The concentration of semiconductor production in Taiwan is usually discussed as a map problem. Taiwan is close to China, the Taiwan Strait is tense, and TSMC sits at the center of leading-edge chip production. All of that matters. The map is still only half the problem.
Roughly 90 percent of the world's leading-edge semiconductors are made in Taiwan. These chips support data centers, defense systems, medical equipment, and scientific instruments. If production stopped, the impact would travel through industries that do not think of themselves as semiconductor businesses.
Building fabs elsewhere reduces some of the geographic risk. It does not recreate TSMC overnight. Advanced chip production depends on thousands of process decisions, experienced operators, specialist equipment suppliers, chemicals, and years of yield improvement. A new building can be finished before the team inside it can produce at the same quality and volume as an established site.
Procurement teams usually see this dependency indirectly. A supplier rarely says that one specific chip has stopped an order. Instead, a delivery moves by twelve weeks. A quoted component disappears before the purchase order is placed. An equipment maker refers vaguely to "component availability" and offers no reliable recovery date. The disruption arrives as a series of small delays until a production plan no longer works.
TSMC is expanding in Arizona and Japan, while the ESMC venture in Dresden will add European capacity. These are useful investments. They will also produce different technologies for different markets. A fab making an established process for automotive customers does not replace a leading-edge line making 2nm or 3nm chips. Capacity is not interchangeable simply because both sites produce semiconductors.
The same limitation applies to people. The engineers who can run an advanced process at production scale are not waiting in a large labor pool. Moving experienced staff helps a new site start, but it can also stretch the original operation. Local suppliers and technicians need time to learn the process and its failure modes.
This is why the "Silicon Shield" argument makes me uneasy. The theory says that Taiwan's importance to the global economy discourages conflict. That may be true, but it asks political decisions to remain economically rational under pressure. A procurement plan should not depend on that assumption.
The sensible response is not to predict a crisis date. It is to identify where a single chip, foundry, or process node can stop delivery. Some risks can be reduced through redesigns, approved alternatives, or additional inventory. Others cannot. Those need honest lead-time assumptions and a recovery plan measured in years, not quarters.