US Chip Strategy Requires Fabs, New Architectures, and Skilled Workers

Semiconductors—materials including silicon and gallium that regulate the flow of electricity—are the operating substrate of modern electronic systems. They determine how devices function and how information moves within them. That makes chips foundational not only to consumer technology but also to US economic competitiveness, artificial intelligence, and defense systems.

The United States produces about 12% of all chips worldwide. Taiwan manufactures more than 60% of global chips and nearly 90% of the most advanced chips. The result is a heavily concentrated manufacturing base, particularly for advanced production.

Chip supply remains fragile, and this concentration creates economic and national-security risks. The risks are heightened by China’s ambitions toward Taiwan and by US export controls limiting China’s access to advanced chips. Access to critical computing hardware is therefore tied to geopolitical exposure as well as production capacity.

The policy aim is greater domestic manufacturing capacity alongside a more diverse and secure supply chain, reducing vulnerability to geopolitical disruption. But expanding capacity is not simply a matter of building more fabrication plants. The industry must also sustain computing advances as conventional chip scaling becomes harder, manage AI’s energy demands, and develop the workforce needed to operate and support new capacity.

For decades, the industry’s central performance pattern was Moore’s Law: the number of transistors on a microchip roughly doubled every two years while the cost of making the chip stayed approximately flat. That trajectory has slowed, weakening the assumption that computing gains will continue to arrive principally through ever-smaller, denser chips.

The alternatives are architectural and packaging advances rather than a return to the old pace of transistor scaling. Chiplets, 2.5D integration, photonic interconnects, and high-bandwidth memory are presented as routes to continued computing growth. Their intended benefits are lower energy consumption, greater bandwidth, and higher performance.

But this is not a frictionless substitution. These approaches add cost and engineering complexity, while chip development itself is slow. Development can require more than $100 million and more than two years—a combination that makes rapid iteration difficult even as demand for new computing capacity rises.

The pressure is especially acute for AI and machine learning. Growing demand for those workloads is driving unprecedented energy needs. Chiplets and photonics matter in this framing not merely as performance technologies, but as technologies intended to improve efficiency through lower energy consumption alongside higher performance.

Building domestic fabrication capacity will require more than capital investment. The semiconductor workforce is itself a constraint. The US semiconductor sector is expected to support an additional 115,000 jobs by 2030, but roughly 67,000 of those roles—58%—risk going unfilled.

That gap makes talent development a manufacturing and security issue, not simply a labor-market concern. The proposed remedy is closer collaboration among educational institutions, industry, and government to develop, attract, and retain skilled semiconductor workers. Training people is insufficient if the sector cannot also recruit them and keep them in the field.

Further manufacturing investment is also necessary. The CHIPS Act is cited as the kind of effort intended to build domestic capacity and mitigate geopolitical risk through supply-chain diversification.

The policy challenge runs across linked horizons. Innovation must sustain computing progress as conventional scaling slows. Manufacturing investment must address geographic concentration. Workforce policy must ensure new capacity can be staffed. Energy-efficient chip technologies must keep pace with AI’s growing computational requirements.