The Crisis Nobody Saw Coming (Except Everyone Who Did)
The global semiconductor shortage of 2021-2023 exposed just how fragile the world’s most advanced supply chain really is. At its peak, the auto industry alone lost an estimated $210 billion in revenue due to chip shortages, according to AlixPartners. Lead times for some chips stretched to 52 weeks. A $0.50 microcontroller — the kind that controls windshield wipers or power windows — could halt a $50,000 vehicle on the production line.
How did we get here? The short version: a perfect storm of pandemic demand shifts (everyone bought laptops and gaming consoles), supply disruptions (factory shutdowns, shipping chaos), a drought in Taiwan (TSMC’s fabs need enormous amounts of ultrapure water), geopolitical tension, and a structural imbalance where the world’s most advanced chips all come from essentially three companies in two countries.
The Geography Problem
As of 2025, TSMC (Taiwan) manufactures roughly 90% of the world’s most advanced logic chips (below 7nm). Samsung (South Korea) makes most of the rest. Intel, once the undisputed leader, fell behind in manufacturing technology and only began catching up with its Intel 3 process (roughly equivalent to TSMC’s 5nm) in 2024.
The concentration is even more extreme than those numbers suggest. TSMC’s Fab 18 in Tainan, Taiwan, produces essentially all of the world’s 3nm chips. That’s one factory complex, on one island. An earthquake, a military conflict, or a prolonged drought could disrupt global technology supply for months.
This concentration wasn’t an accident. It was the result of three decades of consolidation, specialization, and the brutal economics of semiconductor manufacturing. A leading-edge fab costs $20-30 billion to build. The equipment inside — primarily from ASML in the Netherlands, Applied Materials and Lam Research in the US, and Tokyo Electron in Japan — can take 18 months to deliver. The talent to operate these fabs is concentrated in a handful of locations. The industry consolidated because the alternative was bankruptcy.
The CHIPS Act Response
The US CHIPS and Science Act, signed in August 2022, committed $52.7 billion in subsidies for domestic semiconductor manufacturing and R&D. By early 2025, the Department of Commerce had allocated the bulk of that money:
- Intel: Up to $8.5 billion for fabs in Ohio, Arizona, New Mexico, and Oregon
- TSMC: $6.6 billion for its Arizona fabs (a second fab producing 3nm was announced for 2028)
- Samsung: $6.4 billion for expanded Texas operations
- Micron: $6.1 billion for memory manufacturing in New York and Idaho
- GlobalFoundries: $1.5 billion for New York and Vermont expansion
Combined with private investment, the CHIPS Act has catalyzed over $250 billion in announced semiconductor projects in the US. Whether this reshoring actually reduces dependency on Taiwan is debatable — TSMC’s Arizona fabs will still be run by TSMC, using TSMC’s proprietary processes, with key materials and equipment coming from the same global supply chain.
China’s Response
The US has also been restricting China’s access to advanced semiconductor technology. The October 2022 export controls, expanded in October 2023 and again in 2024, effectively prevent Chinese companies from acquiring advanced AI chips (NVIDIA A100/H100/B200 and equivalents), EUV lithography equipment from ASML, and US-origin chip design software.
China’s response has been a national mobilization around semiconductor self-sufficiency. The Big Fund (China Integrated Circuit Industry Investment Fund) Phase III, launched in 2024 with a reported $47 billion, brings total government semiconductor investment to over $100 billion. Huawei’s HiSilicon division surprised the industry in 2023 with a 7nm chip (the Kirin 9000S) manufactured by SMIC, proving that advanced chips can be made without ASML’s EUV tools — at significantly lower yields and higher costs, but possible.
SMIC, China’s leading foundry, has been aggressively expanding mature-node capacity (28nm and above) even as advanced-node progress remains constrained by sanctions. Analysts project China will account for nearly 30% of global mature-node capacity by 2027, up from about 15% in 2020. The risk isn’t that China catches up in cutting-edge logic — it’s that they dominate the less glamorous but essential chips that go into cars, industrial equipment, and consumer electronics.
Lessons Learned
The 2021-2024 semiconductor crisis taught several uncomfortable lessons. First, just-in-time supply chains optimized for cost efficiency are brittle. The auto industry’s decision to cancel chip orders in early 2020 (expecting a prolonged demand slump that never materialized) then scramble for supply in 2021 was the most expensive forecasting error in industrial history.
Second, geographic concentration creates systemic risk that markets don’t price in. Insurance against Taiwan Strait disruption doesn’t exist because no insurer could pay out the global economic cost.
Third, reshoring is expensive and slow. A semiconductor fab takes 3-5 years from groundbreaking to production. The CHIPS Act projects won’t begin volume production until 2026-2028. The supply chain remains vulnerable during that transition.
The semiconductor industry learned the hard way that efficiency without resilience is fragile. Whether those lessons stick until the next crisis is the real test.