ANN ARBOR—The United States invented the semiconductor industry, built the first transistors, and powered the digital age—then gradually outsourced large-scale chip manufacturing overseas.

Today, amid geopolitical risk, supply-chain shocks and surging demand from electric vehicles, AI systems and data centers, researchers say the path back does not run through copying yesterday’s chips. It runs through building the chips of tomorrow, and Michigan is unusually well positioned to help lead that shift.

Researchers at the University of Michigan argue that the future of U.S. semiconductor manufacturing will be defined less by who can make the smallest transistors today and more by who can commercialize new materials, new architectures and AI-driven design methods faster than anyone else.

The University of Michigan Lurie Nanofabrication facility

From American invention to offshore dominance

Semiconductors were born in the U.S. Bell Labs developed the transistor in 1947. Early integrated circuits emerged from American firms such as Fairchild Semiconductor, Texas Instruments and Intel. For decades, U.S. companies both designed and manufactured the world’s most advanced chips.

That began to change in the 1980s and 1990s. As fabrication plants—known as fabs—became enormously expensive, many U.S. companies shifted to a “fabless” model, keeping design in-house while sending production overseas. Meanwhile, governments in East Asia made long-term, coordinated investments in chip manufacturing infrastructure, subsidies and workforce development.

The result is today’s extreme concentration: Taiwan Semiconductor Manufacturing Company (TSMC) is responsible for roughly 68% of global chip manufacturing and about 90% of the most advanced chips.

“TSMC is a great company, but it’s a single point of failure,” said Valeria Bertacco, Mary Lou Dorf Collegiate Professor of Computer Science and Engineering and vice provost for engaged learning at U-M. “There’s no benefit to anyone in having one company that can produce the most advanced technology node.”

The pandemic exposed the risk

The COVID-19 pandemic turned that vulnerability into a crisis. Semiconductor shortages left parking lots across the U.S. filled with unfinished vehicles lacking key electronics. New-car prices jumped nearly $10,000 between late 2020 and the end of 2022—a rise that previously took eight years.

Even before the pandemic, electronics accounted for about 40% of a new vehicle’s cost. Today, with EVs, advanced driver-assistance systems and software-defined vehicles, that share continues to grow—making semiconductors as essential to modern transportation as steel once was.

Why the U.S. shouldn’t chase yesterday’s chips

Trying to outspend global foundries at today’s cutting edge is not realistic, researchers say. TSMC invests roughly $40 billion per year in equipment and research. By comparison, the CHIPS and Science Act provides about $52 billion spread over five years.

At the same time, the traditional engine of chip progress—Moore’s Law—has slowed. Shrinking transistors into the nanoscale is becoming dramatically harder and more expensive.

“Moore’s Law drove an explosion in chip capability and manufacturing scale,” said Gus Evrard, the Arthur W. and Alice R. Burks Collegiate Professor of Physics at U-M. “But we’re now in a period where miniaturization alone isn’t delivering the same returns.”

Even Apple has paused its rush to the smallest nodes, opting for 3-nanometer transistors in its latest chips while holding off on 2-nanometer designs due to cost and complexity.

That pause creates an opening.

The technologies that can make chipmaking “U.S.-friendly”

Rather than playing catch-up, U-M researchers say the U.S. should focus on technologies where research strength, system integration and software-driven design matter more than raw scale.

1. Advanced materials beyond silicon

New materials are becoming central to next-generation chips:

  • Silicon carbide (SiC) for EVs, charging infrastructure and grid-scale power electronics

  • Gallium nitride (GaN) for high-frequency wireless, radar and data centers

  • Ultra-wide bandgap materials, including diamond, for extreme environments such as aerospace, defense and geothermal drilling

“These technologies don’t all use silicon,” said Becky Peterson, director of U-M’s Lurie Nanofabrication Facility. “They enable capabilities traditional chips simply can’t.”

These fabs are smaller, more specialized and better aligned with U.S. automotive, industrial and defense needs—areas where Michigan already has deep expertise.

2. Chiplets and advanced packaging

Instead of one massive chip, the industry is moving toward chiplet architectures, where multiple smaller dies are combined using advanced packaging.

This approach:

  • Lowers manufacturing costs

  • Improves yield

  • Allows production to be geographically distributed

Value shifts from a single mega-fab to system-level engineering, an area where U.S. companies and universities excel.

3. Domain-specific chips

Future growth will come from chips designed for specific uses, not just general-purpose computing:

  • Automotive safety and power management

  • Industrial automation

  • Medical devices

  • Energy systems

  • Defense and aerospace

These chips prioritize reliability, security and long lifecycles—attributes that fit Michigan’s manufacturing DNA.

AI: the force multiplier

Artificial intelligence may be the single most important technology enabling a U.S. manufacturing comeback—not just as a user of chips, but as a designer and operator of them.

AI in chip design

AI-driven electronic design automation is transforming how chips are created:

  • Exploring billions of design permutations

  • Optimizing power, heat and performance

  • Compressing design cycles from years to months

This lowers barriers to entry, allowing smaller U.S. teams to compete with global giants.

AI in fabs

AI is also reshaping manufacturing itself:

  • Predictive maintenance for billion-dollar tools

  • Yield optimization and defect detection

  • Faster tuning of new materials and processes

“AI reduces the historical advantage of massive scale,” researchers say, making agility and innovation competitive again.

Michigan’s strategic edge

Michigan sits at the intersection of these trends. The auto industry’s shift toward EVs and software-defined vehicles is driving demand for specialized power and safety chips. U-M’s Lurie Nanofabrication Facility already serves as a bridge between lab research and industrial prototyping, allowing companies to test new materials and architectures before scaling production.

With automakers nearby and a skilled engineering workforce, Michigan could anchor a regional semiconductor ecosystem focused on automotive, industrial and defense applications—rather than chasing smartphone processors made elsewhere.

The long game

Researchers caution that rebuilding domestic semiconductor manufacturing will require sustained investment.

“The CHIPS Act was a critical start,” Bertacco said. “But if it’s a one-time effort, it won’t be enough. Semiconductor ecosystems are built over decades.”

The takeaway, they argue, is clear: America’s path back into chip manufacturing isn’t about recreating the past. It’s about inventing the future—with new materials, smarter architectures and AI-driven design—and building those chips at home.

For Michigan, that future may already be taking shape.