The chip shortage is coming and nobody's ready

I’ve been watching something build for months and I’m now fairly sure it’s going to become a major problem.

Car companies are quietly reporting that they can’t get enough chips. Ford, Toyota, Volkswagen. They’re cutting production. Not by a lot yet. But the language in the quarterly reports has shifted from “minor supply disruption” to “significant constraint.” And I think this is the tip of something much bigger.

Let me walk through why.

The supply chain you’ve never thought about

Every car made today contains somewhere between 1,000 and 3,000 semiconductor chips. Not just the fancy infotainment system. The engine control unit. The anti-lock brakes. The power steering. The airbag sensors. The tire pressure monitors. Each of those runs on a chip.

Most of these aren’t bleeding-edge chips. They’re mature technology, built on older process nodes (28nm, 40nm, 65nm), manufactured by foundries that also make chips for industrial equipment, medical devices, and consumer electronics.

Here’s the problem: the pandemic created a demand spike in consumer electronics (everyone bought laptops and monitors for working from home) at exactly the moment it disrupted manufacturing capacity. Fabs can’t spin up quickly. A semiconductor fabrication plant takes 2-3 years to build and costs $10-20 billion. You can’t just add a shift.

So when demand surged and supply stayed flat, the foundries had to choose who to serve. The car manufacturers, who had actually cut their orders in early 2020 (because they thought a pandemic would reduce car sales), got bumped to the back of the line.

Now they want their chips back. And the line is very long.

Why five companies matter

This is the part that keeps me up at night.

The entire world’s supply of advanced semiconductors comes from a remarkably small number of sources. TSMC in Taiwan makes roughly 54% of the world’s contract chips. Samsung in South Korea makes another 17%. GlobalFoundries, UMC, and SMIC make up most of the rest.

Intel makes its own chips in its own fabs, but they’ve been having manufacturing problems for years.

So when people say “chip shortage,” what they really mean is: a handful of fabs, concentrated in East Asia, can’t keep up with global demand, and there’s no way to fix that quickly.

Taiwan. A single island. 54% of the world’s chip manufacturing.

I keep thinking about what happens if Taiwan has a drought (which would affect the water-intensive fab process). Or an earthquake. Or if geopolitical tensions in the Taiwan Strait escalate. I don’t want to be alarmist, but the concentration of such a critical resource in such a small geographic area seems like something more people should be worried about.

The 18-month lag

The thing about chip manufacturing that most people don’t understand is the lead time. From the moment you decide you need more chips to the moment those chips roll off a production line, you’re looking at 3-6 months minimum. Building new fab capacity? 2-3 years.

The demand spike started in early 2020. The supply response won’t fully materialize until late 2021 or early 2022. That’s 18+ months of shortage.

And in those 18 months, every industry that uses chips (which is every industry) is competing for allocation from the same limited pool of fabs. Cars. Phones. Laptops. Game consoles. Medical devices. Military hardware. Industrial robots.

According to Reuters and Automotive News, some car manufacturers are already looking at weeks of production halts. This is September 2020. I think by mid-2021, this will be front-page news.

What I’m watching

Three things.

First, TSMC’s capital expenditure plans. They announced a massive expansion, including a new fab in Arizona. But that won’t be operational until 2024. That’s four years away. A lot of pain between now and then.

Second, the political response. Governments are waking up to the strategic importance of semiconductor manufacturing. The US, EU, Japan, and South Korea are all talking about onshoring chip production. But talking and building are very different things.

Third, the secondary effects. If you can’t get chips, you can’t build cars. If you can’t build cars, dealership inventories drop. If inventories drop, prices rise. If car prices rise, people hold onto their old cars longer. If people hold onto old cars longer, the auto industry contracts. And the auto industry is, directly or indirectly, about 3% of US GDP.

All because of some silicon wafers in Taiwan.

I might be overreacting. Supply chains are resilient. Markets adjust. Maybe the shortage stays contained to a few quarters of mild inconvenience.

But I don’t think so. The structural vulnerabilities are too deep and the demand surge is too persistent. I think we’re about to learn, painfully, that the entire modern world is balanced on a very thin, very fragile piece of silicon.

Stay tuned. I’ll be writing more about this.

Related thinking: