You're fabbing it wrong: Chip shortages due to lack of investment in the right factories, says IDC Semiconductor shortage issues will continue through the first half of 2022 as the industry attempts to build up inventory to normal levels, according to research firm IDC. It cites limited investment in mature process technology as one reason, with many vital components for the automotive industry and other sectors …
I wonder how Rochester Electronics is doing through all of this? They specialize in obsolete and EOL ICs. For a lot of their business, they're a conventional broker. They buy up stock that's going EOL, and later sell those chips to those in need.
Another part of the business is to actually build obsolete ICs. They buy up the rights and the masks to continue building chips after the original manufacturer wants to quit.
No clue if they hands their own fab, or if they rent time/subcontract to someone like TSMC.
Historically we're had leading-edge manufacturing making goods to a higher and higher quality.
Then in parallel we the movement of production going off-shore, allowing us all to buy the high-end for less.
Last few years have shown us that the two progressions above have left holes on our manufacturing.
Whereas previously a skilled person with a machine lathe could have made your required legacy widget for a handsome markup, there are now holes. You could offer $50 for that $5 chip your car needs - but nobody can make it for you - they're now making $500 parts and don't feel the urge to answer your RFP.
The motor manfacurers (as an example) have now learnt that battering down suppliers on price doesn't work for high-investment parts, like it maybe previously did injection molding the plastic on a wing-mirror.
My guess is that this is a "one time" lesson - long term availability will be baked into contracts going foward - or like Tesla they'll design a product with replaceable components.
>Historically we're had leading-edge manufacturing making goods to a higher and higher quality.
I wish. We've (US) typically gone for the 'first to market' and 'grab the IP' route. We may make something if its small run, bespoke and high margin but generally we like to sit back, let someone else do the work and make the profit on markups. (It used to be called the "Smiley Face" curve.)
The problem with this approach is that eventually the vassals making the stuff for imports figure out how to sell the stuff themselves. Since IP protection only goes so far eventually that commodity income stream evaporates.
The UK fell into a similar type of trap a generation ago or so.
Long term contracts? Ok, as long as the client has to pay for not manufacturing.
The thing is, why sign such a contrac with no warranty? It is not in the interest of the foundry.
They did in the past, and were left holding the bag. They don't want to do it anymore.
As for designing with repleaceable components.. ha! They mostly don't design the cars, just RFP parts of the cars to be designed, squeeze the suppliers and at most put the 3rd party parts together as cheaply as possible.
"long term availability will be baked into contracts going foward"
In a similar way as a Service Level Agreement guarantees the competence of the service supplier and the impossibility of service disruption.
the naivety of some people is hard to believe.
Maybe.
Or maybe the power balance is about to permanently shift.
In the Good Old Days, a car company was basically three things. An engine Designer, a chassis manufacturer, and an assembly plant.
Of these, the engine design ruled the roost. Without the engine, the car is going nowhere.
In the new electric world, which isn’t here yet, but soon, the electric motor just isn’t the specialist bit of kit. Loads of companies make them as commodities. The battery is specialist, but the car manufacturers neither make nor design that. Nor do the car manufacturers make the electronics that now constitute 50% of the value of the car.
So why would the owner of the car assembly plant really be the car marque? Why isn’t the car marque just the electronics prime, who subs out a bit of assembly, which is 10% of the car value. Which is done by robotics mostly anyway, bought from ABB, Fanuc, Motoman. What exactly does Volkswagen or Ford bring to the table that they should be paid loads for?
The chassis is one thing. An unacknowledged fact is that the Tesla is a pitifully poor chassis from the 70s. But apparently people still love them.
In just the same way as the actual microphone, speaker and voice quality on a current iPhone is far worse than a 2000-era Nokia. But nobody is buying the Nokia, because they don’t actually want a phone as such.
Maybe an 'open spec" for car computers is in order? You know, like PCs.
Problem is that gummints seek to REGULATE TOO MUCH (especially Cali-Fornicate-You) when it comes to engine performance and whatnot. With minimal safety standards, and NOT insane regulations on fuel utilization, having people (and car dealers) able to shop around for better 'after market' equipment that plugs right into the vehicle might be the BEST option overall... and a potential "2nd source" for stuff that's currently in short supply.
As for electric cars, it will be a VERY long time before they replace gasoline and diesel. In my opinion they are still HIGHLY overrated.
/me pointing out that car computers and other related components most likely have to be "approved" by bureaucrats before they can be put into cars. Therefore, swapping out a part with a minimal design change to use something that IS available is no longer an option. Bureaucracies are JUST too inefficient to allow for that kind of flexibility. And so a shortage on a 2 cent part stops the line...
As I noted, so long as the market for cars continues in the “Old World Order”, a standardised open spec for car electronics sounds like a great idea. And it *might* happen that way. But it depends on the assumption that the value of the car, and car brand, lies in the engine and chassis. Then the standardised electronics is not differentiating in any way.
But it alternatively might happen that *all* the value, and the brand, is in the consumer experience. Then quite the opposite: the *electronics primes* come together to write a standardised commodity spec for the low-bidder mechanical assemblers to meet.
I reckon there’s a high likelihood from your post that you own a car, enjoy driving to the performance of the car, take pride in your driving, the next car you buy will be the newer model of what you drive today, and you don’t Uber. Certainly I fit all of those. But I’m *old*. I learnt to drive when I was 17, same as everybody else, and I drove the fastest 10year old car I could afford to insure and write off.
When you talk to 23 year olds and even 30years olds now, not only do most of them not drive, many of them would be *shamed* if they turn up to a friends house or party driving a car. They Uber everywhere. Being able to drive at all, is something that marks you out as an Uber driver or DadCore, and therefore a *total loser*. And they don’t need the car to do the “the big shop”, because that’s either delivered or they’re having Deliveroo anyway. And Ubers come in two flavours: it’s either an unnamed car with the right number of seats, or it’s a Limo. Limos have zero performance, but they do have plush seats, a minibar, WiFi, space for six plus anyone they stop for on the way. and a tricked out gaming rig possibly with wraparound screens.
The direction of travel, is that it’s what our generation used to call a car, which is anonymous unbranded commodity, whereas it’s the electronics platform that has a value.
All inventory stockers and brokers including Rochester have been doing OK, I've used them even before the shortages hit. Rochester are now on Digikey's marketplace.
The problem is that if you're on small to medium production runs, they often tell you you have to buy full reels of, say, 10,000 units, but you might only need 1,000. For passives it's usually not a big deal because they're sub-pennies, but for semiconductors it's a different kettle of fish, so you end up sitting on a lot of inventory.
JIT manufacturing is dead.
I had a visit to my CEM (contract electronics manufacturer) just before the new year, and I've never seen their stores so full: there are so many projects with parts on back order. Just one part in the BOM missing and you can't do a production run as you can't test your board or assembly..
The worst part about it is that there inventory for most parts exists, but it's sitting in shady warehouses in China. Predominently Chinese businesses have come out of nowhere, and buy up everything on a speculative whim. As they're the only ones stocking the parts, they can demand any price they like, and it's not just a 10 or 20% premium, it's at least three times but often ten or 15x the manufacturer's price.
Many of us are reworking designs to use what's available, but it's been a game a whac-a-mole for over a year now: you redesign for different part(s) and suddenly they disappear too.
The end result is that for the first time ever I've had to increase prices on my products by 20%. You will also notice that some products will just disappear or not even make the retail market at all despite a marketing push: Intel NUCs have been a good example of that for example.
Regarding TI in the article, if you look at their own inventory, I'd say about 70 to 80% of their products are out of stock. TI is quite big in a number of areas, and power supply parts in particular voltage regulators is a problem for everyone now. These aren't fancy state of the art 5 or 7 nm nodes, they have no need to be. There's a move by some to go back to good old fashioned discrete solutions suing jelly bean parts: they're significantly more effort to design, but it you're not space constrained it can make sense. The increased parts count will inevitably hit the COG.
Lead times for many parts are extending out to 2023 now. As a result, everybody is stockpiling, it's like Mad Max. While the bog roll thing we saw was just pathetic, for businesses if they have no parts, they have no product, they have no sales, they have no business.
>Predominently Chinese businesses have come out of nowhere, and buy up everything on a speculative whim.
Those damn Chinese capitalists. If only the socialist banks, hedge funds and investors in the west had the profit motive, organization experience and access to cheap capital to allow them to do this
JIT manufacturing is dead.
I hope you are right. Decades ago I partnered with a materials management guy to produce software to assist with improving inventory control by developing more accurate sales forecasts. I'd like to resurrect this. Being re-done as a cloud-based system is probably a good idea, either private OR public cloud. Or maybe just a web-based system. I think I'll look into it...
looks like we need more companies doing things like Rochester and T.I. do, _ESPECIALLY_ if they build fabs in UK, EU, USA, Mexico, Canada, and Japan. Big ones. Lots of robotics and automation. Without the need of a swarm of slave wage minions to get work done, building things OUTSIDE of China becomes profitable again.
This is an extremely good opportunity for the USA, Europe, Canada, even Japan to start building high tech node manufacturing plants.
There will plenty of buyers waiting in line now due to shortages and then later plenty more with electric vehicles coming on board. Screw the cheap labor BS. First off so much of it can be done with robots now. And, high paid employees in a country with highly trained and educated employees and a favorable business and political climate are big pluses.
If I was smart and had a few hundred million bucks laying around I'd build a chip factory right now. And, I bet I would be able to sell every chip that comes off the line for a good profit.
Not so fast.
The front end costs for any industrial stuff is generally the same across the world.
You want a 5 axis alll singing all dancing CNC machine complete with intergrated robot handler.... best you pony up some $$$ to germany.. transport costs are naff all in comparision... and just need a warehouse with a decent concrete floor to install it on.
Then comes the variable prices such as power, waste disposal and finally paying for the staff
So if you're bashing low margin stuff, you dont want to pay any more than you have to for the staff, high margin stuff needs the staff so thats going to cost you plenty of $$$.
Of course you could use the no brain method of hiring expensive staff to build up your multi-million factory, then firing the lot and hiring min wagers as 'everything is setup with instructions for doing the job'
Which works very well until the min wagers who dont give a toss dont follow the instructions exactly........and your multi-million factory stops because 1 of the machines has just been turned into scrap metal.....
"Of course you could use the no brain method of hiring expensive staff to build up your multi-million factory, then firing the lot and hiring min wagers as 'everything is setup with instructions for doing the job'"
Also know as the "South African Method".
"Which works very well until the min wagers who dont give a toss dont follow the instructions exactly........and your multi-million factory stops because 1 of the machines has just been turned into scrap metal....."
Yup...has ZA written all over it.
It is the legacy nodes (this has also been biting IT) that are hard to come by. Older plants, building older chips on older processes.
Even smartphones have been caught by this. Those older facilities build chips for everyone, including things like USB converters etc. in older, larger processes. There is no benefit for making them on smaller processes, it just makes them more expensive for no benefit.
Part of the problem with automotive was that they missed their windows and didn't order far enough in advance or have high enough priority (number of units ordered) and, because of close-downs, they cancelled some orders.
As the chip plants went offline due to COVID - limited raw material supplies, workers infected etc. - the production orders were cancelled or pushed into the future. During the re-jigging of production plans, small orders or less profitable orders were pushed further back, or fitted in, where there was free capacity. Plus, when the automotive production plants came back online, they suddenly wanted orders fulfilled at short notice, when the chip plants were already working over capacity.
The problem for car makers is that they order in relatively low quantities, compared to the IT industry, for example, or consumer and white goods.
IT and tech products (PCs, server, smartphones) need those legacy nodes as well, and they buy in the 10s of millions of units, and they probably source some of their high-value nodes from the same supplier as well, meaning they can apply more pressure to get their legacy nodes prioritised (hey, if you can't supply those legacy nodes, we won't need out high profit nodes from you either). consumer electronics and white goods are next in line and so on, with automotive being somewhere near the bottom of the pile and order on their JIT cycle, instead of stockpiling.
I work in a different industry, but we were struck with the same sort of raw material supply problems. Luckily, our purchasers saw the problem early and convinced management to bulk order on certain materials, so we had our yards full of raw material, but at least we could continue production. Likewise, strict controls on testing and quarantine meant we didn't have any COVID related production outages.
We have people that design chips the world uses then we outsource making them. How totally stupid. We rely on chips in everything so it is a strategic blunder to outsource, same as it is a strategic blunder not to be able to produce steel, aluminium, nylon and all manner of other useful materials on UK soil. That we dont have to would be a relief, but that we cant is a danger.
You have to concentrate on what you have competitive advantage, what you do best, and let others do what they do best, this is what has given so much more advancement to the world.
The problem is, with the huge increase in costs, etc etc, we are getting worse at the crucial things..
The current crises are making people re-evaluate this.
With COVID, US trade embargoes limiting supply of many essential items from China. the Russian sabre rattling, increased cyber attacks etc. many governments and regions are looking to bring at least some of the production and expertise back within their borders.
It gives better control of the supply chain, fewer chances for unwanted "parts" being added to mainboards, for example, or rogue code getting in during manufacturing - one of the unsubstantiated claims against Huawei, for example. Plus cases, such as the NSA slipping in custom ROMs on kit going out to supposed USA allies.
Independence seems to be a growing trend, again.
@Dave 15
"We have people that design chips the world uses then we outsource making them. How totally stupid."
Why? Can you imagine trying to make these things here with huge costs for energy, labour, regs, etc. Makes sense to design them here but manufacturing is probably better done elsewhere where its cheaper
I thought the chip shortage started with the trade war, supplies got disrupted, the covid and its sons continued taking its toll, now all three are combined. US factories cannot produce enough supplies, prices go crazy, incompetent industries unable to compete can reap a lot of profit.
Every hot war in the past was preceded with a trade war (1 and II both). Are we on the path to the III? SCS and Pivot to Pacific seems to indicate so....
Nothing helps efficient industries like fair competition. Nothing destroys a society faster than a trade war....
American and European semiconductor manufacturers are outsourcing most of their production to China, creating a dependence on the CCP for low-tech (relatively) yet critical components in cars, white goods etc. I attribute this to sheer laziness, nothing more.
If the Chinese can produce these semiconductors more cheaply, we should study why this is. And it can't be all put down to labor costs. Environmental laws and labor shortages also play a big role in the pricing of these parts.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
