Tag: Nvidia

Guest Post: How AI Has Transformed Semiconductor Scheduling at NVIDIA and TSMC

Professor Misty Blessley raises an interesting AI issue-chip plant scheduling.

Semiconductors have been called the brains of the modern world. Switching between conducting and blocking electricity, they are essentially “on/off” switches that control the flow of power, and this unique characteristic makes them the building blocks of modern technology. They are behind computing, communications, and energy innovation, and our information dependent society is always hungry for more. Leading chip makers, NVIDIA and TSMC, both use an AI to advance semiconductor design, scheduling, and manufacturing, report ElectronicsUSA (June 4, 2026) and NVIDIA News (May 31, 2026).

TSMC ( Taiwan Semiconductor Manufacturing Company), the world’s largest semiconductor foundry, makes the most advanced chips on the planet. NVIDIA is a global leader in accelerated computing and AI. For decades, TSMC has manufactured NVIDIA’s chips, and this partnership has come full circle. TSMC uses NVIDIA’s AI technologies inside its fabrication plants (“fabs”), which in the semiconductor industry refers to highly specialized facilities where silicon wafers are processed into microchips.

NVIDIA’s AI models are now embedded directly into TSMC’s manufacturing workflow, transforming production scheduling. A single wafer, a subcomponent of a semiconductor, may require hundreds of tools and thousands of tightly sequenced steps. TSMC can now evaluate millions of scheduling combinations in seconds. This optimizes job sequencing, and instantly rebalancing schedules when tools go down or urgent orders arrive. The result is smoother production flows, less idle time, and higher overall fab productivity. AI comes as close as ever to “running the plant.”

AI driven scheduling is also advancing through emerging capabilities such as predictive dispatching, where models forecast bottlenecks hours ahead and reroute wafers to prevent delays. Another is crossfab load balancing, which evaluates capacity across multiple TSMC sites and shifts work to maximize throughput. Both approaches are expected to reduce fab cycle time 5–10%.

NVIDIA’s AI scheduling helped TSMC cut critical chip production workloads 20-50%. This addresses society’s insatiable demand for semiconductors. It also fuels the next wave of breakthroughs in computing, communications, and energy innovation, the domains powered by these tiny on/off switches.

Classroom discussion questions:

  1. What are the strategic and operational risks if a company relies too heavily on AI to automatically reroute work without human oversight?
  2. Discuss how AI’s ability to evaluate millions of combinations in seconds changes a manager’s approach to bottleneck scheduling compared to manual or sequential sequencing rules (like SPT or FCFS).

OM in the News: PepsiCo Turns To Digital Twins To Rethink Plants

We posted recently about the joint nuclear fusion digital twin work of Siemens and NVIDIA. Today’s news is that PepsiCo is working with the same two firms  to change how it designs, tests, and expands its plants and warehouses using AI and digital twins. “Physical industries are entering the age of AI. For companies with real-world assets, digital twins are the foundation of their AI journey,” said NVIDIA’s CEO.

By modeling factories and distribution centers digitally before making physical changes, PepsiCo hopes to cut down on costly mistakes while improving speed and capacity.

With AI-driven digital twins, teams can simulate plant layouts, equipment movement, and supply chain operations in detail, reports SupplyChain (Jan. 7, 2026). Instead of expanding facilities the old way, which can be slow and expensive, they can test changes virtually and see what works before spending money on physical upgrades.

“The scale and complexity of PepsiCo’s business is massive—and we are embedding AI throughout our operations to better meet the increasing demands of our consumers and customers,” said PepsiCo’s CEO. The digital models recreate machines, conveyors, pallet routes, and even worker movement, helping teams spot problems early and test different setups in weeks instead of months.

By finding bottlenecks and unused capacity in a virtual setting, teams increased throughput by 20%. The same approach has also shortened design cycles and helped cut capital spending by 10-15%. Testing ideas digitally first, teams can plan ahead, compare options, and move faster without the usual surprises that come with physical expansion.

Classroom discussion questions:

  1.  How is PepsiCo employing digital twins?
  2. How do AI and digital twins work together?

OM in the News: Digital Twins and Nuclear Fusion

Digital twins, which we cover in Module F (Simulations and Digital Twins), is a big topic at Nvidia and Siemens as they work together to make nuclear fusion a commercial reality. In that chapter (see p. 847), we define a digital twin as:  “an electronic virtual replica of an operation that allows organizations to mimic how a product, process, or system will perform.”

Workers at Commonwealth Fusion Systems’ campus in Devens, Mass

Fusion engineers at the Nvidia/Siemens venture, called Commonwealth Fusion Systems (CFS), will use its digital twin to run simulations, ultimately to hasten the goal of producing fusion energy at a commercial scale. CFS “will be able to compress years of manual experimentation into weeks” with the AI assistance, said its CEO.

Nuclear fission, which splits atoms to produce energy, is already in use in power plants, reports The Wall Street Journal (Jan. 7, 2026).  But many companies see fusion, the energy process that powers the sun by joining atoms together, as a longer-term bet because it can provide much more energy in a cleaner process. Nuclear energy appeals to tech giants because it releases minimal carbon emissions while providing round-the-clock power—particularly as they look to fuel their AI ambitions.

CFS said it was working with Google on an AI project, and explained that that effort has created something like a co-pilot for its fusion machine, while the digital twin plan “is the virtual airplane.” Google also recently signed a power purchase agreement with CFS to secure energy from what could be the first grid-scale fusion plant.

“The race is on for AI. Everyone is trying to get to the next frontier,” said Nvidia’s CEO.

Classroom discussion questions:

  1. Provide other examples of how digital twins can be used.
  2. Why is this fusion project so important as an OM tool?

OM in the News: AI Will Soon Take Your Order at Taco Bell

operations management blog artificial intelligenceYum Brands, owner of Taco Bell, Pizza Hut, and KFC is partnering with Nvidia to build a range of new AI-driven services in its restaurants. The first—AI-powered voice-ordering at the drive-through lane and on the phone—was built using tools from Nvidia, and will begin rolling out at 500 Yum Brand restaurants this year.

The ultimate goal is to move all orders through digital channels instead of human order-takers, an effort Yum says will boost sales. Yum will also use AI to enhance a number of the company’s internal operations.

Other planned changes include the use of computer vision to spot fumbled orders and AI that filters internet chatter on the restaurants for useful feedback for their managers. “Yum and other quick-service-restaurant chains, like McDonald’s, have been leaning into more digital experiments for efficiency gains and improved customer satisfaction as inflation squeezes low-income diners,” writes The Wall Street Journal (March 19, 2025). 

Voice-ordering has been a priority for Yum for some time as it works to receive 100% of its orders through digital channels rather than through humans. Currently it is above 50% including orders that come through its app or online, up from 19% in 2019. Consumers end up spending more when they buy via digital channels because the restaurant can upsell, personalize and entice eaters through notifications.

Yum is currently evaluating whether existing CCTV cameras can provide images sharp enough for computer vision to determine whether the food received is what was ordered or whether it’s missing any ingredients. “Order accuracy is a big problem that a lot of quick-service restaurant companies face,” says the Chief Technology Officer.

In-restaurant workers won’t disappear. Instead, they will focus more on customer service, for instance, helping people with orders.

Classroom discussion questions:

  1. How else can AI be used in this industry?
  2. What are some disadvantages of depending on AI?