Toyota was one of the leaders in just-in-time manufacturing, yet they're doing just fine with the chip shortage. They stockpile parts, and try to understand how they work in depth.
> After the [Fukushima] catastrophe severed Toyota’s supply chains on March 11, 2011, the world’s biggest automaker realised the lead-time for semiconductors was way too long to cope with devastating shocks such as natural disasters.
> That’s why Toyota came up with a business continuity plan (BCP) that required suppliers to stockpile anywhere from two to six months’ worth of chips for the Japanese carmaker, depending on the time it takes from order to delivery, four sources said.
> The sources said Toyota has another advantage over some rivals when it comes to chips thanks to its long-standing policy of ensuring it understands all the technology used in its cars, rather than relying on suppliers to provide “black boxes”.
This is really ironic given the fact that Toyota was really at the cutting edge of 90s JIT manufacturing and this was one of worries that was raised in the. Their response, "we have multiple sources for parts." [0] But over time those multiple partners migrated to the same geographical locations, and then an earthquake hit.
[0] I'm totally paraphrasing here as I remember reading it in some magazine years ago.
It certainly is ironic, but it is also an interesting case study for the many, many idiots who keep applying LEAN manufacturing as a dogma.
Look! See! Toyota isn't dogmatic about everything being Just In Time. They are only doing it because it is the best option at the time and it works. If it isn't the best option (or doesn't work) then they stockpile like a clever person.
There are a few people who really need their noses rubbed in to this little example.
Dell was amazing at this 10-15 years ago. They had less than 24 hours of production volume of harddisks (and other components) on their own shelves, but forced their suppliers to have them ready. Which resulted in the suppliers building warehouses close to Dell manufacturing locations, essentially offloading some of the inventory costs upstream in the supply chain. They could do that because suppliers were competing for Dell order volume. And Dell to their customers never specified the brand of the components just generic specs like 7200 rpm, 100 GB.
They famously maintained production after that earthquake. It's a common case study of how to adapt to supply chain shocks by being good at running a supply chain in the first place. All the things they needed for resilience -- plentiful data, high trust relationships with suppliers, flexible and skillful workforce -- were things they had built up to support their JIT manufacturing capability.
It's not so ironic, you can do both. What differentiates what Toyota is doing today from what Toyota did before TPS is, without understanding where the bottlenecks were in their manufacturing operations, the stockpile sizes were uncontrolled and thus only represented waste. Here, lean manufacturing principles still apply, albeit with an intent to pay a certain amount of cost to maintain a stockpile of a certain size, for BCP purposes.
Yes. No buffer is bad, and too much buffering is also bad. It's about understanding all the parameters that determine the optimal buffer size given the goals, and maintaining that optimal size.
At least they had it planned, and very well planned. And they still understand what JIT really meant. The actual idea and concept is now only known to Apple and Toyota. For various sort of reason ( Mostly Politics ). Others are only applying what they think they know to Supply Chains.
Most on HN would likely have little exposure to supply chain, take a look at another example coming from Toyota that anybody should have at least heard of it in Software development.
Remember Kanban board?
Holy mother of God that great idea quickly turn into a shit show.
That has always been the core innovation of JIT. Offloading inventory costs onto suppliers makes your books look a lot better. Toyota did make a bunch of other improvements to streamline their manufacturing processes, but the just in time part is really just bullying suppliers into taking on some of your cost and risk.
Partially - suppliers can offset that by increasing their cost to do business; but also most contracts in that world are multi-year deals with ranges of minima / maxima for quantity, time etc.
It also means you can customise your product closer to the delivery time. The vast majority of parts are the same between dozens of final variations of the cars. Having low final product inventory, but plenty of parts stockpiled, means if demand shifts between your final variations you can respond to that quickly while also still being resilient to sudden increases in total demand.
Having the supplier store them still uses the same distribution network as if they were shipped directly. If Toyota stored them, they'd need to develop an equally robust network from their warehouses to their factories. It's more expensive to maintain your own distribution network and warehouses than to just pay the supplier to store stuff
To be more specific, it is cheaper to maintain warehouses for the products your company produces and more expensive for products your suppliers are producing.
1 chip warehouse per supplier vs 1 chip warehouse for each customer.
Toyota is an expert in storing cars, it is not an expert in storing thousands of different car parts.
I had the same question, I'm not versed on manufacturing processes but "Just in time" would seem to imply inventory is nearly 0 and as parts come in, they are used for assembly and go right back out the door.
Perhaps they realized the limits of JIT and have looked at storying inventory based on the impact to assembly continuity?
> storying inventory based on the impact to assembly continuity?
It's always been about that.
JIT can be looked at as, "right part in the right quantity at the right time." That can mean "milk runs" where assemblies are delivered to the final assembly plant twice a day. An AM delivery where the last assembly is bolted to a car as the PM delivery arrives. Or it means "kanban orders" where 1 week supply of parts is delivered on a Monday and a refresh the next Monday. Or it means you buy in bulk and get 3 months supply at once.
JIT and everything around "Lean manufacturing" and the "Toyota Production System (TPS)" are ideals to strive for. Not everything in Toyota operates according to the ideals. There are certain fundamental things (e.g., stability in material, personnel, machinery, and methods) that need to be in place before starting to reduce inventory.
When suppliers have trouble with providing defect free parts, the orders start to increase in quantity and decrease in frequency. These can be subject to 100% incoming inspection if the supplier's quality comes into question enough.
From Toyota perspective yes -- they can maintain appropriate kanban stocks in their individual factories and still pull as needed.
Sounds like the suppliers might be "strongarmed" by Toyota purchasing power.
Except if Toyota's business is worth the cost of keeping 2-6 months of product in the warehouse, it isn't really like an "unneeded resource" that the JIT approach would want to avoid. They are being paid (in a sense) to hold those parts, so it is not wasted.
Thanks for all the responses, makes a lot of sense. It seems like the answer is “Yes, from Toyota’s perspective” in that Toyota gets to focus solely on the manufacturing process, and let the others figure out the chaotic bits to get the required parts in the right place at the right time, because that’s a different skill set that certain other corporations are far more adept or have specific resources like rolling stock.
To use an analogy: a surgeon operating on a patient, holds a hand out and says “scalpel” and a scalpel instantly placed in hand. Does it matter that their are trays and trays of parts and other people in the room if we are just evaluating the surgeon’s efficiency?
I think the article is implying that they moved away from a "pure" JIT model, to somewhat of a hybrid model. Their in-depth knowledge of their suppliers is orthogonal to their stockpiling, but it still helps them plan better.
This is a bit of digression on this thread, but it's stuff like this that makes it seem like traditional car companies are still out to lunch vs Tesla. They're just too slow to adapt and give up what has worked for decades.
Isn’t it their first plug-in hybrid? Seems like they could be testing the waters and ended up with way more demand than they anticipated. I guess we will see in a year or two if they can ramp up.
Toyota’s blunders with electric vehicles has nothing to do with its supply chain and everything to do with its strategic decisions. The Rav4 Prime has a long wait list because they chose to make a small quantity of them.
They partly choose to make so few of them because they’re still trying to make hydrogen happen.
That's backwards. It would had prevented the shortage.
The shortage is happening because manufacturers cancelled their chip orders from their suppliers because of reduced sales forecasts due to COVID.
When the manufacturers cancelled their orders from their suppliers, the supplies cancelled the orders from the foundries, which cancelled their slots and gave them to consumer electronics which had increased demand.
When sales of new cars rose, manufacturers called their suppliers for more parts, which called the foundries to increase production only to find out their previous slots had been sold.
Well also the climate induced power outages in Texas knocked out three (4?) foundries in the Austin area from suppliers that tended to supply embedded customers. Foundries really really hate to be interrupted and getting them back up seems to be taking longer than anticipated.
I think I know what you're getting at: If all of the manufacturers suddenly requested private inventories to be set up, the chip manufacturer would be dedicating all of its production to filling up warehouses rather than shipping.
Most comments I've seen on this and other related articles on HN are too focused on the high-end chips doing fancy stuff. There are not only supply constraints on these sorts of high-compute chips, but also big shortages on practically all ICs in general.
I'm a hardware engineer for a medical device company, and we've been dealing with supply constraints not only with our MCUs but other ICs like high-side power switches, LDOs, memory, and more. It's tough when we're low volume (a few thousand a year) and the huge automakers and other huge consumer electronics giants are gobbling up all the parts.
We recently tried to get 150pc of the same uC that's on the teensy 3.6 - no chance. Luckily we weren't bound to the exact model and could substitute for a pin compatible model with different flash sizes and without a certain feature.
Then I had to design a board because a simple 5 to 12 V boost converter from TI became unobtainium.
We are a prototype shop and do quantities of 10 and less most of the time.
I saw Paul S (who makes the Teensy) posted recently that he’s having trouble sourcing some components.
I’m projecting that a hobby project I make is going to be delayed by not being able to get the LDO I worked into the design. I’m contemplating switching out the component for something a little more common. (AP2112 for an LM1117 for anyone who’s interested)
It's sad seeing the price inelasticity causing a decreasing productivity in this way. I hope this is a lesson for all future considerations combining JIT with price inelastic goods
Do medical devices and auto tech use similar components? I understand there's a lot in common, but I would think that safety constraints for one domain would cause it to not overlap with the other.
Forgive my ignorance on this, as I am but a humble software engineer who is forever in amazement at how all ICs are basically flattened rocks we shoot lightning through to make it do math real dang quick...
Often yes. "Medical devices" covers a large array of products from in-vitro/small implantable to things similar to laboratory equipment (think patient monitoring) or industrial (MRI and the like, autoclaves, etc). "Safety" is not really achieved on a per-component basis, but as a sum of the parts.
As for other comments in this thread, depending on your device, it can see a great deal of thermal stress (steam autoclaves) or vibration (MRI), and not all components in autos are experiencing the total vibrations from internal combustion engines. I've typically designed in components with extended temperature ranges, most commonly 85C+ or even 125C. This increases the reliability and extends the life of your product.
"Automotive-grade" typically means extra lot testing, and some parameters may vary in the datasheet due to the way the parts were validated by the manufacturer, even though they're technically the same part. For instance, I am using a Texas Instruments part that became nearly impossible to source. I had the automotive version in my bill of materials because at time of the design, it was the only version out. An alternative made by TI looked like a good fit, but a few key parameters in the datasheet didn't quite line up. It was because the automotive was tested at like 13.8V input whereas the newer "industrial" part was tested with 24V and slightly different loads. Same exact component, tested slightly differently.
Similar but not the same. There is an insane variety of nearly identical MCUs, for example. For an otherwise identical part, you can often get it in half a dozen or more packages. (i.e. the chip has different physical form factors in terms of how it attaches to the PCB, pin type and spacing etc... but the actual die inside is the same) Then there are subtle variants for a given MCU (differing types/amounts/speeds of RAM/Flash/interfaces etc.) The matrix of variants quickly gets out of hand. It's probably easily 10-50x (depending on the part) the number of variants Intel/AMD come out with each year. The autos also typically use parts with an extended temperature range.
For the subtle variants, I believe it's how they price discriminate. It's all the same part, but they use fuses to disable various peripherals. Way easier to do that than to fabricate a whole bunch of different chips.
Most likely when it's moderate/major features. Sometimes it can be as simple as a case of packaging (i.e. fewer pins so they can't bring everything out.) I've always assumed that some of the packaging options were driven by large customers who wanted something just a bit different for whatever reason. (PCB space, power consumption, just to be difficult... who knows)
In addition to the other good comments regarding medical/automotive overlap, is that often those chip lines have long lifecycle commitments built into them - ie. we promise to make this set of parts for at least 10 years. For both both of those customer areas, not having to redesign your boards and recertify devices on an external timeline is of value.
One of the most common additional requirements on automotive parts is extended range, often below -20 C and above 70 C. That is often not relevant for medical products which operate only inside, near room temperatures.
After much confusion as to why the healthy looking child, and that no I wasn’t rushing her to the ER, they finally figured out that for head scanners don’t work right at 20 degrees.
The solution of course was to have someone sit on it when not in use.
Would you kindly confirm your story involves an outdoor rink? (My mind conjures a miserably cold indoor rink and I’m shuddering; the story as told is unnecessarily confusing because the ice rink detail is immaterial, the important detail is using the device outdoors in cold weather)
And then there's also the vibration issue. I don't think most medical devices need to stand up to continuous vibrations like you'd have in an auto engine. If they are not soldered well, eventually things can loosen up and fall off, or minimally just break contact. I think the car industry is closer to the aero/space industry than medical industry.
It took 20 years but the circuitboard on my Nissan's dash had deteriorated enough to warrant resoldering. Now that is done, it is as good as new. It's remarkable how dash clusters look inside, absolutely pristine behind the plastic lens in spite of their environment.
Aero has a few unique issues that make car parts unusable. One is the lower bound of the temperature range, it's not unusual to see -55 celcius for an aircraft. And pressure, air is really thin at the 41,000ft most large aircraft are certified for.
Since January of this year, I have been waiting for a certain MCU (STM32F042K6T6) to be available but still no luck. I get the bigger picture now: there's a shortage at the source. I have to move fast and secure a stock of the rest of the ICs, otherwise this supply-chain issue can be disastrous to small hardware companies.
I'm convinced that, if in-person IEEE meetups were still a thing right now, you could drop a tray of ST Micro chips in the middle of the meeting and watch it devolve into a fistfight in seconds.
+1 for having to be sealed.
We have a handful of companies wanting to sell us needed chips but they are odds and ends, not always packaged. Were always suspect.
I think thats two factors. One, the analog chip manufacturers that deal with power ICs have been consolidating, and two the power outages in Texas hit a number of foundries that tend to focus on embedded parts.
For a run we're doing of 800 boards that use 8 commodity ICs, 3 of those ICs are having significant supply issues right now.
Just spent a fun 2 days soldering a bunch of 0.1mm wires to boards and adding test sockets so we can sort out which supplier to pay 4X the previous price to vs which ones sent bogus or old oxidized stock.
Hey, if you need help finding replacements I might be able to help - I've literally been doing mostly this for customers since January. It's absolute bullshit and I've never seen this before in all my time in this industry.
It's funny you mention this, a lot of the stress in Automotive right now is about these random IC's as well. CAN transceivers, and other dumb supporting HW is in very short supply. Less so the high-end silicon
> What has made the auto industry particularly vulnerable is its reliance on just-in-time delivery, where parts are brought in when needed, rather than being stockpiled.
It's fascinating that the response is to close production rather than increase prices.
Those warning of inflation point to events like this as support. But inflation requires that higher producer costs be accepted by consumers.
And for that price transmission mechanism to work, there needs to be supply actually available at the higher price. It appears that just-in-time economics mean that in the event of a shortage your supply just goes offline. You don't get higher prices, just empty shelves.
It sort of does help; chip prices have risen because customers are out-bidding each other. If you're willing to pay 10x 2019 prices you can probably get whatever chips you need. Many companies are deciding that's not worthwhile.
>If you're willing to pay 10x 2019 prices you can probably get whatever chips you need. Many companies are deciding that's not worthwhile.
The problem as I understand it is a company like TSMC sells capacity for the some time period - like a year. The automakers thought people wouldn't buy cars, so they relinquished their slots. Turns out car demand exploded, so they went back to TSMC and said "hey we need those slots" and TSMC said "sorry your slot already got sold to nvidia".
Chrsyler could offer to pay TSMC 10x more, but that that ultimately wouldn't do much, the slot is gone and they would have to buy the capacity from nvidia. nvidia is in no position to sell because they are also facing extreme demand. How would it look to nvidia if, while gamers across the world can't get their hands on 3080s, they sold their fab capacity to Chrysler for $$$$?.
I think your simplifications could change the logic here.
Chrysler does not directly buy from TSMC, they buy from i.e. TI, Renesas, Microchip, NXP, ...
Many of those companies do still have fabs, and even if not, they are not competing with Nvidia. They are producing on the old, old fabs, at 40, even 80nm. We would need better numbers on where these microcontrollers are fabbed to be able to tell how this interferes with i.e. Nvidia and Apple on TSMC 5nm.
I'm not even sure it's strictly speaking still true that automotive only uses outdated fabs. Tesla seems to have disrupted the use of low-end chips at least in high end cars, like Jaguar Land Rover.
Change "probably" to "definitely" and I'm with you. Automotive chips need to be a lot more reliable than a PS5 GPU, and that kind of reliability takes time to test and prove. Not to mention that automakers try not to change parts unnecessarily because their economies of scale are only as useful as their ability to use one component across several models and years. If you care little about thermal efficiency and power consumption, and you're not doing extremely abstract computations, there's no reason whatsoever to spend tons of money on a smaller lithography. I was looking into this recently, and the price difference between some of the equipment involved is around two orders of magnitude from the leading edge to the trailing edge.
Not even 80nm. Most stuff in that space is 130nm+, some extremely old, but still produced ICs from early nineties are on 300nm+ for use on equally old automotive parts.
If you look at the concurrent user numbers on Steam, it seems like there's overwhelming engagement in the gaming space right now. I thought the same as you until I looked into it a bit more.
(Also, bitcoin is mined on ASICs nowadays, but I get what you mean.)
I've never heard of other crypto currencies referred to as Bitcoin by anyone other than noobs. It's like calling the Mexican Peso a Dollar. The semantics are important - i.e. for doing something like currency conversions.
Are we noobs here or is this a tech forum where people might reasonably be expected to know the difference between Bitcoin and Ethereum? I'd argue it's the latter.
You can't say that Bitcoin is responsible in a discussion about GPU shortages. It's flat-out wrong and obfuscates the actual problem. It's Ethereum that's mostly responsible for said shortage, and seeing as how Ethereum is moving to proof of stake soon that problem might be alleviated. Bitcoin uses all custom hardware for mining and is not affecting the GPU shortage. The problem with Bitcoin mining is power consumption, not chip contention.
That will be actually fun. Make a company invest in some random crypto coins, presents movements. Just call all of them bitcoin... Argue that at this point it is generic name... Skim your fees from top.
Chip shortages are at a point in the US were government contractors who already have right by law to jump ahead of anyone else in line when ordering and can probably pay whatever price they want are still being told that new shipments won't come for 16 months.
Yes. It’s really a testament to the praise of Tim Cook as a master of operations.
It’s quite a statement that Apple launched four variants of the same product (M1 Air, M1 Pro, M1 iMac, M1 iPad) in the middle of a supply chain apocalypse, all within a few weeks and is both delivering on schedule and even putting the product on sale.
Call your HP rep and try to buy an LCD monitor right now.
10-15 years ago, Cook's level of competency wouldn't have been anything special, but that of competent operations lead.
The industry in the West not only shipped everything to outsourcers, but even forgot how to manage outsourcers as well, being fully reliant on turnkey services.
To be fair, this only works because it is difficult to make money off of protein folding. It is kind of a waste, but not wasteful enough to criticize it.
It's not correct to think that increased prices solves shortages. I work in automotive chip industry. We have many cases where no amount of money could fix the shortage within the next 9 months. Only choices for carmaker is to produce fewer cars or produce the car without the chip that is short (which is not always possible).
There are contracts that might prevent that from happening, just like lead times don't disappear, even if you outbid everyone else (if it was a bidding issue).
Throwing money at a problem to try and solve it can be impossible in various ways.
Yeah, but -- though I haven't run the numbers -- it seems like Jaguar would be better off taking smaller profits while overbidding for the chips rather than just go idle and make something. Even selling at a loss may be preferable to the huge damage to brand from such a discontinuity. (Remember how hard the major internet sites work to avoid downtime.)
It's not just the normal customers, though. A predicted shortage caused a run on chips by people looking for a profit. I've been working on a PCB design for months in which I specifically picked a chip to design around that was newish, in current production, and plentiful at all the major suppliers. I got my first test run done, made a few changes, ordered a second run, which I intended to populate myself, and there were exactly 4 chips left at a single major supplier. Yet, Octopart suddenly lists dozens of new suppliers I've never heard of.
If you have a stockpile of your supplies you'll notice that you can't replenish them long before actually running out and you can therefore increase the price of your product to extract the highest amount from whatever limited volume that you can still produce. Except of course when you are working off an order backlog, but then emergency procurement in an empty market will be even more ruinous.
I wonder if it's possible to retool production lines to keep going without the chips, then add them later. It increases WIP but we saw this in the 80's with computers; they soldered sockets for those chips then only needed to snap them in to complete the machine.
At some price point, you could probably substitute the chips with FPGAs? Though I'm not sure if that would work for cars without requiring some sort of recertification.
The empty shelves will cause cost-push inflation, as people try to buy cars from a smaller available pool (e.g. substitute new cars for used). The price mechanism for cars will respond to a shortage of cars.
Edit: it’s maybe bad luck that this could coincide with a demand-pull inflation pressure, as people ease back into personal transit.
Jaguars And Range Rovers aren't "I need a car" cars. They are optional luxury goods. The lack of new Range Rovers and Jaguars won't keep anyone from getting to work or the store. So I don't think the effect will be nearly as drastic as it would be if the number of Camrys and Hiluxs had to be curtailed.
Wait, so you think if a Veblen good becomes more scarce there will be no price increases around it?
won’t keep anyone from getting to work or the store
I can tell you that luxury vehicles transport plenty of people to work or the store just fine, and if they break down, written off, or are stolen they require replacement just the same.
> It appears that just-in-time economics mean that in the event of a shortage your supply just goes offline. You don't get higher prices, just empty shelves.
If the demand is there, the price increase will find its way to the market, either in the resale market or in alternatives to Jaguar Land Rover.
Exactly. Look at what is happening in the user car market for now. Used cars that people bought a few years ago are appreciating because the new ones are slow to hit the dealerships.
Also cars are interchangeable, if a consumer is not able to buy a model of a luxury car, they can always buy another one that might be higher priced.
Except it takes weeks/months to prep a fab to start making these chips. If the car companies start getting their orders then switch back to JIT manufacturing, then the fabs won’t want to waste these months idle thrashing the manufacturing lines across products.
Raising prices only helps if you can get more supply at a higher price, and you can’t when it takes months to start.
> You don't get higher prices, just empty shelves.
Trying to buy on the used market right now. I promise you we are seeing higher prices because people who had 30k to buy new, are now spending more for the higher quality 1-3 year old used stock.
You have empty shelves, workplaces start going in person again and demand for cars becomes less elastic. Potentially used cars start selling for more than they cost new.
I bought a used car last year for 13k after taxes and fees. 9 months later, I sold to Carvana for 16k (even with a few scratches that would have been $300 to fix, plus a $200 scheduled service coming up in a few weeks). Carvana needs to make a profit, so presumably they're going to mark it up further. It's a crazy world right now.
Carvana is the Internet's BHPH lot. They don't care how much they spend to buy the car because they're gonna sell it for 50% down with 20% APR to someone who can't get financing otherwise.
Used cars went up in value last year. Rental car companies sold their fleet cars into a better market and did not acquire new cars. This left rental car companies short on supply and able to raise rental car prices to compensate as people travel again.
A couple of weeks ago, with seemingly no large events occurring, all rental cars in San Antonio were rented. This week, they are available, but at higher prices than I would normally expect.
I'm even seeing stories about U-Haul trucks all being rented out by tourists because rental cars are unavailable, causing a shortage of rental trucks for movers.
Interestingly enough, Toyota modified their supply chain operating model post-Fukushima. So the solution isn't as coarse as a binary "stockpiles or JIT", but an intelligent assessment of risk and managing that risk granularly. Kaizen at its best.
"Toyota may have pioneered the just-in-time manufacturing strategy but when it comes to chips, its decision to stockpile what have become key components in cars goes back a decade to the Fukushima disaster."
"After the catastrophe severed Toyota’s supply chains on March 11, 2011, the world’s biggest automaker realised the lead-time for semiconductors was way too long to cope with devastating shocks such as natural disasters.
That’s why Toyota came up with a business continuity plan (BCP) that required suppliers to stockpile anywhere from two to six months’ worth of chips for the Japanese carmaker, depending on the time it takes from order to delivery, four sources said."
"“Toyota was, as far as we can tell, the only automaker properly equipped to deal with chip shortages,” said a person familiar with Harman International, which specialises in car audio systems, displays and driver assistance technology."
I was under the impression it was Japanese cars' superior quality/reliability to price ratio that caused them to take over the industry a few decades ago. Just in time might allow for lower prices, but would it also have resulted in the higher quality?
One way to look at it is that JIT is about reducing or eliminating "waste." A quality improvement program is also about reducing waste because reworking a bad part is wasted time and effort compared to making the part correctly in the first place.
So it's not that JIT resulted in higher quality, but an emphasis on quality across the board likely lead to JIT being implemented.
Chips become obsolete very quickly, and car makers create new models every year or two. You are not going to keep a year worth of inventory. Not convinced maintaining more stocks would have bought them much more time.
That's really not true, Car parts are heavily interchangeable. Many parts are used across generations of cars, across different models in a make's lineup, and even across makes.
It takes them 5 to 6 years to design a new model. The chips they use usually come with a guarantee from the manufacturer that they will supply it for the next 15 years.
And I believe the automotive engineers are much happier to keep using old proven parts if there is no reason to replace those. Unlike software, where every new thing will have to be done with something different.
I don’t think chips for cars become obsolete so quickly.
Well, self driving hardware excluded, but the other stuff can probably run very well on 10+ year old designs.
Exactly even those will have same the chips as the regular cars. Just because you have big expensive chip and stuff for self-driving doesn't mean the other systems doesn't use standard off-the-shelve solutions like ABS and so on and on.
People just think the CPU or GPU, but what about power supply, peripherals, busses? All sorts of auxiliary stuff...
I don’t mind patiently waiting for a new Xbox but if Microsoft suddenly doubled the price I’d strongly consider patiently waiting for a PS5 instead.
I’m guessing whatever companies expect they could gain (or even retain) from raising the price in the short-term doesn’t exceed what they expect to lose in the long-term.
I have been looking to buy shocks for my motorcycle for months but the cost was very high and I wasn't sure I could spend the money. But every few days would check out the shocks online. Then I started noticing the shocks I wanted started selling out everywhere. You can not buy them on ebay even. I literally found the last place with a set and bought them. I did so because I noticed that after everywhere sold out the manufacturers website has suddenly increased their MSRP by $50. The place I found them at had them at the old price so I pulled the trigger before they went up.
The model your are using is to simple. Since there is a shortage of new vehicles it pushes up the price of used ones. Then any supply of new vehicles that hasn’t been sold yet goes up. People will keep cars longer since the price of new vehicles has gone up. The drives higher repair rate of vehicles and parts prices will start to rise. Since there is a chip shortages you may not even been able to get certain parts new so it drives up the price of used parts containing the chips.
As people put off buying new cars when cars start becoming more widely available the prices stay higher until the shortage worked out and then prices of used and new car drops. Manufacturers are likely to overshoot the number of new vehicles since the models they are using assume the increased demand. This means once it is worked out a significant drop in prices and you can get vehicles really cheap. See 2008 cash for clunkers where it caused a spike in used cars price temporarily. It also had the effect that leases where cheaper since manufacturers believe they would be able to sale them used at a higher price which was only temporary causing them to lose money.
You do get increased prices for new stock that don’t have existing contracts. If you screw up your estimate of what you need and your suppliers don’t have the capacity to make your parts, now you have a shortage. Now you have to shutdown because you can’t get parts but a manufacturer who did a better job of estimating will be able to charge a higher price. In this case it would be Toyota which didn’t cut their chip orders when the crisis hit. Of course they are also likely to have some shortage of certain vehicles since there will be a shift in demand for their vechiles. So if you own work trucks the value of them will skyrocket right now since most of them in the US are made by the big 3 which don’t have the chips to manufacture them.
Inflation means some things become economically impossible to produce, that is why the rest of the stuff becomes more expensive.
Prices always lag behind because they are the result of all the economic processes that take place, not the cause, and this is just the beginning of the process.
It's worse than that - the auto companies reduced their orders because they predicted covid meant that their sales volume would drop - turns out they were wrong, and now they are bitching and moaning because they can't get the stuff they didn't order
> What has made the auto industry particularly vulnerable is its reliance on just-in-time delivery, where parts are brought in when needed, rather than being stockpiled.
In another version of my life I was a truck-driver bringing parts to Ford/Chrysler/GM/Honda/Toyota on a daily basis. I've done this for many trucking companies, and the warning/threat that you get at 'orientation' (aka training for 15 minutes) is the same for all of them:
GM Charges us $24,000/hour if we are late for our window by more then 15 minutes.
I'm sure the other automakers had similar threats, but I only ever heard it about GM.
“Other automakers purchase much less valuable silicon content, and become less of a priority when compared to Tesla, who designs chips in house, secures wafer supply from foundries directly, and buys chips directly from the various chip designers like NXP, Infineon, and so forth,” according to a note from Cho Research. “They don’t outsource the design of their chip stack; they in-source wherever possible and work extremely closely with their suppliers.”
Tesla designs mostly FSD related chips. They buy tons of other of the shelf chips like any other manufacturer.
They have stopped S/X lines, their highest profit margins models by far, for refresh. But that’s been already delayed by many months, and it’s not expected to start production anytime soon. Their metal moving volume sellers got couple of price increases in a short time.
One thing we know is that we cannot trust Tesla not lie about their production issues. During model 3 ramp up they were repeating how great they’re doing. And recently Musk acknowledged that it was such a chaos, that they were few weeks away from bankruptcy.
While Tesla doesn’t admit they are affected by chip shortage, if it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck.
I’m willing to bet Tesla runs into this in just a few months. This is coming in waves across the industry, and I doubt Tesla is equipped to deal with it any better than any other manufacturer.
(Full disclosure: I work for an Automotive OEM. I do not have any inside knowledge about the shortage in general, or how it’s affecting my company. All my comments are from my own observation of public news sources.)
Well, maybe. Here is the thing, during the peak of the pandemic Elon Musk stated that he sees everybody else reducing orders, they were not gone do that. They said they will continue to go full board ahead and put in orders expecting continuing growth over the next year.
The chip makers specifically called out the some OEM for instantly reducing orders and then instantly demand more again. Tesla didn't do that.
That doesn't scale. The reason Tesla's not having problems is because they have first dibs. That only works for them because the companies who don't have first dibs are taking the hit, not because this is some solution to the issue of shortages on the macro-economic level. If everybody did what Tesla did and in-sourced, somebody who in-sourced would be taking the hit.
Just because Tesla and Toyota did something that ended up right doesn't necessarily mean all the other auto manufacturers did something wrong. I don't the blame the other auto companies and think their actions to be prudent at the time. We have a global pandemic for the first time in 100 years where people were required to not travel and stay indoors. Seems logical to expect demand for automobiles to go down and cut back on ordering chips. Turns out reality was the exact opposite. Certainly something I wouldn't have been able to predict.
That is exactly the situation they are in. They have plans/contingency plans to make sure they don’t end up in a similar situation while the other manufactures don’t. Their plans paid off and the others’ bet not on having the plan...well that also paid off.
Is there a good article around that explains the specific semiconductor bottlenecks everyone is facing right now? I’d like to know more. It seems like a huge opportunity, but also I am guessing it’s a very hard or impossible problem since it hasn’t already been solved by infinite money?
When the pandemic started, most companies thought demand will go down for things like cars, etc so didn't place orders for semiconductors needed for things like cars. Instead demand went up for devices used at home like laptops, webcams, etc. When companies order component, they don't buy on demand, but they order far ahead. So you might say something like, I want to buy 10million units of X in 6 months, or next year. The manufacturers operate at peak capacity already so they can't crank up output on demand. You give them enough time to be able to meet your needs. The companies that ordered say 1/3rd or 1/2 of what they needed realized they need more since the pandemic didn't slow sales down much. They went to order some more components and the manufacture says, "sorry! lead time is now 12 months" because they already got contracts with other folks. This of course is not a one supplier to one product problem, but possible across a chain of supplier/producer relationship. The organizations that saw there might be a supply problem earlier on, aggressively stocked up and bought up more than they needed which placed additional strain across the supply problem. The supply problem ends up cascading across industry. This is know as the bullwhip effect - https://en.wikipedia.org/wiki/Bullwhip_effect. No one wants to invest in new plant because by the time it's done, the issue might be resolved and even if it's not, the plant might not return the ROI. That's how we ended up where we are.
Thanks for bringing up the Bullwhip effect. Do you know about any format where one can get insights into the current state of the supply chain world? I don't get why podcasts and tv channels talk about all and everything 24/7 but never manage to pull up an interview with top tier supply chain/logistics/purchasing manager people. Why isn't this something that should be publicly debate?
this is why I roll my eyes at anyone trying to deride the stimulus bill etc about worries about inflation. The supply shortages are temporary as expectation of lower demand did not come true....it is not a fundamental limitation of supply capacity...truly our capacity to supply products can be ramped up enormously and fairly quickly.
Fabbing older chips is not a particularly fruitful business for a startup, and the established players have their hands full. Some face water shortages. Others have suffered major factory fires. And all would need several months at a minimum in order to bring on more capacity.
Part of the problem is that when the supply chain is long and fragile, a perturbation anywhere affects customers everywhere. Kulicke & Soffa has said that they're having trouble getting chips for their machines. That's problematic because they are a semiconductor equipment manufacturer.
The result of shortage is that lead times get longer (not unlike how packets queue for a lossy channel), incentivizing chip customers to order more than they need. This, in turn, further constrains the market.
The trailing edge lithography scanners don't get up to the 180 tons that an ASML EUV machine might weigh, but they're still large and you can't exactly get them with Amazon Prime two-day installation.
Acer and AMD said the shortages would ease this year, but TSMC says they will continue to next year and Intel said yesterday that they think it will take a couple of years.
I have friends working in the physical side of the chip industry- there is a lot of money being poured into the industry right now to increase fab output, but it takes time to build the physical infrastructure. Chip fabs are highly hazardous environments and need to conform to pretty intense standards.
And of course, every time anyone on their specific team within the fab gets COVID-like systems, the _entire team_ goes home for two weeks. This has happened several times over the past year.
You'd think that a fab would be one of the least likely work places where one would pick up covid. The air is constantly filtered, you're wearing bunny suits with face coverings and the density of people is pretty low.
This fab is in an ultraconservative city where masks and social distancing are not practiced. The workers have been getting COVID from the community, not at work.
But if I have covid and you and I are both wearing those weird space suit looking things (I don’t know what they actually are called) you would think the odds of me giving you covid would be minuscule right?
Not everyone at the plant wears a full suit. A lot of the workers are sitting in offices most of the day monitoring systems, or are working on supporting systems like chemical pipes, or loading and unloading trucks.
And then everyone sees each other outside of work anyway. Some of the people at the plant are family, or as close as family. Some go to the same church. Some go camping or fishing together on the weekends
The summary there seems to be "Insufficient investment in 200mm wafers". Combined with everyone's demand forecasts being screwed up due to COVID and being either too high, or too low. Finally, you can't just move from FabA to FabB (even for the same wafer and process size) quickly, so customers can't just quickly migrate to whatever fabs ended up with excess capacity.
A lot of people seem to have learned economics 101 but not control theory, and seem to think that the economy will respond to signals instantly rather than having a finite frequency response.
The high level of uncertainty doesn't help; how sure are you that there won't be a pandemic related demand crash in six months?
In past shortages, that demand crash is precisely what happened. Considering how long a new fab takes to pay for itself, and how thin the margins are in fabbing trailing-edge chips that have been around for years, it would be a major risk to lay out more capex than your road map from 2019 suggested. Meanwhile, if you just keep doing what you're already doing, utilization is high and prices are rising so you're guaranteed to turn a rewarding profit. There's not a major private-sector incentive that would justify costs whose amortization would require consistent demand for several years.
- Digital transformation from Covid (increasing demand for chips)
- "Double booking" feedback loop (companies preorder more than needed, fearing fab capacity limits --> makes it seem like there's less fab capacity --> companies book more capacity, fearing capacity limits)...this is not specific to Covid, but heats up in this kind of situation
Even infinite money can't set up new fabs overnight - factories are tough. Unless you're a government paying for the strategic advantage of domestic manufacturing, you have to decide whether the plant will still be profitable when the shortage ends - and whether it's currently profitable enough to merit building.
It's not unlike the situation with PPE manufacturing circa March 2020.
Yeah I guess I'm just surprised. It feels to me (who knows basically nothing about it) that it's been a pretty big problem for at least a few years, and it also seems like the combined buying power of the entire world's hardware companies would make the investment in new fabs trivially worth it.
Maybe fabs take longer than a couple years to set up regardless of money spent?
Maybe fabs take a level of expertise that only a handful of people in the world have, and it's a matter of ramping up promising undergrads to that level which will take like 15+ years?
Maybe analysts expect demand to even out so that capital outlays right now don't make sense?
But all of these are surprising to me to the degree that we still don't have enough semiconductors. I'm just hoping for a sort of comprehensive overview of the issue, if such a thing exists.
> Maybe fabs take longer than a couple years to set up regardless of money spent?
Maybe fabs take a level of expertise that only a handful of people in the world have, and it's a matter of ramping up promising undergrads to that level which will take like 15+ years?
From the start of permitting to first tape out, I've heard five years for cutting edge, three to four for mid range (under 100nm), and two years for older processes. Subtract a year for expansions of existing facilities where hazmat permits and infrastructure already exist.
Far more than a handful of people in the world are qualified to set up a fab but it really is hard, time consuming work that requires a lot of expertise. While most of the equipment in a fab is customized off the shelf (not totally custom but not cookie cutter), each piece has to be carefully calibrated to fit in with the whole.
Stuff like HVAC, which is normally pretty predictable in commercial buildings, has to be custom designed for seasonal variations which differ from region to region while taking into account each piece of equipment's heat generation and stability needs. Tens if not hundreds of different robotic positioning and material handling systems need to be calibrated so that they can move wafers entire meters with a precision and accuracy measured in the tens or hundreds of nanometers. The design of the factory even needs to take into account regional variations in day to day humidity which usually takes a year just in data collection.
Building the concrete shell for a fab is easy. It's filling it with equipment that actually works together to produce cutting edge technology that is the expensive part.
Right! I get it for the very high end chips used in CPUs and GPUs that are in demand for home computers and cryptocurrency mining are constrained by a very limited amount of fabs, but what is holding back the older process chips?
There's a confluence of things affecting the older-process, more commodity chips as well:
- Demand for electronics as a whole, at all scales (from computers to cars to IOT to industrial controls and all kinds of other things that we don't think about) has been shooting up for the last decade (without compensating expansion in fabrication capacity). This is the 3rd serious supply shortage we've seen in electronic components in the last few years (anyone else remember seeing MLCC lead times hit 100 weeks in 2017-2018?).
- All of these supply lines running all the way back to the raw silicon operate with very lean (or no) supply buffers. Think "suppliers making deliveries directly to the factory floor" lean. These sorts of systems are not resilient, and do not respond well to shocks (like suppliers having to shut down due to a global pandemic).
- As has been commonly noted, there have also been demand shocks, where electronics consumers (especially car companies) who are normally operating on timelines forecasting out 6-24 months forecast reduced customer demand, reflected that in their orders, and then had to adjust back when demand for cars came back.
- Similarly, with people locked down and working from home, demand for electronics both for professional use and as a replacement for entertainment options has expanded.
- Finally, there are rumors (that I haven't been able to confirm in a meaningful way) that some large Chinese manufacturers have been stockpiling components in anticipation of further tariffs, sanctions and trade tension.
> (anyone else remember seeing MLCC lead times hit 100 weeks in 2017-2018?).
I did, even Chinese domestic market still has months long passives shortages even today.
> - Finally, there are rumors (that I haven't been able to confirm in a meaningful way) that some large Chinese manufacturers have been stockpiling components in anticipation of further tariffs
Actually plain speculations, some times by people very far from semiconductors market, just like with masks, baby formula, apartments, pork, cement, aluminium, stocks etc.
> Actually plain speculations, some times by people very far from semiconductors market, just like with masks, baby formula, apartments, cement, aluminium, stocks etc.
I can’t speak for semiconductors but it seems like supply chains everywhere are hurting. I work in bio and it used to be you could order something and it would arrive in 2 days. Now literally every product I order is backordered 2 weeks to 2 months. Even innocuous things like centrifuge tubes or common chemicals!
I don’t know any chip manufacturers that shutdown due to COVID. Everyone has been shipping the entire time and fabs are at capacity. The way chip manufacturing works is that you get a slot and if you cancel they fill your slot with something else. There have been fab issues but these have been unrelated to COVID and would have happened anyway.
To me this seems like auto industry trying to shift blame away from their supply chain management to their vendors.
Texas Instruments has a facility that has had output reduced due to COVID staffing issues for at least the past 6 months. They produce parts that I use in my 1000 pc/year medical device that Toyota and GM vacuum up at a much higher rate. Hard to compete against those heavyweights for available stock. I'm sure there are other manufacturers encountering the same. This isn't just the fancy very-small nm processes, it's effecting the whole industry.
Has there been any detailed reporting on the root cause of the fire? It's strange that an electrical fire can start in the first place in what should be an extremely well-engineered factory, let alone spread to over 6000 square feet, and if that's possible there, it'd be great to learn more to avoid it elsewhere.
What I couldnt find out so far is what the chip manufacturers are producing instead of automotive chips then? There must be another industry which has a significantly increased need or will get a lot of chips earlier than planned. Wondering which industry that will be.
Basically everything we use? Many many things have electronics inside that is chips and as demand have moved from services to things there is more things bought, thus more demand. Then add to this some distortions... These are chips in question, not that there is not also demand on high-end. But that is also makers learning that it's better for them to constrain supply. RAM for example was at one point very cheap.
AMD growth is gone through the roof in general computing, and they're the sole supplier for both PS5 and Xbox's CPU and GPU chips. Unlike Intel who have their own fabs, AMD is locked into TSMC - and when auto manufacturers relinquished slots, they went in.
Nvidia ships a boatload of chips for Nintendo Switch plus GPUs for almost all altcoin miners that can be mined with GPUs and everyday regular gamers.
Car manufacturers released all of their slots at the beginning of 2020 so those were sold to the entertainment industry instead. Turns out there is a price on not dying and it's "a car instead of public transportation".
> What has made the auto industry particularly vulnerable is its reliance on just-in-time delivery, where parts are brought in when needed, rather than being stockpiled.
This idea has gotten a lot of play lately. But the unstated alternative is to somehow perfectly forecast future demand for parts. That's very difficult in general and doubly difficult during a global pandemic. And, in fact, well-practiced lean outfits are better at knowing which inputs are potentially most disruptive, because they already obsess over lead times for everything.
Without lean practices you just wind up with giants piles of almost random inventory. That you'd have wound up with a giant pile of CPUs is a total crapshoot. But you would absolutely positively have a bunch of stuff you don't need and never will. And that inventory would choke the whole company to death.
The whole idea that JIT destroyed some glorious, flawless past is the Nirvana fallacy. "Oh, supply chain disruptions happen at all, therefore JIT is entirely useless". It's just a silly idea and needs to be mocked at every opportunity.
You don't have to be a rocket scientist or have a PHD in supply chains to understand there are second/third order consequences by cancelling significant amounts of your forecasted production and then expecting your suppliers to magically pick up where you left off.
Its why you keep production/manufacturing warm rather than do stops (which is what they essentially forced their suppliers to do by cancelling purchases).
This is BCP 101 and I give the auto industry no sympathy since most of this pain was entirely self inflicted. They tried to push all the risk to their suppliers, left them holding the bag and had a shocked Pikachu face when they couldn't ramp back up to prior levels.
We're talking about different things. A company running to optimise EOQ and favouring long production runs with few changeovers would be just as screwed by misunderstanding lead times of critical components. JIT doesn't excuse you from supply chain shocks but neither does Lost Magical Process Of The Wonderful Times Before JIT.
I find it bothersome that this situation is referred to as simply a shortage without any mention of the incorrect predictions or decisions the automakers made that lead to it.
It isn't. If you are instantly going to reduce orders you try to reduce your own cost and save money. Some automakers said we rather order to much then to little, and those producers have far less problems.
But it's affecting everyone in electronics as far as I know. My company is not in automotive, but pretty much every lead time has exploded, prices have increased and some pieces are just not available at all.
The automakers might have messed up their predictions, but they're not disrupting the whole electronics market like that.
They actually did disrupt the entire market like that. They cancelled all their orders, causing fab capacity to be redirected to other markets, then they came back and asked for their orders back, at any price, and everyone who was already making something else switched their now fewer available slots to service the car industry. See https://news.ycombinator.com/item?id=26842924
"Just-in-time" logistics are by definition highly susceptible to disruptions. Of course it can make sense in some cases, but I've never understood it as a near-universally applied doctrine. There are serious drawbacks.
Unscientific dogma: Mother nature is telling us through Taiwan's drought to slow down and redesign chips to be secure.
My late model American vehicle is my home. Earlier this week, the locks kept popping open when I was trying to go to sleep. Nobody else besides me has a remote entry key.
I hopped into a Jeep Cherokee that I thought was mine and drove it down the road for a minute before realizing I was not in my vehicle. Keys are funny like that.
You'd think keyfob security would be worked out by now (just look at the prevalence of skeleton key software defined radios amongst the grayhat crowd). There's probably a limitation in security complexity due to limited power availability in the keyfob.
Ah, cool. Sounds like there's no excuse. My mistake.
I mistakenly assumed you'd need some kind of cryptographic key signing handshake via active electronics from within the keyfob to achieve secure comms with the host vehicle, and that such a requirement may have been some kind of implementation limiting factor, security-wise.
I’m curious if automotive systems used more generic controllers for cabin systems if shortages like this would be less or more severe.
If the AC and the power seats and the cabin lights used the moral equivalent of a raspberry pi compute engine, would we see supply harder to disrupt, or massive consolidation that just makes these problems worse, for instance every single car plant shutting down for two weeks out of the same three months as the pause wends its way through the supply chain. Today a Chevrolet plant gets the only truckload, but Ram doesn’t run out until tomorrow due to transit delays.
The mass market parts probably wouldn't stand up to car environments, these are expected to operate for years with temperatures that range from well below freezing in the winter to over 150 parked in the sun when it's hot out.
> the moral equivalent of a raspberry pi compute engine
I did not mean mass market parts when I said “moral equivalent” since that was not clear. I meant automotive grade controllers.
I guess the real problem, besides shortages of the IC on the board as someone else said, would be connectors. The circuit boards aren’t just built different, the connectors are ruggedized and in some cases water/dirt resistant too. There’s no way you’ll use the same connector for a five way adjustable seat as for the cabin light controls. And if you did you couldn’t put any of them next to each other or someone plugs the defroster into the cruise control and somebody dies.
Best you could do is a standard circuit board and custom housing that has the wiring harness, and at that point things look pretty similar to the status quo.
Also relays in that situation are probably a bigger part of the wiring than I’m allowing.
There are automotive rated parts which comes down to an increased temperature range and (where applicable) vibration tolerance, electrical transients, etc.
Sometimes it involves different materials but usually it’s just more testing on an increased range.
For example a part which handles electric current might be rated to 100mA at an ambient temperature of 20C but only 40mA at 60C or -40C.
Regular consumer electronics you can just ignore these ranges as unreasonable but a car is going to get that hot or that cold so you have to have parts rated for those conditions.
The RPi compute engine is a PCB, a fairly large number of components, and a complicated chipset meant for video processing and Linux. This is not what your power seat wants, and it’s probably considerably more expensive than what your seat wants.
In any case, the current bottleneck is, to a vague approximation, a bottleneck on total output. An RPi compute module has a lot more mm^2 of silicon than a little ECU. You also likely can’t make a good RPi compute-like module on the larger-feature-size fabs that make ECUs.
Problem is everything needs to be automotive grade components, and if the supply is already low such a type or in contention, they are going to cost a heck of a lot more than the absurdity which is the 100% additional cost for a automotive grade resistor or IC versus a consumer level component.
> I’m curious if automotive systems used more generic controllers for cabin systems if shortages like this would be less or more severe.
that wouldn't help. generic and not-generic controllers would still use ICs, and ICs are still what we're slowed down on. a more generic design might have to use more (or more complicated) ICs as well, to have more tolerant inputs.
more generic/standardized controllers might help if there were limits on printed circuit board manufacturing, or board assembly capacity. (as those things get cheaper/easier/faster the more identical units you make. and yes, that applies to ICs as well, but all of the widely used microcontrollers / jellybean parts are already produced in huge volumes.)
This is already the case. A new Chevy pickup and a new Ford pickup have the same transmission. BMW and Mercedes buy transmissions from ZF or other vendors. Bosch provides engine management systems to like... everyone. I'd be astonished if there are more than a dozen power seat motors in use across the entire new car fleet.
Tesla is a notable exception because they actually make their own components, like the Model X Gull wing door actuators.
I meant within a single vehicle but that’s probably the best you’re going to get.
It wasn’t until I was an adult living on the coast that I understood this and why Midwestern auto shops were so resistant to foreign cars. This process was still in development, and so while one set of brake assembly might fit twenty Chevy and Ford vehicles over a five model year period, you were always having to special order parts for the “weird” cars because you couldn’t keep them in inventory, and so you had to deal with a hostile customer who didn’t understand why their wife’s car got brakes in one day and the Subaru was going to take a week. They storm off to another shop and get the same answer.
Ah yeah that is different then. I do wonder where these supply chains broke down specifically. Is it compute modules or specific chips/components? How feasible is it to have interchangeable compute modules with different internal implementations?
My car was available with I think six different batteries depending on options but they are also two entirely different chemistries (AGM and lead-acid). If there is a lead-acid battery shortage for some reason presumably everything could have been AGM.
The car itself doesn't care about the battery chemistry at all - it's actually not even a different chemistry, only a different way to build them -, it's always a nominal voltage of 12V and a charge voltage of about 14V. You can put in whatever battery you want, in theory.
Practically, your choice will be influenced by:
- price (lead-acid are cheaper)
- usage frequency (AGM have lower self discharge rates and can stand up for a year or more without charging, while a lead-acid will grow sulfate crystals in a matter of months)
- peak current capacity and temperature range (again, AGM are better, and you must use a battery that can supply the crank with enough amps... and note that the crank will need more power in colder climates)
- capacity (obviously... depending on options like a beefy sound system, service lights, antitheft systems or seat heaters, you need higher capacity)
My car absolutely cares about the battery chemistry because it optimizes the charging based on battery performance and capability. It also uses this information for the condition based service to indicate when the battery needs replaced and as an input to identify other electrical problems such as excessive discharge.
I hope countries learn from this mess and start working to build up their local manufacturing capacity, even if only to cover their own needs when global disruptions like this hit. It doesn't have to be cutting edge or competitive cost-wise. Even late 90s era tech would do.
ICs, like electricity, are indispensable to modern life. They're availability must be secured and protected in the same way.
But not ideal. Though all that, I wonder if this was a perfect storm and with the increases and demands for all things smart, it may be that we are playing catchup with a moving goalpost for a few years yet.
Not relevant to production of consumer vehicles under a temporary shortage of high-tech parts (though, under a longer-term shortage, it might be) but the Soviets had an interesting approach to high-tech dependencies in their military equipment:
I first encountered the idea of "Monkey Models" in Suvorov's book (referenced on that page).
The TL;DR is that the Soviets would design their equipment with the best high-tech sensors, weapons, countermeasures, et c., that they could reasonably manage, but also design the equipment to function with much simpler parts & manufacturing processes. So a high-tech Soviet tank might have an electronic targeting system, but also be designed to work with a simpler glass-and-steel rangefinder that could be built with relatively simple tools, in a half-decent machine shop shed. They might fit their best models with advanced armor plating, but design a variant that replaced all that with a little extra steel. They'd do this with practically everything, including aircraft.
Why? Multiple reasons: 1) it let them export "new" equipment to allies and puppet-states at a lower cost and in much greater quantities, by selling them "monkey models" with much of the high-tech gear & parts swapped for low-tech counterparts (older generations of top-end gear would be sent to the closest allies/puppets or, more often, to domestic reserve units, in a kind of tiered system), 2) since most of the Soviet gear the West encountered was in direct or proxy wars with Soviet ally, client, or puppet states, the West couldn't gain much insight into the actual capabilities of modern Soviet equipment, 3) so-equipped allies would be starved of gear that could threaten the actual Soviet military, in case they became adversaries, 4) less-advanced allies could more easily maintain gear without so much high-tech junk in it, and 5) perhaps most importantly, it gave the Soviets a kind of supply-line defense-in-depth—they had not only designed these weapon systems so they could be built (as weaker versions) without high-tech manufacturing, but practiced doing it. In the event of a shooting war with, say, the US, the Soviets could keep shipping (inferior, but much better than nothing) tanks & aircraft to the front lines even if all their high-tech facilities were bombed out of existence and they lost access to advanced materials (say, high-tech armor material), with hardly a hiccup.
I knew about the monkey models from Iraq, but I never made the connection to supply chain resiliency and wartime production. Thanks for the new perspective!
Rovers, boats, drones, planes, helicopters, submersibles, all done with Arduinos or similar. It seems it can be done if one really wanted to. When I think about all of the things my drone can do that are seemingly more complex than a non-self driving car can do it seems more than plausible. https://ardupilot.org/ardupilot/
If by "basic vehicle functionality" you mean enough to run the engine and move the vehicle, there's a lot of other aftermarket ECUs available, and I suppose any engine can be converted to completely non-computerised operation with the addition of a carburetor and points ignition (at the expense of emissions and fuel economy.)
Doing computation in hardware is lightening fast and highly desirable on mobile devices however this chip shortage for vehicles would get resolved if the computation on those chips was shifted into software and executed on commodity x64 cpu hardware ... possibly expedited if the computation intended to get run on these hard to obtain chips already has a software level simulation solution ... would be interested in learning of any show stoppers to this proposed solution ... turnaround time to go this route would be worth it especially since it eliminates such a chip shortage in the future ... is it that much simpler to implement computation in purpose built hardware than software ?
I bet Tesla's decision to design and manufacture their own machine learning chips for their "self-driving" functionality is looking pretty smart now. They have their own contracts with the fab companies to produce them. These are not the only kinds of chips that go into cars by a long shot these days, but it is the biggest ticket one.
That was a multi-billion dollar, very bold and risky bet that paid off. How many car companies do you know where they decided to take on industry leaders like Nvidia and Intel and actually produce a better product? That's really quite remarkable.
If GM said tomorrow that they were going to build better machine-learning chips than Nvidia, we'd all get a good laugh at that.
But I get a sensation that they are more into off the shelf parts than traditional automakers. SpaceX is also known to use consumer grade chips in their rockets
Maybe. Things that succeed in the face of an issue no one predicted don't really stand out as smart - just lucky.
It might prove smart in the long run for many other reasons of course.
For all we know, Tesla could be locked into TSMC like everyone else, while Nvidia has their current high-end chips on Samsung as well relationships with TSMC.
The beauty of luck is that it can be used practically any time as explanation for success. Tesla bet big on vertical integration. Who knows what they took into consideration but it was surely bold since they were not as flush with cash as they are now.
I checked a Jaguar dealer page now and I see they're still offering finance incentives which I didn't except since their inventory is taking such a hit. Plus luxury cars have been selling for over MSRP on most popular models for the last year.
Many finance incentives are usually restricted to current dealer inventory and previous model years. They also get people in the door, since often times, people don't truly qualify for the advertised rates (if we are talking about financing terms). Jaguar is still suffering from depressed sales (though it has been getting better), so they may feel the demand isn't there to get those vehicles off their lots and increasing prices/reducing financing incentives would potentially make that worse.
Of course, it might just be that they are slow to react and changes will be reflected in the coming weeks.
Jaguar is probably not considered on the same level of demand as say, BMW, Benz, Audi, or even Lexus. The fact that they were notorious for having problems probably doesn't help.
They make gorgeous cars though, with good handling and driver engagement. A V8 (or maybe electric!) F-Type is my attainable dream car.
It's better to raise the net price by raising the price and then offer incentives from a sales perspective than to leave the price alone and remove incentives. People feel good when they get "discounts" or "deals".
What I (foolishly) hadn't realized until I saw it in EE Times the other day: the shortage is in parts built on older tech (e.g. larger feature size). These factories don't produce the high margin parts so investment in them has lagged.
Allegedly you can bring up a fab large node (still sub micron) in just 4-5 months -- there's a lot of surplus / used gear out there, but will anyone bother (try might not earn back your cap ex).
Someone with the right background and contacts should put it together, I think the smart move there is to not only plan on selling chips into regular markets, but to get into talks with customers to become a guaranteed supplier that derisks correlated shortages from other sources.
One problem is the car companies have such razor thin margins that investing for them specifically is pretty bad. They are literally shaving pennies here and there on the backs of their suppliers, which reminds me of the (physical) toy and game business.
It’s no wonder there was inadequate response from their supply chain.
How much of our collective chip-manufacturing capacity is being devoted to Crypto mining? (a useless and environmentally destructive activity)
Could Land Rover petition the UK Government to ban the exchange of Proof of Work coins in the UK in order to alleviate this chip shortage? Mining activity is a function of the price of the coin, and even bans from minor economies like Turkey have been enough to decrease prices.
I fear our upcoming high inflation will be abused as an example to legitimize harsh austerity in the future, because liars will draw the causality with the expansive monetary policy of the 2010s and not with the Corona pandemic. They already conveniently forget that there was a flu pandemic in the beginning of the 20th Century.
same. but I truly don't think we will have large inflation. This is just creakiness of a machine restarting after being turned to a lower setting based on unmet expectations of a long span of limited consumer demand.
You'd think with decades of CAD experience we could figure out how to build a chipless car that has all the benefits of computers in cars and none of the downsides of carburetors. How long will this have to go on before someone considers it?
Even if you could technically meet the legal requirements (which I'm pretty sure you couldn't), it'd be impossible to mass manufacture and completely undesirable to consumers. Imagine having to change out the paper rolls/ink in your analog EDR every couple hours, while making sure the battery doesn't explode every time you charge or discharge it. The dashboard alone would be a sight to behold.
Honestly, it'd be fun to see how far you could get. Backup fiberscopes vs cameras?
Funny nobody considers designs which don't rely on complex chips - and simple ones can be made cheaply in many places. Are we that demanding that WiFi and USB are required for cars?
I'd buy a simpler car instead of having no car any day.
Even if you get rid of wifi and USB -- even if you get rid of the radio -- you would still want electronic ignition, fuel injection, antilock brakes, traction control, airbags, onboard diagnostics... the list goes on.
It's not realistic nowadays to expect a carmaker to sell vehicles that don't depend upon silicon. Ironically, it's not the complex chips that are the issue. It's the little basic ones that have been around for years. Hall effect sensors and things like that. Perhaps nobody ever thought they could be jolted by abnormal market behavior.
Not to mention that prettyuch all the major automakers have laid out a timeline for some or all of their product lines to go electric. Not sure how you would build an electric vehicle without semiconductors.
I believe it is impossible design and to sell a car that does not use these chips in Europe. Probably also other large markets. You can not get sufficient safety and emissions without them.
My whole point is that it is, in fact, possible to have many - practically all important - of those features without complex chips - at best with more simple ones, which don't require latest semiconductor technologies to manufacture.
I won't itemize to GP, but at least remember that onboard computer is a relatively recent addition for cars.
> you would still want electronic ignition, fuel injection, antilock brakes, traction control, airbags, onboard diagnostics... the list goes on.
Fuel injection is quite important; I drove a car with carburetor, but fuel injection was winning even before onboard computers. However you can have one with hardly any electronics; a rather small and simple analog circuit can work.
Electronic ignition begs question: what's that electronic in that? Electrical spark was used in cars, I guess, from the invention of internal combustion engines; but it's not electronic. Timing of the ignition? Cars can and were built without that; even if exists, it's something which sounds rather similar to fuel injection issue.
Antilock braking was invented and used as a mechanical system. Same with airbags. Traction control at least evolved from mechanical systems. Onboard diagnostics is nice, but periodical checkups served people well for a long time.
In other words, we can have a working car with necessary functions made possible either completely without electronics, or with analog electronics. In the worst case we might use small-scale integrated circuits - and that is already for functions which weren't that popular less than a generation ago.
My point, which I believe was misunderstood, is that it looks like a mistake to design a modern car in such a way that it can't be produced without relying on several exclusive manufacturers of high-quality semiconductors, while those semiconductors aren't critical for basic functions of the car. Those custom chips can't even be replaced with alternatives! Make at least an extension module, which would provide all those relatively recent features. I don't understand that it worth stopping car production in case those non-essential - even if quite convenient - functions can't be supplied at the same time as basic necessities.
I think car manufacturers have been specializing their chips and electronics more and more to avoid 3rd party servicing and now we see the fallout which could still be a win for them. Just jack up the prices! How about we standardize the chips used in automotive manufacturing to help alleviate this problem and cut down the number of unique SKUs fabs have to make.
That's an extremely bad take. Cars have had and shared ICs since they became mainstream. Bosch shared electronics between Volvo, Saab, Porsche even BMW bikes as far back as the 80s. Often times parts like air fuel meters, sensors, etc are all designed and manufactured upstream by companies like Bosch or Febi or other companies that sell systems to manufacturers.
We have platforms, like the GM2900 that many cars were based on as well to reduce numbers of skus
I think it's fair to say cars today are packed with more and more full blown cpus since 2010 onwards and so much of it it just junk. Maybe partially due to safety regs like mandatory backup cameras but android auto and car play? Notice how difficult it is to get aftermarket android auto (in my experience). Why can't it just hook up to my cell phone? But to be fair I am totally shooting from the hip here, and the article isn't making it clear which chips are in short supply. But as some other commenters pointed out, if it is really purely a shortage, why don't they just jack up the price. These are luxury brand cars I don't think consumers aren't willing to pay.
Have you heard of the automotive right to repair initiative from 2012? Why do you think such a thing was necessary? And guess the main method used to circumvent right to repair automotive laws that were passed in the USA? More complex electronics.
> I think it's fair to say cars today are packed with more and more full blown cpus since 2010 onwards
ISO standardized the CANbus, which is a networked message-based protocol for components talking to eachother. It was first developed by Bosch. Intel started making chips in the early 90s.
By the time you get to the Saab 9-5 (1999) you have CANBus running all over the car. The drivers seat? There's 5 motors, a heater, adjusters, buttons,etc. That all talks to the central computer and can be diagnosed by talking to it with a diagnosis device you plug into the OBD2 connector. this is fantastic because the microcontroller can detect the position of the seat for example - run tests and use hall effect sensors to determine what is the problem. This sucks if you don't have a TECHII diagnosis tool.
OBD2 is what I'm talking about. Required by law. But guess which auto manufacturer has the port accessible, with the pins only outputting power and ground today ( no data )? It's a new trend. Data now is sent wirelessly over a proprietary protocol.
Pardon my elementary understanding of economics, but I own a lot of Taiwan Semiconductor stock, and I would have thought the answer would be for them to raise their prices until supply meets demand, and that would probably not result in luxury cars being priced out. Also would help my TSM stock quit languishing as it has been doing for a while now.
> After the [Fukushima] catastrophe severed Toyota’s supply chains on March 11, 2011, the world’s biggest automaker realised the lead-time for semiconductors was way too long to cope with devastating shocks such as natural disasters.
> That’s why Toyota came up with a business continuity plan (BCP) that required suppliers to stockpile anywhere from two to six months’ worth of chips for the Japanese carmaker, depending on the time it takes from order to delivery, four sources said.
> The sources said Toyota has another advantage over some rivals when it comes to chips thanks to its long-standing policy of ensuring it understands all the technology used in its cars, rather than relying on suppliers to provide “black boxes”.
https://www.reuters.com/article/us-japan-fukushima-anniversa...