIA chips more than two decades ago. For Shoquist, those relationships strengthen through each shared challenge on the factory floor, because every new generation of technology demands more from the people and tools behind it.
When asked about the global network she relies on, she begins with people. A supplier in Taiwan once called her at 3 a.m. with a solution to a fresh production problem. Factory workers in Vietnam asked her only for more tools so they could get more work done. She also remembers partners who supported NVIDIA when the company was smaller, and the work those relationships have done to help both sides grow.
Shoquist meets key suppliers quarterly and spends six weeks each year traveling to partners in Taiwan, China, Korea, Vietnam and the U.S. The conversations often move from meeting rooms onto production floors, where the details that matter most become visible.
Scaling Beyond Chips Into Systems
NVIDIA’s manufacturing task changed as the company moved from selling chips to building data center systems and platforms. When it was primarily a chip manufacturer, the semiconductor industry was already at scale, so the focus was on finding the best technology match for each product.
As NVIDIA became a data center systems company, it needed more partners. It also entered newer technology areas where the supply base is still developing.
Optical components are one example. Some partners have the technology and the willingness, but not yet the scale or capital needed to build fabs for the volume NVIDIA requires. That forced NVIDIA to support and invest in partners such as Coherent and Lumentum.
In areas such as co-packaged optics, NVIDIA is helping create entirely new supply chains with partners that may be experienced in manufacturing but new to optics. Integrating talent across the full supply chain is central to moving these technologies into high-volume production quickly.
Precision Manufacturing in Vietnam
Amphenol provides a clear example of what NVIDIA asks of partners. Its cable cartridge is the most complex interconnect assembly the company has ever built. Shoquist recently visited the Vietnam factory where the product is now in full production and was struck by the precision required.
The large cable cartridge contains 1,500 small cable legs. Each leg requires very precise assembly to reach a quality level of less than one defect part per million. Amphenol built over a hundred automated lines to produce the component at the speed and scale NVIDIA needed.
That kind of work requires technical capability, but it also requires commitment. Creating something that complex and precise with a supplier strengthens the partnership. The partners NVIDIA maintains are often the ones that supported the company earlier in its growth and are still willing to scale with it today.
Trust Earned on the Factory Floor
Shoquist’s time on production lines has taught her lessons that do not appear in charts or executive meetings. Over 20 years, she has met not only CEOs, but also the people running the production floor. Those individuals know the real problems of manufacturing.
When relationships are close enough, people deeper in the organization feel comfortable telling NVIDIA what it needs to fix: better technical support on the floor, more quality engineers, a better design, more automation, or more material buffers. That openness comes from years of working together and building trust.
Demand for NVIDIA technology is already massive and growing. Keeping production volume in step depends on relationships. When NVIDIA decides to push a line harder, it must understand the partner’s processes, know the people involved, and have enough trust to ask for focused, reasonable help. When both sides succeed, the partnership becomes stronger.
Managing Scarce Chips With Weekly Rules
Resource allocation is another part of the job. The team first identifies the most critical components, then protects and prioritizes production around them. Wafer supply is treated as an essential input, and the TSMC chip is a critical piece of the system.
Each week, as chips come out of assembly, NVIDIA decides which factory receives which chips. Factory performance is reviewed weekly. If one plant has low yield, or if its time of ownership is not at the desired level, it may not receive more chips that week.
The company tracks every chip from the identification number on the die all the way to the customer invoice. Suppliers must support that traceability process. With it, NVIDIA reduced the time from chip out at TSMC to system out at Foxconn and Wistron from five weeks to two weeks when supply and demand require that pace.
She describes the supply chain as a circulatory system with pathways everywhere. When one branch needs acceleration, the team has the capability to push it. That control is what makes operations both practical and powerful.
The GB200 Test Automation Crisis
Not every plan works the first time. NVIDIA, Foxconn and Wistron were preparing volume production of the compute board for NVIDIA GB200. The final step was a test process managed by an automated handler, a new approach developed to handle the throughput needed for complex products. During the initial ramp, throughput was significantly below plan.
The operations director for Taiwan manufacturing decided that protecting the product and shipments required disabling the entire automated process and returning to manual test loading. The manufacturers backed the decision without hesitation and brought in the manpower needed to keep production moving, literally hundreds of operators. That solved the immediate shipment risk, but it did not solve the throughput required for full production.
The root cause had multiple parts: test tray assembly, test tray design, automation hardware design, automation control software and product yield. Resolving it quickly was difficult.
NVIDIA responded by pulling in teams from design engineering, operations industrial, product engineering and test engineering, along with Foxconn automation engineering. They flooded the line with talent, redesigned the process and reactivated it before the larger production ramp arrived.
This kind of response is common in operations when a problem appears. The teams act with urgency and do not wait for organizational boundaries to define the solution. The mission comes first.
Why This Matters for AI Hardware
For readers outside manufacturing, the story shows that AI progress is not only about model design or chip architecture. It also depends on whether a company can find, fund and coordinate partners capable of building new technology at scale.
NVIDIA’s ability to deliver systems depends on long-term supplier relationships, weekly production decisions, traceability systems and the willingness to step onto factory floors when problems arise. Optical components, co-packaged optics and complex interconnect assemblies show how much of AI infrastructure sits in physical manufacturing, not just software or design.
Shoquist’s award highlights that part of the story. Her team is constantly changing the supply chain to meet demand for new products. The challenges are high, but the company provides the resources needed to solve them. The main limitation, in her description, is creativity.
Comments (0)