The Mysterious Microchip Shortage
When the press reports a microchip shortage, as it has done regularly since the beginning of the year, the economist does not believe it on faith. How can there be a shortage if market prices can freely adjust? A shortage is not simply a high price, for how useful would be a special word meaning exactly the same as “high price”? For the economist, a shortage is a situation where it is impossible to get something at any price, that is, by bidding up the current price. To cite just one textbook, see Arman A. Alchian and William R. Allen, Universal Economics, edited by Jerry L. Jordan (Liberty Fund, 2018), chapter 10. Since microchip prices are not, for all we know, capped by some domestic government under penalty of fine or jail, we need to find out what is happening.
At least, we need to ask the right questions. Let me suggest a few and propose some answers.
Of course, a temporary shortage in a segment of the market is not impossible, although it stretches and fuddles the meaning of “shortage” a bit. Perhaps I can say that there is a shortage of croissants at my preferred bakery every day at a certain time. But if I offered one million dollars (with a $100,000 deposit) for a croissant, the bakery would rapidly turn around or, if necessary, somebody would jump in a plane and bring me a croissant within at most 12 hours. Similarly, if your local grocery store is out of baguettes, you are facing a localized and temporary shortage of sorts—but only until the grocery supply truck returns.
In the same way, the so-called chip shortage is not across the whole market. Computer manufacturers, smartphone makers, and others apparently have no problem getting them, although they may have to pay more as they renew their supply contracts. And, of course, it takes more time to manufacture a microchip (typically a few weeks or months after an order) than to cook a croissant or deliver a baguette—the time dimension of shortages.
Another factor to consider is that up to 50% of the chips that car manufacturers need are non-generic, that is, they are not standardized commodities. The spot market for chips is limited but that does not prevent prices from rationing demand, if only in indirect ways through special fees for quicker delivery, for example.
What happened recently in the chip market appears to be that the car manufacturers’ demand suddenly increased last Fall as the economy recovered. This pushed up the price of chips and their components, including the ubiquitous wafers. Just over the past three months, chip prices have increased by an estimated 20%.
For one million dollars per chip instead of a few cents or a few dollars, a car manufacturer or a car part supplier could certainly find a chip maker who would be willing to pause its new contracts (usually signed months before delivery) and start producing for the new buyer as soon as it could. At such a high price some car companies would presumably be willing to sell part of whatever stock they have or to sell their own chip supply contracts with close delivery dates.
Car manufacturers aren’t willing to bid up chip prices to $1 million. They obviously calculate that, at current prices, another batch of chips would increase their marginal cost of car production by more than their marginal revenues from car sales. This is why many car manufacturers in the world have announced production cuts and are waiting for lower prices or later deliveries. Some car manufacturers reallocate chips from their less profitable cars to their more profitable ones. For a student of economics, this is a reminder that possibilities of substitution in production exist—in this case, of an input between two products.
Why didn’t car manufacturers order their chips or chip-containing parts several months before the recovery and stock them? Because they did not need them during the slump in car demand caused by the pandemic. Stocks are expensive—they use space, cost interest, and require management—especially in uncertain conditions when they might not be needed. When the demand for cars picked up last Fall, manufacturers all started placing new orders. In the meantime, manufacturers of computers and other consumer electronics had increased their own demand for chips as consumer demand for their products had grown due to lockdowns and remote work and education. The variety of the chip supply had changed (with some retooling) to meet the changed demand. (Car manufacturing typically uses about 10% of chip production.)
Car manufacturers probably realize now that it was an error not to stock chips or chip-containing parts while waiting for the recovery of car demand. Risto Puhakka, president of VLSIresearch, an industry analysis firm in Silicon Valley, says that it was a big “car industry failure.” On a free market, of course, localized errors happen—all the time. But profit-seeking firms have an incentive to correct them for the future. Automobile manufacturers will no doubt bring more attention to their chip procurement.
To summarize how this reported shortage looks from an economic perspective: If we say that there is a current shortage in the microchip market, we must immediately add that it is a localized and temporary “shortage.” It is localized in that it is limited to certain products—for which users do not want to pay higher prices, which does not make it a shortage at all. It is temporary because the production cycle takes some time during which supply is fixed.
Feb 22 2021 at 7:17am
I’d love your thoughts on the electricity price saga in Texas. My understanding is that prices were allowed to float pretty freely, being able to go up nearly 100x from the average price before the storm. Yet, on the surface, this flexibility didn’t seem to bring new supply to market or decrease demand to a level that was inline with supply. What went wrong? If prices were capped at a much lower increase, would we have expected to see thousands die from lack of heat instead of the estimated 70?
Feb 22 2021 at 10:20am
Dylan: I know nothing about the Texas electricity market. If it is as you describe, your conclusion is correct. For electricity as for BMWs, there can be no shortage if consumers are not prohibited to bid up the price (just as there can be no surplus if producers are free to bid down the price). Depending on supply and demand, however, the price can be too high for many people to want to buy. Now, note that immediately after a natural (or other catastrophes), prices can rise very high as it takes some time for “the grocery supply truck to return.” An “electricity truck” takes more time to move, hence the large temporary price jump.
Feb 22 2021 at 7:48am
How can you have chronic labor shortages in a free market economy? Labor shortages are the result of employers paying below market clearing wages to their employees. All an employer needs to do is pay a market clearing wage to eliminate a labor shortage.
Feb 22 2021 at 10:07am
Richard: Whether you wanted to be ironic or not, you have it exactly right. Employers who don’t know economics and thus speak loosely often complain that there is a shortage of labor: “I can’t find the workers I need.” What they actually mean is that they can’t find them because the wage (price of labor) they would have to pay is higher than the value of the new workers’ marginal product. In fact, they don’t want to find them. If they did, they would just pay more to get them. For example, if there is a shortage of warehouse employees (think Amazon) or department store clerks (think Walmart), the businesses who suffer the shortage will just bid up wages until quantity demanded equals quantity supplied. At $10,000 an hour, many engineers and academics would become greeters at Walmart. This is also why not every worker is paid the minimum wage.
Feb 22 2021 at 11:08am
I suspect the reason why auto manufacturers made planning errors in stocking chips was because they assumed the drop in demand they experienced was like that of a typical actual recession: a prolonged slump which would last for many years before returning to pre-recession levels. What actually happened a much faster recovery of demand.
I am reminded of a line from Steve Horwitz’s contribution to a collection of essays in honor of Leland Yeager: “changes in demand to not come marked with their cause”-indeed, and nor do they come marked with the duration for which they will last.
Market process and entrepreneurial discovery are key.
Feb 22 2021 at 3:15pm
Andrew: Good point.
Feb 22 2021 at 4:56pm
What actually happened a much faster recovery of demand.
Well, yes, and not only because the economic downturn was shorter and shallower than expected but also because consumers stopped spending on entertainment, dining, and travel, which left a lot of money for buying cars and houses. This is something was really without precedent as far as I know (easy to understand in retrospect but perhaps not so easy to predict).
Feb 22 2021 at 11:44am
I have worked in semiconductors all my career and I have discovered the best analogy for non-specialists are commercial aircraft. For leading edge chips (only some in automotive) like commercial aircraft there are just two manufacturers in the world, TSMC and Samsung (like Boeing and Airbus). At a very long time scale, requiring new factories to be built, the market will respond, of course. But that is very slow. In effect, if you didn’t order enough chips/planes about 5 years ago, then you are not going to get any because the factories to build them were not built.
Most automotive chips are not leading edge so much of this doesn’t apply. But many automotive chips are specialized chips built just for one or a few companies. If they were not ordered, they simply don’t exist, so no price will get you one. The reality is that chips take many months to manufacture, so even if capacity is available, you simply cannot order a chip (or a plane) today and get it next week.
So automotive manufacturers (OEMs in their jargon) are not cutting back production because they don’t want to pay double for their chips. Even if they paid a million dollars per chip, they still would not get any.
Feb 22 2021 at 3:15pm
Paul: Thanks for your comment, which I think only contradicts me on the following. Chips do exist because some just received their shipments (even if ordered 5 years ago) or will receive them imminently or have a stock. If somebody can make more profits with them than the owner, he will be able to offer a price that the latter cannot refuse. Prices allocate supplies to their best use even if supply is fixed in the short run. Note also that existing factories or production lines can be retooled depending on price expectations (as apparently happened during the past year). Of course, if the short run is the next 2 minutes, not much can happen.
Feb 24 2021 at 1:07pm
“If somebody can make more profits with them than the owner, he will be able to offer a price that the latter cannot refuse”. In some technical sense this might be true in the extreme. But how much do you think Ford would have to pay Volkswagen to shut down their entire production and give them all their chip inventory? It doesn’t seem a very plausible scenario. Of course, in the limit, Ford could buy Volkswagen (the company). But that is even less plausible. Even if from a financial point of view it might work, Ford is simply n0t going to buy Volkwagen and lay off all its workers in Germany (and clearly the German government would not let them anyway, which is not an economic supply/demand thing of course).
“Note also that existing factories or production lines can be retooled depending on price expectations (as apparently happened during the past year).” Not so much. Each process generation ends up with separate factories, and while it is always possible to manufacture old technology in a new fab, it is simply uneconomic. The other way round is impossible. The fabs are full. How much would you have to pay TSMC to break its current contracts with, say, Nvidia?
To give you an idea of the amounts we are talking about, a modern fab costs about $15bn to build (so it depreciates at about $50/second). The latest processes require EUV lithography machines only available from ASML for about $200M and backordered for years. So yes, you can always say that the shortage is “temporary” in the sense that if you paid enough money, maybe (and it is a big maybe) ASML could increase production by, say, 2026. Five years is “temporary” only in the sense that it is not forever.
I actually think it is worth looking at the problem split in two. First, what is the situation with the current fixed infrastructure (fabs, auto companies, semiconductor equipment vendors, and so on). Second, what could happen in the very long term if you consider changes (new multi-billion dollar fabs in a few years’ time, auto companies going out of business, and more of that sort of thing). There is enormous overcapacity in global auto manufacturing. But that unused capacity (half empty factories) is wasting money, but it is not using chips.
Mar 2 2021 at 2:39pm
Paul: That if somebody can make more profits with something than the current owner, he will be able to offer a price that the latter cannot refuse, is not only true in the extreme, it’s a near-daily occurrence. Firms not only sell brand names but shareholders and executives even sell their whole firms–everything.
Comments are closed.