The Chip Curtain

In 2023, China spent more money importing computer chips than it spent importing oil. The chips ran to about 349 billion dollars. The crude oil came to about 338 billion. For a country that the world still calls its factory, that is a strange and revealing fact. China can assemble almost anything. It cannot yet make the one part that matters most.

That single number is a doorway into the most important supply chain on the planet, and into a question that sits much closer to home than it first appears. The semiconductor supply chain is the most globally segmented system human beings have ever built. It is also the most fought over. And the fight, once you follow it all the way down, turns out to be a fight about people who know how to do hard technical work. That is our beat. So let us walk through it.

The most segmented machine ever built

A modern advanced chip is not made in a country. It is made in a relay race across several of them, and almost every leg of that race is run by a near-monopoly.

It starts with design. The software used to design chips, and the core intellectual property they are built from, is dominated by the United States, which holds roughly 74 percent of the global market in electronic design tools and core IP and about 67 percent of logic chip design. If you want to design a leading-edge processor, you almost certainly do it with American tools.

Then the design has to be fabricated, and here the concentration becomes extreme. Taiwan accounts for about 92 percent of the world's most advanced logic manufacturing capacity, meaning the cutting-edge nodes below 10 nanometers, with South Korea holding most of the rest. One company, TSMC, runs nearly all of it, on one island, within easy reach of the most contested strait in the world.

The machines that do the fabricating are their own chokepoint. The single most advanced step, extreme ultraviolet lithography, is performed by a tool that exactly one company on earth knows how to build: ASML, in the Netherlands. Not a leading share. The entire supply. As of 2025 there were a few hundred of these machines in service worldwide, and not one of them was in China.

Memory chips lean on South Korea, which holds about 59 percent of that market. Final assembly, test, and packaging, the less glamorous back end, is concentrated in China and Taiwan, with the United States doing about 2 percent of it. A landmark 2021 study by the Semiconductor Industry Association and Boston Consulting Group counted more than 50 points along this chain where a single region controls over 65 percent of the global market.

No country can make an advanced chip by itself. That is not an accident of the market. It is the most efficient division of labor ever assembled, and also the most fragile.

The United States illustrates the fragility better than anyone. American companies sell close to half of the world's chips, about 48 percent by revenue. But the United States manufactures only about 12 percent of them, down from 37 percent in 1990. The design genius stayed home. The making of things went abroad. That gap, between who profits from chips and who actually builds them, is the anxiety driving everything that follows.

The fence goes up

For decades this segmentation was treated as simple economics. Make each piece wherever it is cheapest and best, ship it across borders, and everyone wins. Then policy makers looked at the map and saw something else: a set of pressure points, most of them friendly to the United States and its allies, that could be turned into leverage.

In October 2022, the US Bureau of Industry and Security did exactly that. It restricted the export to China of the most advanced computing chips, the kind that train artificial intelligence systems, and of the equipment needed to make them. The rules drew a line at specific technical thresholds, cutting off the top tier of Nvidia accelerators and the tools required to run leading-edge logic and memory fabs. A year later, in October 2023, the rules were tightened again to close the workarounds that had appeared in the meantime.

The strategy has a memorable name. National Security Adviser Jake Sullivan called it a "small yard, high fence" in a 2023 speech: keep the protected zone of technology narrow, but defend it absolutely. Individual companies had already been walled off one at a time. Huawei was placed on the US Entity List in 2019, SMIC, China's largest chipmaker, in 2020. The export controls turned those individual cases into a category.

Crucially, the United States did not act alone, because it could not. The fence only holds if the other chokepoint countries hold it too. The Netherlands, home of ASML, and Japan, home of much of the world's chipmaking equipment and materials, aligned their own export rules through 2023. ASML has never shipped an extreme ultraviolet machine to China. That is the whole game in one sentence. The most concentrated step in the most concentrated supply chain on earth sits inside an alliance, and China is on the outside of it.

China's wall, and the gap in it

China has not accepted this quietly, and it has more room to maneuver than the headlines suggest.

The money is enormous. China's "Big Fund," its national semiconductor investment vehicle, has run three rounds totaling close to 690 billion yuan, including a third round in 2024 of about 47 billion dollars, the largest yet. The political goal is older: the Made in China 2025 plan set a target of producing 70 percent of the chips the country uses by 2025.

It did not get close. Depending on how you count, China produces somewhere between a fifth and half of its own demand, and by the strictest measure, chips actually made by Chinese-headquartered firms, the share was in the single digits at the start of the decade. The 70 percent goal was a statement of intent, and the distance between intent and reality is the clearest measure of how hard this problem is.

And yet. In August 2023, Huawei released a phone, the Mate 60 Pro, built around a processor that outside analysts confirmed was a 7-nanometer class chip made by SMIC, inside China, without any access to ASML's forbidden machine. They had done it the hard way, using older lithography tools and clever multi-step patterning. It was a genuine surprise, and it was also a ceiling. You can push older equipment a long way, but you cannot push it to the leading edge economically, and you cannot buy the machine that would let you. The wall has a gap in it. The gap is not wide enough to climb through at scale.

So the segmentation holds. Not because of any single law, but because each layer, the design tools, the one lithography company, the single dominant fab, the allied equipment makers, reinforces the others. It is the most durable structure in the global economy, and the entire weight of US and Chinese industrial policy is now pushing against it.

Reshoring is a teaching problem

Here is where the story comes home, and where it stops being about geopolitics and starts being about us.

The United States decided it could no longer accept making 12 percent of the world's chips, so in 2022 it passed the CHIPS and Science Act: about 52.7 billion dollars in direct funding, including roughly 39 billion in manufacturing incentives, plus a separate 25 percent tax credit for building fabs. Money started moving fast. TSMC committed to a giant complex in Arizona. Intel broke ground in Ohio. Samsung expanded in Texas.

And almost immediately the industry ran into a wall that no amount of money can knock down quickly: there are not enough people who know how to do the work.

This is not a soft worry. It is measured. In 2023 the Semiconductor Industry Association, working with Oxford Economics, projected that the US semiconductor workforce would need to grow by about 115,000 jobs by 2030, and that roughly 67,000 of those jobs, well over half, were at risk of going unfilled at the current rate at which Americans earn the relevant credentials. The shortfall is not mostly PhDs. The single largest piece of it, about 39 percent, is technicians, the people with a two-year degree or a certificate who actually run and maintain the line. Engineers with four-year degrees and computer scientists make up another 35 percent. The advanced researchers are the smallest slice.

Widen the lens and it gets larger. The same analysis put the broader economy's need at 3.85 million jobs requiring technical proficiency by 2030, with about 1.4 million of them at risk of going unfilled. Deloitte, looking globally, estimates the semiconductor industry alone needs more than a million additional skilled workers by 2030, on the order of a hundred thousand a year.

You can appropriate 52 billion dollars in an afternoon. You cannot train a semiconductor technician in an afternoon, and you cannot import one if the policy goal is to stop relying on imports.

The people who wrote the workforce report were blunt about the fix, and it is not glamorous. They called for certification boot camps, apprenticeships, and training programs at community and technical colleges located near the new fabs. The binding constraint on American chip independence is not capital, and it is not even the lithography machine. It is the pipeline of trained people, and that pipeline runs through classrooms, community colleges, and credential programs.

This is the part that does not appear in the trade-war headlines, and it is the part that matters most to anyone who works in education or operations. A fab is not just process engineers. It is procurement planners, logistics coordinators, import and export compliance staff, and the whole operations apparatus that keeps materials flowing through a factory that runs every hour of every day. When TSMC stands up a 165 billion dollar complex in the Arizona desert, it is hiring supply chain talent alongside the silicon experts. Intel's Ohio project is projected to create thousands of direct jobs and many thousands more across its supplier and logistics ecosystem. The same labor math that constrains the fab constrains the supply chain around it.

The lesson of the chip curtain, in other words, is not only that the world's most important supply chain is dangerously concentrated. It is that rebuilding it, or building the next one anywhere, is downstream of a question every educator and operator already knows: who is teaching people to do this work, and how fast.

That is the work SCM Education Solutions exists to do. Reshoring is, before it is anything else, a teaching problem. We will keep writing about the rest of it from there.

If a colleague, a student, or a program director would find this useful, forward it along. And if you are running a supply chain program and want to talk about how we are helping students earn the credentials this moment demands, just reply to this email.

Sources

• Semiconductor Industry Association and BCG, Strengthening the Global Semiconductor Supply Chain in an Uncertain Era, April 2021 (value-chain shares, the 92 percent figure, the "50 chokepoints" count).
• Semiconductor Industry Association and BCG, Government Incentives and U.S. Competitiveness in Semiconductor Manufacturing, September 2020 (48 percent of sales, 12 percent of manufacturing, down from 37 percent in 1990).
• Semiconductor Industry Association and Oxford Economics, Chipping Away: Assessing and Addressing the Labor Market Gap Facing the U.S. Semiconductor Industry, July 2023 (115,000 new jobs, 67,000 at risk, the 39 / 35 / 26 percent breakdown, the 3.85 million and 1.4 million economy-wide figures).
• China General Administration of Customs, reported by the South China Morning Post, January 2024 (2023 chip imports of 349 billion dollars versus crude oil of 338 billion).
• US Bureau of Industry and Security, advanced computing and semiconductor manufacturing equipment export controls, October 2022 and October 2023 updates.
• National Security Adviser Jake Sullivan, remarks at the Brookings Institution, April 2023 ("small yard, high fence").
• TechInsights teardown of the Huawei Mate 60 Pro / Kirin 9000s, September 2023 (7-nanometer class, SMIC, no EUV).
• The White House, Fact Sheet on the CHIPS and Science Act, August 2022 (52.7 billion dollars, the 39 billion in incentives, the 25 percent credit).
• Deloitte, Global Semiconductor Talent Shortage (more than one million additional skilled workers needed by 2030).