Do you remember the global chip shortage? Somehow it seems so long ago, but it’s not even really been three years yet. Somehow, I had entirely forgotten about it, until two random mentions about it popped up in short succession, and brought it all flooding back like a repressed bad dream.
Playing the role of the ghost-of-chip-shortage-past was a module for a pair of FPV goggles. There are three versions of the firmware available for download at the manufacturer’s website, and I had to figure out which I needed. I knew it wasn’t V1, because that was the buggy receiver PCB that I had just ordered the replacement for. So it was V2 or V3, but which?
Digging into it, V2 was the version that fixed the bug, and V3 was the redesign around a different microcontroller chip, because they couldn’t get the V2 one during the chip shortage.
I saw visions of desperate hackers learning new toolchains, searching for alternative parts, finding that they could get that one chip, but that there were only 20 of them left and they were selling for $30 instead of $1.30. I know a lot of you out there were designing through these tough couple years, and you’ve all probably got war stories.
And yet here we are, definitively post-chip-shortage. How can you be sure? A $30 vape pen includes a processor that we would have killed for just three years ago. The vape includes a touchscreen, just because. And it even has a Bluetooth LE chip that it’s not even using. My guess is that the hardware designers just put it in there hoping that the firmware team would get around to using it for something.
This vape has 16 MB of external SPI Flash! During the chip shortage, we couldn’t even get 4 MB SPI flash.
It’s nice to be on the other side of the chip shortage. Just order whatever parts you want and you get them, but don’t take for granted how luxurious that feels. Breathe easy, and design confidently. You can finally use that last genuine STM32F103 blue pill board without fear of it being the last one on earth.
(Featured image is not an actual photo of the author, although he does sometimes have that energy.)
I remember the shortage, because today i found the reason, 25+ raspi’s in a box at my new work, so thats where the world wide stock went!
One silver lining to that chip shortage is it opened me up to the wacky and wonderful world of cheap Chinese 8051-clone microcontrollers, where 5+ companies make near-identical pin-compatible general purpose micros that all have many sourcing options… My favorite “brand” of these was HolyChip.
Can you short a pin for vorpal angoras or do you need SPI for that?
(1) is Russia still scrounging washing machines for chips for their T-80’s and T-90’s? (I understand neither side mostly doesn’t eveb bother to field MBTs anymore because you don’t need a $20,000 javelin, a $2000 drone is just as good), (2) I wouldn’t whistle past the graveyard just yet — the decade ain’t — between Xi and DJT and all the drama around companies like TSMC, Intel and Nvidia (and OpenAI) and all that I fully expect another breakdown in supply.
So, Russia’s BMPT “terminator” tanks are basically a bunch of close support weaponry bolted onto T-72 and T-14 frames.
They were supposed to be “indestructible” urban support tanks but apparently are (1) very expensive, and (2) not very good at their missions.
The use of washing machine chips was referenced a few times — I always assumed it meant they needed OTS MCU’s.
I don’t like war porn so I’m not linking to videos of them being destroyed by low cost drones but you can find a number if you look. By looking at older articles it seems like the Russians fielded about 10 of the BMPTs and lost at least 3 or 4 to FPV drones. Looks like maybe one more within the last week or two – that one was supposedly on a T-90 frame, but I’m not sure if that’s actually one of the BMPT terminators or if the Ukrainians are just calling it that.
https://www.forbes.com/sites/davidaxe/2023/08/13/a-swarm-of-tiny-ukrainian-drones-just-knocked-out-one-of-russias-nine-remaining-terminator-vehicles/
https://www.businessinsider.com/us-says-russia-using-chips-from-dishwashers-in-tanks-sanctions-2022-5
https://www.newsweek.com/russia-ukraine-tank-imr-3m-2133513
I never saw any real evidence that Russia was harvesting chips from washing machines anyway – yes the troops were stealing them from Ukraine but that was mainly because they were in awe at the western decadence of tarmac roads, indoor plumbing, and everyone having a TV & washing machine.
The idea you’d buy huge bulky appliances to desolder a $2 chip that you could probably source from your buddies in China seemed tenuous at best, also the idea that any significant military hardware was sharing a chip with a household appliance in the first place.
No nostalgia for the 2018 MLCC shortage?
cracks and shorts
I agree with the others: you’re celebrating too early. Look at what just happened with DJI, that’s a slippery slope.
COVID was just a taste of the disaster brewing. The US Administration has a campaign to ensure domestic production of everything is possible as a means to feel good about national security.
If this transition occurs poorly, it’ll be worse than COVID. Stocking up on TP today.
It’s not a shortage per se, but my low sodium diet is very much short on chips
So, what was the FINAL conclusion, will US set up and run its own chip factories or not?
I bet it was “let’s wait and see” and here we are again, setting up ourselves just for such thing once more to happen, because China eventually will try to take back Taipei/Taiwan with all the chip making factories and engineers included. They’ll just wait for the Taco Wind to pass over and make it happen, because they have patience.
I still do not understand what stops the US monsters (Intel, etc) from building their own US factories on scale comparable to Taiwan. The suppliers (say, ASML) are mostly european, so why don’t we return the 1980s?
It costs roughly twice as much to construct a fabrication facility (fab) in the U.S. compared to Taiwan. Estimates show that while a fab in Taiwan costs $10 billion–$12 billion to build, a comparable one in the U.S. costs more than $20 billion.
This higher cost is driven by more expensive labor, stricter regulatory hurdles, lengthy permitting and environmental review processes, and more expensive materials and utilities.
U.S. construction timelines are also roughly twice as long as Taiwan’s, due to permitting delays and the inability to maintain 24/7 construction schedules.
Once a fab is operational, engineer salaries and other operational costs are significantly higher in the U.S. than in Taiwan.
In 2024, Taiwan Semiconductor Manufacturing Companyproduced an estimated 12.9 million 12-inch-equivalent wafers, compared to Intel’s estimated 1.3 million to 2 million worldwide. This means TSMC’s capacity is at least 6 to 8 times greater, and the majority of the most advanced chips are made in Taiwan.
Intel is investing heavily to build capacity in the U.S., it is still playing catch-up. Intel’s new foundry unit, Intel Foundry Services (IFS), has struggled to attract major external customers and remains unprofitable, which hampers its ability to generate the cash needed for aggressive expansion.
Meanwhile TSMC is also aggressively building in the U.S., investing $65 billion in its Arizona facilities
I still do NOT get it.
Even if it takes twice as long and costs twice as much, it will be the US engineers’ salaries paid, moneys WILL return to the US economy while keeping talent/expertise local.
Chip-making is not just one industry, it is also a network of sub-industries, so in very real sense, it will returns things like well-paid engineering/technicians (not managers) jobs and as any Keynesyan Acolyte will tell you, give US economy an unexpected boost. In the long run it WILL lower the cost of ICs, and WILL return proper R&D.
Obviously, hiring H1B contractors to cut the costs will not work. It’s been tried, and the results are mostly disastrous. It will have to be US citizens.
TSMC has became a monopoly in its own (mostly unchallenged) field, so US is afraid of challenging TSMC monopoly? Why? What happened to the competition? Protecting offshore monopoly is more important? Invent newer/better practices that will NOT cost twice as much and require twice as long, then! Like NOT hiring top-heavy management drones and focusing on training/hiring the smartest MIT graduates instead? What happened to that?
I honestly do not get it.
It wont lower prices. When your costs are 2X, when your infrastructure is decades behind and will cost significant investment which must be recouped, the prices will be HIGHER to compensate, even with tariffs.
Its not a matter of fear of or protecting an offshore monopoly. There are players chasing domestic silicon. They are just too far behind the game and too small, EVEN with billions of government subsidy behind them, to make a significant dent in the lead TSMC already has.
The reason you arent running things your way is that you just dont get it.
“Even if it takes twice as long and costs twice as much, it will be the US engineers’ salaries paid, moneys WILL return to the US economy while keeping talent/expertise local”
That might be the way to think if you have any kind of central planning (whether it’s communism or WWII- era state capitalism). But in our neoliberal economy, it’s not anyone’s job to think about this stuff.
If a public company makes something in the US when they could have got it from Taiwan for $1 less, their shareholders will sue them. If a government tries to compel sustainable behavior (outside the defence sector), every corporate media outlet will work in lockstep to hound them from office. That’s why Lina Khan was out of a job no matter who won in 2024.
When businesses aren’t compelled, or even allowed, to work in the broader interests of any particular country, then you get an economy run for quarterly profits and nothing else. But it works great for the handful of people who matter!
Would you as a customer buy a terrible terrible smartphone that performs like its from 2004 because it was built ‘at home’? And almost certainly pay extra to have that ‘superb’ quality over the imported model?
So no the cost of IC’s won’t get lower for bringing it home, not even close unless you also artificially inflate the costs of import. So manage to entirely shut out the cheaper/better stuff ones and stuff made with them – those better established players that just make better chips than you can or match you for vastly lower prices will just be what 99% of the consumers buy, which makes the domestic stuff a really tiny niche production run so even more expensive than the mass produced better ones. Which leaves the niche domestic product versions something you only select because you are ideologically motivated, or have no other choice.
Also it is the more advanced of those sub-industries and potentially profitable but high investment R&D can be and usually still are kept at home as those are the the bits that its economically viable, even ideal to keep in the developed nations with rules and higher wage requirements. Generally the developed world has the better education systems (assuming you can get into them) and attracts the skilled and educated migrants largely because they already have larger established research facilities (For a relatively cheap example you don’t need to build a new windtunnel when you can more easily do that research in one already built with running hours available, so likely located in Britain mostly for the F1 teams now they are limited in windtunnel time). Which is why as I understand it the best lithography machines in the world are Dutch, probably the most loved chip architectures are developed in the West even if they get produced by TSMC etc etc.
What bringing the full spread of those sort of industries back locally really does its give your economy and military some security of supply, as while almost all fabrication happens in one place should anything happen to them or the supply chain between you… We all saw what happened with the Covid supply chain disruption, and that was really quite a small event in terms of real world impact on production potential, most just causing chaos in the supply lines.
RE: SpillsDirt, Bobtato, Foldi-One.
I get what you are saying, however, I’ll be honest, those all sound like excuses, not real reasons.
“…When your costs are 2X, when your infrastructure is decades behind…” why? It should be properly/STRATEGICALLY planned to that it won’t be obsolete by the time it is running full-speed. That’s what planning does, allows timely upgrades if/when needed. If one’s planning is short-term goal, then they should focus on selling something else, used cars, or legalized weed, or flipped houses. Those are completely different business models, actually, no real “models” to speak of, really, just “business”, ie, buy low sell high.
“…There are players chasing domestic silicon…” – yes, I personally know ppl working for one of those, and for now they mostly do military stuffs. Obviously, because of the tariffs pretty much HALF of their equipment, ordered and paid for, is still stuck in China. Because they are a tiny shop, and things like that cut their razor-thin budget even thinner. They literally CANNOT grow or hire anyone else now, until the get their (darn expensive) equipment shipped to the US. Catch 22. Military won’t like this, and today’s shutdown pretty much put on hold ALL the payments until further notice. Another catch 22. Have we had these all LOCAL, this won’t happen, but it did.
“…The reason you arent running things your way…” I am immune to the insults, but snowflakes may turn red from anything that even remotely reeks of “you are stupid”. That’s not what I implied, I implied that I see gazillion of greenbucks squandered on propping up for-profit Wall Street Casinos, and those could have been spend wisely investing into US production instead, with far better long-term returns overall, industries of stuffs RETURNED. Chip making is actually an amalgamation of industries, of which the final step, factories, is the tip of the iceberg. Until entire iceberg is returned to the US, there is no point in talking about the tip of the iceberg, and there no shortage of domestic capital, but there is shortage of figuring out how to invest it into growth.
“…That might be the way to think if you have any kind of central planning…” – no, and there is no need for “central planning”. There is a real need for de-centralized strategic DISTRIBUTED planning. If this is to fall into one planning entity, it will become Yet Another Public Housing kind of project. The whole point should be many smaller semi-independent entities, kind of like “liberal capitalism” is supposed to be, and NOT monopoly capitalism (where one entity amasses all the major/main funding – see certain Marvin Minsky and how he single-handedly diverted pretty much ALL R&D funding into AI to his personal projects only – yes, AI R&D funding, and witness the literal cambrian explosion of AI after he stepped away from the role – though, it also coincided with the Murphy’s doubling of processing power every year finally delivering the needed powerful CPUs, which sure helped, too).
I wasn’t exercising my (naive/incomplete) knowledge of planning, since I am not a CEO, but I do know what works – startup kind of entities/organizations, where horizontal, ie, expertise sharing, networking is as important (and sometimes MORE important) as vertical, ie, top-down, networking. AND – this is important – emergent networks of merit (ie, task groups of experts) should have the right to re-arrange/re-purpose their vertical integration; or, equally, move out of a vertical hierarchy into a different vertically-integrated network when they can. This is what “corporate restructuring” supposed to do, btw, prune the dead branches, while retaining the expertise/knowledge. In short, what’s needed are agile semi-independent (and funded) task groups/organizations, akin to early Apple (Wozniak, Jobs, etc) working together with similar task groups/organizations.
“…f a public company makes something in the US when they could have got it from Taiwan for $1 less, their shareholders will sue them..” – yep, Superior Court ruling “Dodge v. Ford Motor Co.”, one of the worse examples of court siding with short-sellers. No matter. Why? Because China’s federal government (and Japan’s Federal government, and South Korea’s federal government) don’t care about THAT ruling – they invest into their domestic chip-making factories, while we split each other’s hairs here.
While I do not like what China’s Experiment is doing (central planning/socialism, local capitalism economy), I acknowledge that they are BETTER at investing into their local chip-making expertise. Oh, and they’ll capture Taiwan the second Pentagon pulls out of the East China Sea, and gain the 50% chip-making in one swift move, while we are still stuck explaining each other why US cannot (or should not) build its own chip-making factories.
So that’s why “I don’t get it”. Not because I sound like I don’t know – but because I see the parts that don’t make sense.
Lol? Chip shortage was about high-end CPU’s, GPU’s, and ARM stuff.
You could buy simple chips in the period. And using different chips in one product is also not new, manufacturers do it since decades.
No you couldn’t. I couldn’t get a simple atmel chip for love nor money.
Plenty of chips blinked out of existence, or were only available to the big boys as the Automotive manufacturer as an example can just buy the entire available stock and next run of the chip in advance leaving nothing for the smaller consumers.
The general public probably only felt it in a few places, but plenty of scrap bin salvage diving or putting projects on hold till a suitable part can be sourced.
The US military could flat jump in line and commandeer entire upcoming orders, too. That was fun.
Mouser added the “XXX on order by date YY” thing in the chip shortage. That feature alone got them buckets more business from me.
Yeah, LOL no. The chip shortage entirely deleted some simple chips from existence because the high demand caused manufacturers to focus on the most profitable in demand chips. So some things went EOL fast.
Our designs from a few years ago were entirely built around chips with multiple footprint compatibility and multiple vendor sources just to avoid any issues as much as possible, and while yeah, this isn’t nee, at this level of adaptability it was.
We also had colleagues who had orders cancelled out underneath them because the US military commandeered the supply, and modules from other companies were modified even after the order was placed because they needed to change out chips that were no longer available.
Good way to demonstrate adaptability working those design skills especially under pressure. That’s why you guys (and gals) make the big money.
It’s a limitation, though – instead of choosing the best options, you choose the options that are “compatible enough” you can shift or adapt between part availability. It was even more annoying when the best parts for your application became discontinued in favor of more generic parts.
This has been happening in RF amplifiers for a while, for instance – chip companies consolidate, their product offerings are too varied and they just entirely drop certain designs even though (unlike digital chips) there aren’t really plug-and-play dropins. Broadcom years ago offed an entire series of just amazingly generically useful amps (the ABA-xxx63 amps) that have no functional replacement. Drives me nuts. The chip shortage accelerated that process a lot.
My friend bought a car from that era that was supposed to have heated seats, but they were removed because the controller chips weren’t available.
I promise you it wasn’t running on Pascal architecture.
In my project, I think I had a backlight driver, DSI-LVDS bridge, USB battery charge controller, and RTC go out of stock at various points. Possibly the audio codec too.
I remember the chip shortage a few years ago because I have to go past a yard full of inoperative pickup trucks. The GM plant in Flint didn’t stop producing trucks but they couldn’t finish the production due to chip shortage, all of those trucks were moved and stored in the field next to I-69.
The automotive sector was both one of the worst hit, also the least badly hit. The paradox: if there are 1000 ICs in a car, the whole production line can be brought down by any one of them.
So while they were able to step in front of many other industries, as mentioned above, they had such a broad exposure that some things had to give. No more heated seats. Return to analog/mechanical dashboard components in cars that otherwise would have had displays. Etc.
But we’re at the bottom of the food chain.
F’rinstance, [Ccecil] on Smoothie board: https://www.robosprout.com/smoothieboard-v2-final-specs-and-update/ Where Ford can whine about dashboard parts, these folks had to pick an entirely new microcontroller architecture and new motor drivers. For a motion control board. It’s like saying Ford had to redo their frame design or something.
Just wait until china goes for taiwan and the entire region blows up. With china, taiwan, south korea, and japan fighting a conflict, good luck getting anything.
Timeline is any time between 2027 and 2030.
All current conflicts are proxy wars fueled from outside (except one that is semi proxy war fueled by a madman). China, Japan and South Korea are not stupid enough to engage in such conflict any time soon. They enjoy this status quo (especially China) and use their political and economic influence to get what they want.
You are sorely mistaken. China is building, testing, and deploying hardware with one singular possible purpose: the invasion of the taiwanese island. With hong kong firmly under their control, it is the sole remaining standout. They pledged to be ready by 2027 and 2030 is seen as a cutoff by which point too much defensive hardware will be set up in the south china sea.
With american interests in taiwan for their chip production and american military bases and assets in japan and south korea, a wider conflict in the region is highly likely and contingent only on the US protecting it’s interests.
Lots of people think (and there’s reporting) that Taiwan has put in all kinds of scorched earth contingency plans against an invasion going against them.
Think lots of plants and facilities and warehouses rigged to be destroyed remotely.
Which would leave mainland China with a rather hollow victory and a lot of expensive smoking real estate.
Patience, intimidation and soft power seem like a much better strategy — they seem to have won EM over and if they can win DJT over then it’s a done deal without firing a shot.
Plus the US isn’t the only country affected by a Taiwan takeover. So they may put their two cents in as well.
My biggest frustration was trying to explain to my boss that, no they could not just spin up a new fab to provide capacity.