I used to think geopolitics was for politicians and pundits. Then I spent six months unable to buy a GPU.
The chip shortage of 2021-2023 wasn’t just “supply chain issues.” It was a lesson in how the world actually works. And if you write software, you probably weren’t paying attention.
One Island, 92% of Advanced Chips
Taiwan Semiconductor Manufacturing Company (TSMC) manufactures about 92% of the world’s most advanced semiconductors.1 This figure is for chips at 7nm and below—the cutting edge. TSMC’s market share for less advanced chips is lower, but the advanced nodes are what power AI training, high-end smartphones, and modern CPUs. The concentration is unprecedented. Not Taiwan the country—one company.
The chips in your MacBook? TSMC. The GPUs training ChatGPT? TSMC. The processors in your phone, your car, the servers running your code? Probably TSMC.
This is insane.
Imagine if 92% of the world’s oil came from one refinery in one country that another country claims as its own territory and regularly threatens to invade.
Oh wait, you don’t have to imagine it. That’s literally the situation.
The Straits of Taiwan
China considers Taiwan a breakaway province that will eventually be “reunified.” Taiwan has been functionally independent since 1949. The US has a policy of “strategic ambiguity” about whether it would defend Taiwan militarily.2 Strategic ambiguity means the US doesn’t explicitly commit to defending Taiwan, but doesn’t rule it out either. The idea is to deter both Chinese invasion (by hinting at defense) and Taiwanese independence declarations (by not guaranteeing protection). It’s a delicate balance that many argue is becoming untenable.
The Taiwan Strait—100 miles of water—is the most important shipping lane in the semiconductor industry. Pretty much every advanced chip in the world crosses it at some point.
If anything happens to Taiwan, the global tech industry doesn’t slow down. It stops.
Why Fabs Can’t Just Move
You might think: if Taiwan is so risky, why not build fabs elsewhere?
Chip fabrication plants cost $15-20 billion and take 3-5 years to build.3 TSMC’s Arizona fab, announced in 2020, isn’t fully operational yet in 2026. Intel’s Ohio project faced similar delays. These aren’t factories you can spin up quickly. They require:
- Thousands of specialized engineers (Taiwan has spent decades building this workforce)
- Ultra-pure water in massive quantities
- Stable electricity (fabs can’t tolerate brownouts)
- A supply chain of specialized chemicals, gases, and materials
- Extreme ultraviolet lithography machines from ASML (the only company that makes them)
TSMC’s competitive advantage isn’t just equipment—it’s institutional knowledge accumulated over decades. That doesn’t relocate easily.
“But Surely Someone Would Fix It”
Intel is trying to rebuild US chip manufacturing. It’s going… okay. They’re years behind TSMC and burning through billions.
Samsung in South Korea is the only other company that can make chips anywhere near TSMC’s level. South Korea, by the way, is technically still at war with North Korea. Cool.
The Netherlands controls the machines that make the machines that make chips—ASML makes extreme ultraviolet lithography systems. They’re the only ones.4 An ASML EUV machine costs over $150 million, weighs 180 tons, and requires multiple 747s to ship. The technology is so precise it uses lasers to vaporize tin droplets 50,000 times per second to generate the light needed to etch transistors at atomic scales. Export controls to China are a whole thing.
Building semiconductor fabs takes 3-5 years and tens of billions of dollars. You can’t just spin one up when things get tense.
Why Should You Care?
Here’s where engineers tune out. “This is politics, not my problem.”
I thought that too. Then:
- Cloud costs spiked because GPUs were scarce
- Hardware lead times stretched to 12+ months
- AI projects got delayed waiting for compute
- My laptop order took four months
The code you write runs on hardware. That hardware comes from somewhere. If you’re not thinking about where, you’re not seeing the whole system.
Systems Thinking at Scale
Engineers love systems thinking. Dependencies, failure modes, single points of failure. We spend hours eliminating SPOFs in our architectures.5 The irony is thick. We’ll redesign entire systems to eliminate a single database SPOF, but we don’t think about the single geographic SPOF underlying all our hardware.
The global chip supply chain has a single point of failure with 92% of advanced production, located on an island claimed by an authoritarian state, across a strait that could be blockaded.
If any of our systems had this architecture, we’d call it negligent.
The AI Race Makes It Worse
Here’s the part that keeps me up at night.
AI development requires ridiculous amounts of compute. The US and China are in an AI race. Both need chips. Most advanced chips come from Taiwan.
The US has implemented chip export controls to slow China’s AI development.6 The October 2022 export controls were described by some analysts as the most significant technology restrictions since the Cold War. They restrict not just chips but also the equipment to make chips and even US persons working on chip development in China. China is not happy about this. China has invested massively in domestic chip production but is still years behind.
Taiwan’s strategic importance just got higher. A lot higher.
What Can You Actually Do?
Honestly? Not much directly. This is nation-state level stuff.
But:
Understand the dependencies — Know where your hardware comes from. Know what happens to your systems if supplies are disrupted. Have contingency plans.
Watch the news differently — When you see headlines about Taiwan or TSMC or chip export controls, don’t skip past. This affects you.
Support onshoring — The US CHIPS Act is trying to rebuild domestic manufacturing. It matters.
Diversify where possible — If your entire stack depends on one cloud provider in one region with hardware from one supplier, you’re exposed.
Vote — This sounds preachy but the political decisions being made now about technology and trade policy will shape the next few decades.
The Bigger Picture
I started this by saying I used to think geopolitics wasn’t my concern. I was wrong.
Everything is connected. The code I write runs on chips made in Taiwan, shipped through contested waters, using technology from the Netherlands, powered by rare earth minerals from complicated places.
Pretending this doesn’t affect me is not neutrality. It’s just ignorance.
You don’t have to become a foreign policy expert. But understanding how the world works—the actual physical, political, economic world—makes you a better engineer. It makes your planning more realistic. It makes your systems more resilient.
And maybe, just maybe, it makes you a more informed citizen too.
Further Reading
- Miller, Chris (2022). Chip War: The Fight for the World's Most Critical Technology — The definitive book on semiconductor geopolitics
- Thompson, Ben (ongoing). Stratechery — Covers semiconductor industry in depth
- Kao, Jeff (2023). Taiwan's Silicon Shield — On how TSMC’s importance affects cross-strait relations
Related
- What Transformers Taught Me About Attention — On the AI systems that need these chips
- The Psychology of Code Review — On thinking systemically about dependencies
Changelog
- 2026-01-05: Initial draft
- 2026-01-29: Published. Added sidenotes with context on TSMC, ASML, and export controls. Expanded on fab construction challenges. Added Further Reading section.