字幕列表 影片播放 列印英文字幕 This April, 19 million new cars were sold in America — the highest level since 2005. But these weren't just any old cars. Today's Chevy Silverados and Ram 1500s are not quite the same as those sold a year or even a few months ago. Something very strange is happening across the industry. Features, packages, even certain parts have suddenly disappeared from recently manufactured vehicles. An exceptionally observant buyer of a '21 Silverado, for example, might notice their truck getting about one less mile per gallon than was advertised earlier this year on the same model. Recent Peugeot 308 hatchback owners, meanwhile, may be surprised to find the digital speedometer shown on TV replaced with an old-fashioned analog version. And the dashboard display on the Renault Arkana mysteriously got a little bit smaller. It's as if they're playing a game: What tiny things can they change without anyone noticing? The actual explanation is a global shortage of semiconductors — the chips that power virtually every electronic device, from rice cookers to Xboxes. Earlier this year, auto manufacturers found themselves in an awkward position — the average car is made up of about 30,000 individual parts, yet they were missing one — worth perhaps only 4 or $5, one-hundredth of a percent of the final selling price. It didn't matter. Without one, single component, everything had to wait. At first, factories simply stopped producing and workers went home. Ford, Honda, Volkswagen, Audi, and GM all shut down at least a few of their production lines. Audi furloughed 10,000 workers, and Volkswagen predicts it will sell 100,000 fewer cars as a direct result. The total damage to the industry could be upwards of $60 billion, despite ordinarily spending only about 5% of its GDP on semiconductors. Eventually, though, many brands resigned to cutting back features, realizing the problem is here to stay. It's not just cars. Sony doesn't expect to meet demand for its PS5 until next year, Nintendo can't make enough Switches, and even deep-pocketed Apple, usually immune to such disruptions, has been affected. According to Goldman Sachs, semiconductors touch 169 different industries, and the worst-case scenario would be a 1% reduction in U.S. GDP. Given how strong the incentives are to alleviate the shortage, what on earth could cause such a costly disaster? Sponsored by Hover. Get 10% off the domain you need to turn your side project, personal brand, or startup into a reality. Early last year, demand for cars around the world plummeted. At the same time, sales of home electronics skyrocketed — routers, monitors, webcams, keyboards, iPads, and so on. Car companies canceled their semiconductor orders, consumer electronics brands ordered extra, and to the extent the two overlap, production lines shifted to producing for the latter. Automobile sales, however, jumped back much faster than expected — almost as fast as they had fallen in March. Already by September of last year, demand for cars was near normal, 2019 levels. Not used to such wild swings in demand, when Americans got back to their usual habit of buying trucks the size of small European countries, car manufacturers rushed to order more parts. The problem is that 'semiconductor' and 'rushed' are practically antonyms. There's nothing fast about producing billions of transistors on a single chip or maintaining conditions 100,000 times more sterile than a hospital operating room. When you hear semiconductors have extraordinarily long lead times, think not days or even weeks but rather months. Many, many months. A single chip takes no fewer than fourteen and as many as 20 weeks just to manufacture. After that, you can expect an additional 6 weeks for assembly, testing, and packaging. This, mind you, assumes the requisite machines are sitting around, ready to start the moment an order is received, and doesn't include delivery time. Ignoring these two giant caveats, each chip will take three to six and a half months from start to finish — longer than the entire dip in car sales last year. But if it were that simple, supply would've caught up with demand months ago. Instead, some analysts don't expect the current shortage to end until later this year, or, even next year. Something else is happening here. The semiconductor supply chain isn't just slow. It also involves many different parties. Steps which would be separated only by conveyor belts or departments in other industries are so complex for semiconductors as to justify entirely different, highly specialized companies. And, while efficient, specialization can also cause problems. Imagine, for instance, last September when Toyota saw its cars start to fly off the lot — it would've immediately adjusted its sales forecast accordingly. If sales the previous month were 18 million, it would've extrapolated a trendline and added some cushion — ordering, let's say, 20 million. But remember, Toyota's order doesn't just go directly to a factory. Instead, it might go to a company like Panasonic, which might be responsible for delivering the car's navigation unit. Panasonic, on the other hand, will see an order of 20 million units, extrapolate, and add some cushion — maybe ordering 21 million. It, likewise, may order from another intermediary, who, again, will extrapolate and add slack. By the time an actual manufacturer, or “fab”, receives an order, it may bear little resemblance to the original. Each step of the process involves an independent, profit-maximizing firm, making the most logical decision given its limited information. In aggregate, however, demand gets more and more distorted. Under normal conditions, this effect is known and manageable. But when orders come flooding in all at once, as they did in September, manufacturers are very easily overwhelmed. And because demand shot up everywhere at once, every company was effectively competing for the same, limited supply. Each one, trying both to minimize the risk of under-ordering and maximize its priority in line, inflated its order. For manufacturers, the difficulty is not so much solving the shortage — with lead time so long, there's really nothing that can be done except build more factories, which takes years. Their real task is to project confidence. If they can credibly convince buyers of their capacity, buyers stop inflating orders, and at least some of the shortage will naturally resolve. It's fundamentally no different from toilet paper, masks, or diapers. If even one person fears a shortage, they'll start hoarding supplies, forcing everyone, including the most rational, to play along. Shipping prices are at recent highs, and Chinese companies may be hoarding supplies for fear of U.S. sanctions, but no one truly knows how much is simply panic buying. In reality, the “shortage” may not be much of a shortage at all. What if there was a shortage, though? Supply may not be severely constrained now, but if this is what merely believing it to be, looks like, what would it, actually? The very nature of semiconductors all but guaranteed the dominance of a small number of companies. Opening a new factory easily costs upwards of $10 billion, and as much as twice that for the most cutting-edge. But because of how fast the underlying technology changes, it will very likely only remain useful for 5 or so years. That means burning through a million dollars every five hours. There's absolutely no room for mistakes or downtime. Factories run 24/7, 365 days a year, with no exceptions. With capital costs this high, it's simply not possible for new companies to dip their toes in the market. Over time, the number of competitors has dwindled, from 25 in 2002 to just three by 2016. The winner in each major chip technology revolution reinvests its earnings back, ensuring they dominate even harder in the next. With just a handful of competitors, the industry is among the most consolidated in the world — more so than airlines, biotechnology, or soft drinks. While not, strictly speaking, a monopoly, Taiwan Semiconductor Manufacturing Company is the clear winner in the contract foundry market — which produce on demand for companies like Apple and Toyota. That's a lot of very important eggs in one basket. Especially given that over 90% of TSMC fabs are located in one, politically fraught country: Taiwan. When semiconductors are flowing, no one pays much attention to just how amazing their existence is. But when they stop, when a single ship blocks the Suez Canal, or when an oil pipeline is held for ransom, the fragility of our modern world quickly becomes apparent. And yet, it's in part that very fragility which makes it, in all but the weirdest of times, so dependable. TSMC needs Taiwan, Taiwan needs TSMC, and at least for now, the world needs TSMC. What you and I need, on the other hand, is an online presence that we control. One of the first things I did when I started this channel was create a website. And to do that, you first need to reserve a domain name — a .com or .io or .app — which you can do with Hover for, in some cases, as low as $2.99. A website is the only place on the internet that you really, truly own. On Twitter or YouTube, we're just users at the whim of algorithms. Even if you don't need a website, Hover can also help you set up your very own email address — your name @ whatever you want .com, for example. Hover is upfront, reliable, and affordable, with only the best customer support. Reserve your domain with Hover and make your idea come to life with 10% off using the link in the description.