Things have changed quite a bit in the last two years with the automotive & industrial semiconductor market — from extreme chip shortage to inventory glut. I wanted to weigh in on our most recent economic news — tariff wars — while I’m still working on the next chapter of the Supply Chain Games series. (Yes, the last part I published was Part Five on cycle time in August 2023; no, I haven’t forgotten.) Part Six is really tricky to write, and I’m trying to figure out how to present it without going into excessive detail. So for today, I’ll just give a short bit of context, and why tariffs affect the situation. Hint: it isn’t good news.
Disclaimers
I am not an economist. I am also not directly involved in the semiconductor manufacturing process. So take my “wisdom” with a grain of salt. I have made reasonable attempts to understand some of the nuances of the semiconductor industry that are relevant to the situation, but I expect that understanding is imperfect. At any rate, I would appreciate any feedback to correct errors in this series of articles.
Though I work for Microchip Technology, Inc. as an application engineer, the views and opinions expressed in this article are my own and are not representative of my employer. Furthermore, although from time to time I do have some exposure to internal financial and operations details of a semiconductor manufacturer, that exposure is minimal and outside my normal job responsibilities, and I have taken great care in this article not to reveal what little proprietary business information I do know.
Photograph "’09" © 2009 kristina sohappy,
used under CC BY 2.0 license, desaturated from the original.
Underdamped Systems: What Goes Up Usually Comes Down
In Part Four I likened the cyclical semiconductor market to a tire swing:
What’s different, and a bit unnerving, is the mature-node situation. If you think of the semiconductor market as a bunch of tire swings, hanging from a big tree, regularly moving back and forth between overcapacity to undercapacity — look, there’s the one for memory and there’s the one for PC/mobile microprocessors, back and forth, there they go…
Well, one of the tire swings covers mature nodes, and it’s been pushed so hard it’s gotten stuck on a rock ledge or a tree branch.
So, the metaphorical tire has come back down, and many of the automotive and industrial companies have had decreases in revenue and increases in inventory while customers are working through their own excess inventory made during the panic of the chip shortage. I’ll quote CEO Kurt Sievers from NXP’s February 2025 earnings call:
Now on the inventory side. So the answer, first of all, Thomas, yes, we continue to digest on-hand inventory at our direct Tier 1 customers. That process continues, and it is both for the U.S. as well as for the European Tier 1 customers. How much it is and how long that still lasts, I can’t really tell you. And I would also not put Q1 into comparison to Q4 in terms of it’s less or more. No, we have individual plans there with each single customer, and we got to see how that plays out.
But that is clearly still weighing on our revenue performance, especially in automotive, keeping it at a sub-end-demand level. At the same time, and I — for completeness, I have to say that also the discipline on the channel inventory holds, you saw that we kept it flat at eight weeks in the fourth quarter for Q1, I had in my prepared remarks, something like eight to nine weeks. It’s a bit hard to say to pin it down exactly, but, say, eight to nine weeks. So, and in total, by the way, a reduction in dollars in the inventory.
So the important piece is we continue to do this from inventories, both at distributors as well as direct customers go down from a dollar perspective. So at some point, that must come to an end because the rate of under-shipment against end demand in a flat SAAR environment, that must come to an end. I gave up to try and call when exactly. But if you do the math, it cannot take that long anymore.
SAAR stands for seasonally adjusted annual rate, and the point is that people are buying cars at a relatively stable rate, so NXP thinks their automotive semiconductor shipments have been lower than the end demand, and therefore their Tier 1 customers are still drawing from inventory.
Haviv Ilan, the CEO of Texas Instruments, also mentioned the inventory situation in TI’s January 2025 earnings call:
So TI, yes, there was a lot of, I would say, a shock wave almost during COVID. It was not only supply/demand mismatch, but behavior of customers and almost anxiety of needing parts. So I think inventory was built. And I think it’s still playing out. I wouldn’t say that this asynchronous behavior across markets is done. We see it still across sectors in industrial. We see it differently between geographies. We even see it internally between our Embedded business and Analog business. So it’s one of these things — and I kind of hate to say it, but time will tell where we are.
Ilan goes on to put a spin on the fact that TI’s inventory buildup is “right” for the company as it comes out of this downcycle:
I am excited about the fact that now we have the right inventory level, the right supply to support growth. I’m encouraged about our progress in the analog market. As I mentioned in our prepared remarks, we did grow the Analog business on a year-over-year perspective in Q4. It was a couple of points of growth, but after about eight quarters of decline. And I think we — I look forward to continue to grow that business, especially when you think about the longer term. And most importantly, the secular growth in the market that we’ve mentioned, industrial and automotive, I’m convinced that it’s there. I’m convinced that our product portfolio is improving and strengthening. And we will have the right level of capacity and inventory to support our customers. So obviously, expectation, that I have for the team, is TI will continue to compete for market share and grow it.
Meanwhile the days of inventory of these two companies are high at the end of 2024; from the same earnings release and earnings calls:
- TI reported 241 days of inventory, \$4.5 billion, up 10 days from the previous quarter
- NXP reported 151 days of inventory, up 2 days from the previous quarter.
There are minor differences among the other automotive and industrial chip manufacturers — go see for yourself! — but we’re basically around the “Glut” stage of the semiconductor cycle. I can’t draw a tire swing very well, so I’ll just show a pendulum:
Inventory buildup or drawdown at the semiconductor manufacturers depends on whether supply exceeds demand or falls short, and pushes the pendulum towards glut or shortage, respectively. The obvious question here is, why a pendulum? What are the analogous components of inertia and gravity?
I talked a bit about the semiconductor cycle in Part Two; in case you don’t remember, here is Integrated Circuit Engineering’s description of the semiconductor cycle:
Swings in production growth rate are closely tied to capacity utilization, ASPs of devices and capital spending (Figure 1-2). For the industry as a whole, when capacity utilization is high, ASPs rise and companies are more profitable, which in turn, encourages capital spending. However, with increased spending, capacity constraints loosen and ASPs tend to drop, decreasing company profitability. The decreased profitability (pre-tax income) then reduces the amount of capital available to invest in future needs.
Historically, this has applied more to commodity ICs such as DRAM or op-amps, but many of today’s chips are proprietary in nature and are “stickier” in designs, so price competition is likely less of a factor. The culprit in 2021-2023 was backlog and inventory buildup, and I’ll talk more about this in Part Six. Either way, here’s my short take on the underdamped model analogy:
- “inertia” = delays in various business processes, such as from decisions of capital spending to actually build new capacity (about two years, as I mentioned in Part Two), as well as customer purchasing patterns
- “gravity” / “restoring force” = control of inventory, desire to match supply and demand
- There is also a positive feedback amplifying oscillation, from customers’ ordering strategies based on the presence of shortages or gluts — more on this in Part Six
The phases of the cycle go something like this:
- Tail end of the glut: there is plenty of supply, lead times are short, customers order what they need
- Economic growth occurs, customers need more, fab utilization increases, semiconductor manufacturer inventory drops
- Start of the shortage: Lead times increase, customers order more in advance, manufacturers are flush with cash and start making plans to expand capacity
- Shortage: manufacturers try to keep up with demand, as more capacity gradually comes on line
- End of the shortage: Lead times decrease, orders get canceled, further capital spending is halted
- Beginning of the glut: in extreme cases, customers have built up inventory, and use it instead of ordering more components
- Fab utilization drops, semiconductor manufacturer inventory rises
- Tail end of the glut: customers work through their inventory, and we start all over again.
And in “normal” times the cycle is somewhat manageable. But it’s been a while since “normal”. In 2018 we had a trade war; in 2020 we had COVID and the unexpected shift to home office / home entertainment throwing a monkey wrench into things. Both of these were disruptions that tweaked the pendulum, giving it a big push. Now what?
Tariffs, Again: The Dangerous Game of Chicken
Well, that brings us to tariffs again. This latest tariff aggression from the United States appears to make 2018 look like a tiny little spat.
There are two economic concerns to bring up with this year’s tariffs.
Politics: Free Trade vs Protectionism
One concern is the politico-economic argument of whether there should be tariffs or not. This gets into philosophies of free trade vs. protectionism. Economists generally seem to indicate that tariffs cause negative economic effects; here’s what economists from the National Bureau of Economic Research concluded in 2019:
Economists have long argued that there are real income losses from import protection. Using the evidence to date from the 2018 trade war, we find empirical support for these arguments. We estimate the cumulative deadweight welfare cost (reduction in real income) from the U.S. tariffs to be around \$6.9 billion during the first 11 months of 2018, with an additional cost of \$12.3 billion to domestic consumers and importers in the form of tariff revenue transferred to the government. The deadweight welfare costs alone reached \$1.4 billion per month by November of 2018. The trade war also caused dramatic adjustments in international supply chains, as approximately \$165 billion dollars of trade (\$136 billion of imports and \$29 billion of exports) is lost or redirected in order to avoid the tariffs. We find that the U.S. tariffs were almost completely passed through into U.S. domestic prices, so that the entire incidence of the tariffs fell on domestic consumers and importers up to now, with no impact so far on the prices received by foreign exporters. We also find that U.S. producers responded to reduced import competition by raising their prices.
and two economists in the Federal Reserve studied effects of the tariffs on the various states in the US, concluding:
We have exploited cross-state variation in trade exposure and related it to U.S. economic outcomes in terms of both employment and output growth during the past year. States that were more exposed to trade with the world have performed worse in terms of employment and output growth than have states that were less exposed. Although we cannot claim any causality effects in this analysis, our findings suggest that the trade war initiated by the United States may have had a stronger impact on the U.S. economy than what standard models of trade have found. Accounting for state heterogeneity is thus key in capturing the negative impacts of tariff increases.
What these kind of conclusions mean for ordinary people is not quite clear; independent groups such as the Tax Foundation are a bit more blunt about the 2018 tariffs:
The burden of higher tariffs will fall disproportionately harder on lower and middle-income households than on upper-income households. The administration has already imposed a \$42 billion a year tax increase on Americans, which will reduce after-tax incomes by 0.33 percent on average. Acting on the threats to impose an additional \$129.8 billion in annual tariffs would further reduce after-tax incomes, falling by an average of 0.92 percent.
Harvard Business School management professor Dante Roscini was also fairly blunt, saying the 2018 tariffs “are likely to backfire eventually, and American consumers will pay the price.”
Now in 2025 there are just as many warnings. (For example, see Tariffs Will Destroy the Best Cure for the Trade Deficit by Shannon K. O’Neil at the Council on Foreign Relations.) In fact, as I write this and try to find a variety of opinions, it is hard to find ones that are arguing for the tariffs with any sort of economic foundation, rather than just parroting sound bites. I’m not going to cite sensationalist news media of any inclination, but here are a few editorials that are from various institutions from right-of-center:
- Competitive Enterprise Institute: An economic, constitutional, and geopolitical disaster
- American Enterprise Institute: Tariff Formula Makes No Economic Sense
- National Review: Americans Will Pay the Price for Reckless Tariffs
(The Heritage Foundation appears to be the one major exception this year, and they’ve been fairly quiet about the issue in the last two months; in 2018 and 2019 they published several articles expressing strong opposition towards tariffs.)
I’m not an economist, but I do support the idea of free trade because we are in a global economy, and free trade improves overall economic efficiencies. We should find ways to mitigate the negative aspects of free trade such as poor working conditions and lax environmental standards — the fair trade movement has become well-known, even if it has had only minor impacts — but there don’t appear to be any easy answers. The nationalistic aspects of protectionism make no sense to me. But that’s just my opinion. I find myself in weird agreement with some economists with the Cato Institute that the 2025 tariffs are not a good idea — I won’t quote them, but you can read the article. And with the Wall Street Journal editorial board, which I would quote, but it’s behind a paywall.
In addition to direct economic issues — you can believe who you want to believe, I suppose — there are also concerns that increases in tariffs hurt the geopolitical standing of the United States among other countries; Professor Robert Gulotty of the University of Chicago stated in a recent interview that the tariffs “are being done in a way that ignores prior treaty commitments, combined with hostile rhetoric about national security and a seeming drive toward imperialism. Decades of diplomatic effort to negotiate these treaties, build trust and develop a system to address economic frictions have been abandoned.” Regardless of the eventual outcome — will the US get kicked out of the World Trade Organization? — the uncertainty is never good for facilitating business.
The second concern of this year’s tariffs is much more alarming, and is independent of whether you think tariffs are good or bad — namely the speed and magnitude of their disruption.
Thwack!
Investor’s Business Daily posted this graph in an article on April 4:
2018 barely shows up as a blip on the graph, and it was more gradual in nature, increasing in smaller steps over the course of about two years, as illustrated in this graph from the Peterson Institute:
The reason the speed of the 2025 tariffs is so alarming, is that there is no way that supply chains can keep up with it. Raise the tariffs by 1% a month over two years? Okay, maybe there’s a chance for the system to respond. But a huge step change is just not something that anyone can cope with. The stock market has taken a steep nose dive in the April 3 and April 4 trading sessions, anticipating uncertainty.
So far the tariffs have exempted semiconductors, but those are coming, allegedly “very soon”.
Again, there is no way for the system to respond in a matter of days or even weeks. TSMC’s factory in north Phoenix, Arizona took about three and a half years from breaking ground in April 2021 to starting high-volume N4 production in the 4th quarter of 2024. And that was after years of preparation and planning; according to the South China Morning Post, a delegation from the Arizona governor visited Taipei in 2013. A follow-up visit from the Phoenix Mayor and Economic Development Director in 2019 apparently kicked off discussions anew, and in May 2020 TSMC announced they would be building a factory in Arizona. For European or Asian semiconductor manufacturers, starting chip manufacturing in the US — even if it did make economic sense — isn’t like flipping on a switch; it is an expensive and lengthy undertaking that requires enough capital resources to allocate billions of dollars over several years before getting even one cent of return. None of the automotive / industrial semiconductor manufacturers are in that position today.
Transfer of chip production between existing fabs (say, from Europe or Taiwan to the USA) isn’t quick either. It’s not like TSMC has a bunch of branch offices scattered throughout the world to offer portability of the same fab processes — most of TSMC’s 300mm fabs are located in Taiwan; the exceptions are Arizona, which covers 4 nanometer and newer, Fab 16 in China allegedly supporting 28nm and 16nm nodes, and the Kumamoto fab in Japan which covers the 40nm - 12nm range. If you have a chip on TSMC’s 65nm or 90nm processes, it’s coming from Taiwan. As far as redesigning an IC to run on a different company’s process: well, that takes time… less time for purely digital chips, more time for designs that include analog processing or need automotive qualification.
What about switching the semiconductors in a design from manufacturers in Europe or Asia (NXP, ST Micro, Renesas) to a different chip from a US-based manufacturer? Because of the “stickiness” of proprietary ICs in a design, and the time it takes to complete firmware development and testing, this is an expensive and lengthy process that can take several years; the embedded systems industry seems to have finally caught up with the disruption from COVID-19 to work around supply shortages.
Even if you can imagine a situation where the semiconductor industry adjusts into some bizarre new reality that satisfies nationalistic concerns, persuaded by the new calculus of US tariffs, and sells chips “Made in the USA” to American companies that then sell electronics to American consumers — why would any company do so, when the tariffs could disappear just as fast as they arrive? Supply chains do evolve; we’ve seen shifts of many manufacturing industries away from the United States, from furniture to shoes to steel to semiconductors. But they do so gradually, in response to economic conditions that allow manufacturers to thrive most successfully. Again, setting up a high CapEx factory is an expensive and lengthy undertaking — whether it’s a wafer fab, or a packaging plant, or an electronics assembly plant — and it won’t happen on political whims that have no long-term guarantees.
So don’t expect any significant tariff workarounds. Exacerbating the situation are last-minute panicking efforts to bring finished products into the US to evade tariffs, like Bitcoin mining machines from Thailand.
And that’s not even considering the increased risk of recession and what it means for electronics manufacturers.
As to the semiconductor industry cycle: well, let’s take a look. Here we are in glut land.
Now we have severe tariffs looming ahead. Increases in prices for products with imported components are likely to hit end consumers, which means lower demand is likely. (This is simple Microeconomics 101 on the price/demand curve: Lemonade Stand from Part One all over again.) That’s a continued push further into the “Glut” end of the cycle, making the amplitude of the pendulum’s swing even larger.
My prediction is that in the next two-to-three years, when it comes back the other way, we’ll have another shortage. Maybe not as severe as 2021, but disturbances that tend to increase the amplitude of the cycle have long-lasting effects. I hope I’m wrong. Of course, with shocks this severe, anything can happen.
The toilet paper shortage in 2020 was a good predictor of the dynamics of a supply chain shock due to COVID, so during the next few months of 2025, look for what happens in other industries with shorter time lags, to get a sense of how bad things might get.
If you are concerned about this situation, and you live in the United States, you might decide to call your Senators or Congressional Representative and pass along those concerns, and perhaps they, in turn, will let this administration know how dangerous and destructive this game of chicken really is.
Meanwhile, enjoy this spring and go outdoors to get a break from society’s ills. I certainly need it.
© 2025 Jason M. Sachs, all rights reserved.
