For a billion dollars, you're not getting anything latest and greatest. And you're definitely not going to get anything in the timeframe of the chip shortage.
I applaud the effort, but feel like it's just going to be people taking advantage of the money for no net benefit.
I was on a panel with one of the study's authors recently, she's very thoughtful, thorough, and precise. I like how she emphasized that there is no simple TL;DR to explain the situation, and her paper does a good job of covering a lot of ground without taking short cuts.
However, one factor not mentioned in the study is that China, which had announced a plan for foundry independence, had its capacity expansion program disrupted by the trade war, as the US restricted China's ability to acquire key equipment to build out fabs.
As a result China could not build the plants it needed to meet its own consumption, forcing them to buy chips from the same vendors that everyone else is also trying to buy chips from. This is an additional shock which is hard to estimate in magnitude but it's probably not small as it represents the missing capacity of entire fabs for many months or years as they struggle to be built.
Thus while the US trade policy may have effectively hamstrung China's plans for foundry independence in the short term, it may also have contributed to a global shortage of semiconductors and its own inflation problems as collateral damage. Trade wars are tricky: without careful and long term policy planning there is a high risk of even more collateral damage.
Another factor that could also be contributing specifically to the difficulties of the automotive supply chain is that China internally had made electric vehicles a strategic technology focus, which means that domestically China's demand for automotive-grade semiconductors was set to surge. Even before the pandemic I was getting tips from buyers that power semiconductors, particularly those in SiC and GaN, were going to be in short supply, due to that shift.
How does your view comport with the fact that the supply shortages did not actually begin with the trade war or the pandemic? The problem worsened, certainly, but global stock of components had been tight for years before any of these supposed causes.
If fault must be found with the US it is in the consolidation and outsourcing of manufacturing to a small number of manufacturers, removing competitors and excess capacity from the supply chain. China's part has simply been to grow; creating consumers and thus demand for products that it had previously not consumed.
Absolutely none of this was a problem before the Trump sanctions took effect (which we can now call the "Biden sanctions," as he has done nothing to revert them.) There were occasional shortages of things like MLCC capacitors, some of them severe, but nothing as far-reaching as what we've been seeing.
It is fairly clear that we'd still be in trouble absent the "easy to win" trade war against China, though. COVID alone would have been enough to upset the proverbial apple cart.
I disagree. Some of Trumps policies were ridiculed by the same people who now have the power and choose to continue the exact same policies. In some cases even more extreme.
Keeping history within view helps to view current events in perspective.
Reagan was ridiculed as well. He did a pretty good job at guiding the end of the cold war into a remarkably non-eventful event ( this could be worded better, not a native speaker ).
The problem with the tech sanctions is that the bullet has left the chamber.
Even if the Biden administration were to say fine - Huawei, you can produce your chips at TSMC, and SMIC, you can buy an EUVL machine from ASML. China would still persue a completely US influence free semiconductor supply chain, as the US has clearly demonstrated its willingness weaponize its dominance.
From a Chinese perspective, the interest of state and its big unruly tech companies are now aligned. Alibaba, Xiaomi, etc. are all making chips now.
It is not all given that the escalated tech war is a longterm winning move from a US perspective.
"It is not all given that the escalated tech war is a longterm winning move from a US perspective."
This is what bothers me most about 'tough on China' folks, whats the next move?
Foreign policy has to think few steps ahead. This is not looking much better than 'invade X country in middle east and figure out what to do next later"
Funny enough during the pandemic, it was better for manufacturing to be off-shore.
That's because American price-controls and anti-price-gouging laws (and chilling effects from fear of them) don't really apply to foreign importers to the US, only to domestic companies. US lawmakers haven't quite gotten around to legislating world market prices yet after all.
So the prices of imported goods could rise to clear the markets. For domestically produced goods you were more likely to see empty shelves, lines and (informal) rationing; like super markets only selling you limited quantities.
Countries with manufacturing under government control haven't fared that well. The US has led a lot of fab investments over the past 1-2 years, with at least seven new fabs under construction from Intel and TSMC. The America Competes Act, if passed will be a great fillip, and the chances are that it will be passed looks high at the moment.
The previous president, (not naming him just to keep things a bit apolitical) also had tons of bad ideas. A huge amount of the US advanced tech industry (such as semiconductors or EVs) depends on foreign talent, and publishing a Stephen Miller drafted Executive Order (No. 13940 - August 2020) with the stated purpose of disallowing any visa holders to work in a government contractor (which includes almost every tech company from Intel to Nvidia to Google) is a stupid move, and would have been suicidal.
If you wanted to keep it apolitical, why are you dragging a completely separate issue related to a certain, unnamed, but very well know person into the thread?
The overwhelming lesson of the preceding four-year administration is that everything is political. You may ignore politics, but rest assured, politics doesn't ignore you. Bad things tend to happen when you try.
Hiring people, which was honestly the most routine thing, has become insanely political. If an organization's employee list becomes a political point, then how are you going to escape that?
This is not an unrelated issue. You want chip fabs in the US. Great. Who is going to run them if Executive Orders prevent you from hiring fab scientists?
Why is that "unrelated"? Chip fabs don't run themselves.
Agreed about the harm already been done, even if sanctions would be retracted now.
> It is not all given that the escalated tech war is a longterm winning move from a US perspective.
But the alternative of not doing anything gives no guarantees either.
Trump was not alone in his view that 'something had to be done about China' with regards to trade and IP ( WTO ). It already was a hairy situation for everybody in the West and some even stated that apparently it took 'a Trump' to confront this situation.
This was already China’s plan before the sanctions. Sanctions only slowed them down but that was the goal so it worked. To buy some time before the US loses one of the few technological edges it still has.
"To see what is in front of one's nose needs a constant struggle. One thing that helps toward it is to keep a diary, or, at any rate, to keep some kind of record of one's opinions about important events. Otherwise, when some particularly absurd belief is exploded by events, one may simply forget that one ever held it. Political predictions are usually wrong. But even when one makes a correct one, to discover why one was right can be very illuminating. In general, one is only right when either wish or fear coincides with reality."
That list includes many fabless players and isn't a good indication of how much the sanctions would play out in the next few years, as we undergo a massive migration to EUV process nodes, which China is effectively barred from.
How much of that 50% is for devices manufactured in China for export? That business is also undergoing a massive shift, especially now we are close to the end of Covid.
To be fair, the majority of bread-and-butter semiconductor manufacturing probably doesn't need EUV process nodes. You only really need that level of technology if you want to make cutting-edge (10nm or so) devices like microprocessors.
The majority of semiconductor manufacturing is still in the 200nm range -- better yields, cheaper processes, more resilient chips. Some of it is still in the 1um (1 micron) range.
Huawei start buying huge amount of chips when there was rumor that US going to imposed sanctions on Huawei, and when the sanction hits, every major chips users in China start buying too, they were afraid that they will be next on sanction lists. They kept buying chips to build up safe inventory level during the early pandemic period when western firms were cutting orders.
Of course, when demands returned, Chinese companies were still on buying spree. The shrewd middleman everywhere, not just in China, saw shortage coming, and they stock piling chips too.
Yeah, most semi-industry people in Taiwan think that US sanctions and the threat to impose more sanctions definitely contribute to the current shortage.
What's better: A $1000 iPhone made in China that Americans can't afford because their jobs were globalized to China, or a $2000 iPhone made in USA that Americans can afford because they work in the iPhone factory or live in communities that are supported by it?
However, one factor not mentioned in the study is that China, which had announced a plan for foundry independence, had its capacity expansion program disrupted by the trade war, as the US restricted China's ability to acquire key equipment to build out fabs.
Page 5 alludes to this, but it does seem to be underrepresented in the report. Erratic policymaking on the US's part incentivized China to hoard semiconductors while they could still get them. For all they knew, their whole country was about to end up on the entity list.
Semiconductor nonproliferation feels as pointless as banning math after school. it’s a sign that the people who run the usa really don’t get along with the people that run China on some issue, and they picked semi conductors as the game of chicken that’s going to settle it.
> As a result China could not build the plants it needed to meet its own consumption, forcing them to buy chips from the same vendors that everyone else is also trying to buy chips from.
How a fab that didn't exist yet can affect current demand?
The biggest skeleton in the room was barely mentioned once.
Chip shortage, or, better say, capacity shortage for affordable nodes been there for 5-7 years, and nobody really gave it much notice.
Lead times for average 180nm project were 12 months+ for last 5 years.
There is still a stratospheric gap in between "common man fab service," and the realm of multi-million dollar semi companies, who can afford service on relatively new, (and not decades old) equipment.
All significant chips which went into shortage were using that "common man fab process" on eighties, and nineties era equipment.
> All significant chips which went into shortage were using that "common man fab process" on eighties, and nineties era equipment.
I don't think I agree with that. A LOT of small microcontrollers went into allocation and they were on 120nm, 90nm, and 65nm processes.
However, I do agree that there really was a shortage even before Covid and nobody cared. Effectively--nobody builds a line for an older node, ever. And there has been a lot of consolidation and sales of facilities such that very few people have been building fabs below 90nm.
The number of fabs built in the 2010's that aren't TSMC, Intel, or Samsung is pretty minimal.
> 65nm? 90nm? 180nm? I'd argue that the processes can be improved and cost can probably come down compared to the older machines.
This is really dubious for two reasons. First, you can do a die shrink and fit many, many more equivalent chips onto a single wafer at a modern node. This is somewhat mitigated by the high cost of mask sets at modern nodes -- but there's just no way it would ever make economic sense to build capacity at 180nm.
Second, the equipment for old nodes is not made anymore. It's cheap, but that's only because there's ample supply on the secondary market. If fabs began to get restarted in earnest at old nodes, that supply would quickly evaporate. Then you would have to buy modern (expensive!) equipment -- at which point why not take advantage of its full potential? Wafer throughput is the same regardless of whether you're making 100 chips per wafer at 180nm or 10,000 at 22nm.
Do you know how much modern masksets cost? More than a MILLION DOLLARS for a set, and completely ridiculous sums for anything <28nm.
So, if your market is <1M dies a year, you have to price your chip >$1 just to break even. And in the car chip market, even 1M dies is a giant amount.
Car market was all about decades old chips, made to stringent automotive specs, manufactured in tiny batches, and sold on long term contracts at sub-10 cent per packaged, tested, and binned die (and low bin chips were usually simply thrown away.)
> So, if your market is <1M dies a year, you have to price your chip >$1 just to break even. And in the car chip market, even 1M dies is a giant amount.
No, this is not the case. If you have a small run, the foundry will "consolidate" your chip with other customers to make a single mask. This makes sense all around, since the foundry would rather not have to deal with a mask set that has only a few wafer starts anyway.
You don't make commercial chips on a shuttle run. If you sit on the same mask as somebody else, you forsake all the flexibility, and you are limited to fabs with own mask shops — just taking your masks, and moving to another fab with them is near impossible. Asking the fab to make just any process tweaks is similarly impossible.
> buy modern (expensive!) equipment -- at which point why not take advantage of its full potential? Wafer throughput is the same regardless of whether you're making 100 chips per wafer at 180nm or 10,000 at 22nm
Yes, I agree, but what I'm saying is that there's a cost optimum that might not necessarily be at 22nm (you get more chips per wafer but if the machines cost 10x as much, let's say, then it's not necessarily worth it)
I'm not even sure everything can be shrunk that easily. Power ICs for example.
> you get more chips per wafer but if the machines cost 10x as much, let's say, then it's not necessarily worth it
That's sort of the point I'm making here: there is fundamentally no real cost differential unless you go really far back to when the equipment was really rudimentary. The old machines are cheap because they're used. If there was someone still manufacturing them, their cost would not be greatly different from modern equipment.
The best analogy I can make is an old car. Being older doesn't mean it's a whole lot cheaper to manufacture. Maybe a little -- no fancy computers or direct fuel injection -- but you lose that benefit really fast because of the shitty fuel economy.
Most of the modern equipment is "backward compatible"; modern CVD or PVD or CMP or etch equipment would work fine at any trailing node, with the exception of the fact that they would work only with 300mm wafers (which is actually an added cost efficiency).
The biggest difference I could imagine is with litho equipment. 193nm immersion is expensive and EUV is even more outrageous. You could probably use old non-immersion litho technology with modern multiple patterning instead.
Apologies for arguing the analogy, but the Benjamin Button car buyer would lose horsepower and weight (=safety and convenience features) to continue to make mandated fuel economy standards. You could buy a 50 mpg CRX HF from Honda over 30 years ago, but you'd probably be very unhappy with its performance today.
Small relative increases in fuel economy don't save you that much money, either
Especially considering that nowadays 180nm is accessible (if necessary, via double-patterning/immersion 405nm to please the DLP's aluminium mirrors) from maskless steppers, which can improve turn-around times by eliminating shared-mask aggregation delays and allows for cheap experimentation (no more minimum order quantity). This is even more significant when one remembers that maskless steppers already have to do stitching for everything significantly more complex than an opamp, and the per-area price scales up to full monolithic wafers without the tiling constrainst Cerebras has to adhere to.
> Effectively--nobody builds a line for an older node, ever.
UMC, Taiwan's No. 2 fab, offered to build a new 64nm line in an existing fab in exchange for a $1B firm order commitment. Nobody bit, they preferred instead to lean on the Biden administration to pressure the Taiwanese government to allow them to skip the line.
A new fab on old process nodes is not viable if it has to compete against fully amortized hand-me-down fabs that used to be cutting-edge 10-20 years ago.
According to standard economic theory, shortages only happen when pricing mechanisms fail.
If prices rise, then demand would lower and shortage will be resolved (?)..
> According to standard economic theory, shortages only happen when pricing mechanisms fail.
No, not it isn't. That assertion is fundamentally wrong at many levels. It seems you're confusing the intro to economics definition of equilibrium between supply and demand with your own definition of shortage. However, in the process you're making the mistake of believing somehow that pricing out demand from the market and being unable to scale up production to meet demand does not involve any shortages of any kind.
Think about it for a second: I'd suddenly food prices skyrocketed to the point a single 1kg bag of rice sold for over €100, and most of the world was thus unable to buy food for being priced out of the market, would that represent a shortage even if some people could still afford it?
When prices "skyrocket", it becomes profitable to use a lot more expensive means to produce (eg: using nuclear energy to grow rice indoors), this will quickly increase supply.
For microchips, it's likely the prices have not increased enough to kickstart this increase in supply.
The prices may be there, but that does not mean supply is able to pick up. There are non-monetary barriers to entry the market. Knowledge, access to tools and materials, permits from gov agencies, source of water, power etc.
And the time scale to establish new capacity from scratch is many years to a decade or more. Look at the capacity build up in China. It has taken decades, and that is with all the power of the nation, and in some cases loose on regulations regarding IP rights.
Considering what extremely performant and reliable chips you get for under $0.10, it is no wonder that few economies can deliver here.
It simply isn't worth it to manufacture these chips without selling billions of units. Not to mention the development, even the small chips today are extremely advanced from a technical perspective.
Opening a lemonade stand is a far more solid business plan. Since owning know how is somehow not worth anything despite heavy handed patent laws, I think the business case is just bad. China seems to work against this with specifically funding establishment of fabs.
You gave a right example, to prove the current problem... the prices are high now... go build a nuclear energy heater to grow rice, I'll wait.
The prices for chips are high now, the process of building a production facility is very very expensive, and takes a lot of time, and the people building them (and quite a few are being built right now), are still unsure if the high prices will hold, to make it worthwhile.
I work in this industry. Nobody is producing significantly more as a result of covid or anything else. Aggregate demand for electronics products has increased a bit, but it's not a significant change. The "skyrocketing demand" mentioned in this paper is not for electronics products, but for their components, and it's happening entirely due to a lack of visibility into future availability and demand, or, to put it more simply, a lack of confidence in the market. Device manufacturers are buying as much as they can because they don't know when they'll be able to buy again. Component manufacturers are not expanding production because they don't know when the rush buying will end and everyone will produce from stockpiled material. This is primarily a confidence problem (caused primarily by automotive manufacturers being absolute shits to their suppliers). The total production volumes of electronic devices have not increased by any significant amount. Many manufacturers are actually producing less now than pre-pandemic, simply because they can't produce more due to shortages.
I have read, regarding the GPU shortage, that silicon wafers are in short supply. Is that true? If so, how does it square with actual component production not rising? Are people that high in the supply-chain stockpiling silicon?
Component production cannot profitably increase because there is no increased total demand. Everyone is buying everything they can just to make sure they can produce, but they're not actually producing more. Component producers know this and that at some point this artificial demand due to stockpiling will fall off a cliff, and they'll be left sitting on underutilized massively expensive production equipment if they increase volumes. Besides, with the current stockpile-driven demand they can ask higher prices. This is why component production is not, and will not, rise.
GPU shortages are driven by a number of factors, and almost none of them are to do with component production (or wafers). There's a big scalper problem for GPUs, where some assholes will buy every GPU they can find and then resell them at a huge markup to either (increasingly desperate) end users and system integrators, or, increasingly, to cryptofuckers who will pay much more than legitimate users to ride their bubble. This means that even if there's plenty of production, unless the production can outspend the scalpers' total capital, it's all going to be swallowed up.
There's no general wafer shortage, but the highest grade wafers that get used in the highest end production have long lead times (because the fabs that use them reserve them way ahead of time). Wafer availability is not a bottleneck since the smallest nodes have such low capacity to start with. What has happened is that wafers have gone up in price in response to the general stockpiling behavior, but the high end stuff (including high-end GPUs) don't care as the wafer cost is basically negligible to them compared to all the process costs.
Or production is on forced holiday because we are lacking µC (ST) for our devices. So it is not only future demand or high tech components, part of it is already here.
If you need help with redesigning your devices to be manufacturable with what's available now, poke me, maybe I can help. I've been doing this sort of shit for over a year now. It's not fun, but it's saved a number of products.
Thank you for the offer. We were just starting to work on that and we have enough engineering power. We though that chips would become available again sooner and it is more about production planning and not wanting to start another model series.
Business is good otherwise we are not in danger of going out of business. But for people in production this is a cut of their income. We did get some supplies though and are expected to restart full production the week after next. Still crazy it happened at all. We never had a supply problem like this. We certainly will try to eliminate the bottleneck though.
This is also my suspicion. Further, since ALL the GPUs are being eaten up by crypto miners, people are calling it Unpresented demand. As for when It shall end? My take is when Ethernum switches to Proof of Stake model. By keeping the switch 6 months away forever, they hare in a way hindering the creation of ASIC miners, forcing everyone to use GPUs..... Once all those crypto GUPs reinter the gaming market, half the computer side supply problems would be gone.
How do crypto miners make use of the GPU's now? I did the sums some weeks ago, and even assuming free electric, the inflated price of the cards made it very difficult to become profitable. Best case, it's like 2 years before you break even
When markets only go up they bank on appreciation of the underlying cryptocurrency, and the ability to resell hardware into a scalped market when done.
Isn't the way to bolster the market to avoid the crazy 8nm stuff and just buy into the industry at the large scales. I can't imagine that buying or building machines to build op amps or voltage regulators isn't within the reach of lots of manufacturers.
I assume that this would allow the advanced manufacturers to concentrate on the newest tech unless the shortages are only at the cutting edge?
I once heard a major blockage was also the number of companies who had the skill to build machines that build semiconductors was low and therefore prices and leadtimes are high but again, someone who already makes high precision stuff building the lower size fabs must be doable?
> I can't imagine that buying or building machines to build op amps or voltage regulators isn't within the reach of lots of manufacturers.
The thing is, these things used to be built all over the place. However, consolidation, environmental concerns (the Silicon Valley is one of the US' most polluted areas due to decades of semiconductor manufacture) and market forces have destroyed lots of the capacity that used to exist, and now everyone is stuck and waiting. Not to mention that the people who operated the old fabs are gone for good, either in pension or in completely different careers.
Bringing the old fabs back or constructing new ones will take years. Capitalism is great for driving down costs in peace times, but usually the cost-saving comes at the cost (sic) of losing resilience in crisis times, and forget about "perfect storm" events like the one we're currently seeing with corona, coin miners, the shift towards electric mobility and the Chinese saber-rattling against Taiwan.
Seems like an issue of incentives at varying scales. A private business has an incentive to produce the quantity demanded and nothing more. Surplus is a cost, and capacity should be designed to meet, not exceed demand. Then, a stress appears on the system, and shortages occur.
A government has an incentive to avoid shortages of specific products and resources. In some industries, this leads to, essentially, a promise to purchase the supply the market does not (agriculture).
Does a similar model need to apply hereafter to certain technological inputs?
Actually capacity is designed to meet the highest demand and scale down for slow times. My company has seen lines go from 100% capacity and turning away customers one year to 20% capacity filling all orders. It is murder on our suppliers and so the most important question we ask when sourcing a new part is can the supplier keep their doors open if we make no orders for two years, yet still meet our peek demand? The reality of the business cycle is on everyone, though some see it worse than others.
I am also curious as to the chip demand by major cloud providers. COVID accelerated migration to the cloud, I have heard numbers from "cloud people" as much as 5x.
I can’t download the pdf but PCs are 15% of global semiconductor demand and the second largest electronics end market (after smartphones). After a decade of stagnant growth, PC sales are are now 40% above pre-pandemic levels. The semiconductor industry does not operate with a lot of slack, and this sort of step-change demand in a single large end market sucks up a lot of that slack capacity by itself. Then of course you have elevated demand and improved mix of overall goods, strong corporate investment cycle, low inventories entering the pandemic, and delayed semiconductor capacity investments from the trade war / inventory downcycle in early 2019 through to late 2020…
78 comments
[ 4.7 ms ] story [ 173 ms ] threadEven 30-40 years old 2nd hand equipment is completely impossible to find now.
Equipment once sold for scrap, or simply thrown away is now sold for more that it was new.
I applaud the effort, but feel like it's just going to be people taking advantage of the money for no net benefit.
However, one factor not mentioned in the study is that China, which had announced a plan for foundry independence, had its capacity expansion program disrupted by the trade war, as the US restricted China's ability to acquire key equipment to build out fabs.
As a result China could not build the plants it needed to meet its own consumption, forcing them to buy chips from the same vendors that everyone else is also trying to buy chips from. This is an additional shock which is hard to estimate in magnitude but it's probably not small as it represents the missing capacity of entire fabs for many months or years as they struggle to be built.
Thus while the US trade policy may have effectively hamstrung China's plans for foundry independence in the short term, it may also have contributed to a global shortage of semiconductors and its own inflation problems as collateral damage. Trade wars are tricky: without careful and long term policy planning there is a high risk of even more collateral damage.
Another factor that could also be contributing specifically to the difficulties of the automotive supply chain is that China internally had made electric vehicles a strategic technology focus, which means that domestically China's demand for automotive-grade semiconductors was set to surge. Even before the pandemic I was getting tips from buyers that power semiconductors, particularly those in SiC and GaN, were going to be in short supply, due to that shift.
edit - emphasized original paper
If fault must be found with the US it is in the consolidation and outsourcing of manufacturing to a small number of manufacturers, removing competitors and excess capacity from the supply chain. China's part has simply been to grow; creating consumers and thus demand for products that it had previously not consumed.
It is fairly clear that we'd still be in trouble absent the "easy to win" trade war against China, though. COVID alone would have been enough to upset the proverbial apple cart.
Surely at this point if not before you can just call them US sanctions, why party-politicise it?
Just my 'two pence' as an outsider!
Keeping history within view helps to view current events in perspective.
Reagan was ridiculed as well. He did a pretty good job at guiding the end of the cold war into a remarkably non-eventful event ( this could be worded better, not a native speaker ).
Even if the Biden administration were to say fine - Huawei, you can produce your chips at TSMC, and SMIC, you can buy an EUVL machine from ASML. China would still persue a completely US influence free semiconductor supply chain, as the US has clearly demonstrated its willingness weaponize its dominance.
From a Chinese perspective, the interest of state and its big unruly tech companies are now aligned. Alibaba, Xiaomi, etc. are all making chips now.
It is not all given that the escalated tech war is a longterm winning move from a US perspective.
But time will tell.
This is what bothers me most about 'tough on China' folks, whats the next move?
Foreign policy has to think few steps ahead. This is not looking much better than 'invade X country in middle east and figure out what to do next later"
Two steps ahead is bringing critical manufacturing back under one's control. Guess who wanted that?
That's because American price-controls and anti-price-gouging laws (and chilling effects from fear of them) don't really apply to foreign importers to the US, only to domestic companies. US lawmakers haven't quite gotten around to legislating world market prices yet after all.
So the prices of imported goods could rise to clear the markets. For domestically produced goods you were more likely to see empty shelves, lines and (informal) rationing; like super markets only selling you limited quantities.
The previous president, (not naming him just to keep things a bit apolitical) also had tons of bad ideas. A huge amount of the US advanced tech industry (such as semiconductors or EVs) depends on foreign talent, and publishing a Stephen Miller drafted Executive Order (No. 13940 - August 2020) with the stated purpose of disallowing any visa holders to work in a government contractor (which includes almost every tech company from Intel to Nvidia to Google) is a stupid move, and would have been suicidal.
Why is that "unrelated"? Chip fabs don't run themselves.
> It is not all given that the escalated tech war is a longterm winning move from a US perspective.
But the alternative of not doing anything gives no guarantees either.
Trump was not alone in his view that 'something had to be done about China' with regards to trade and IP ( WTO ). It already was a hairy situation for everybody in the West and some even stated that apparently it took 'a Trump' to confront this situation.
However it used to be that China's big tech (DJI, Xiaomi, Alibaba, Huawei etc.) had no problem being dependent on US-controlled tech.
Now they see it as an existential threat.
G. Orwell, In Front Of Your Nose
https://www.semiconductors.org/chinas-share-of-global-chip-s...
China’s Share of Global Chip Sales Now Surpasses Taiwan’s, Closing in on Europe’s and Japan’s
And true, some are fabless and the Chinese fabs fab for foreign customers as well.
The majority of semiconductor manufacturing is still in the 200nm range -- better yields, cheaper processes, more resilient chips. Some of it is still in the 1um (1 micron) range.
Of course, when demands returned, Chinese companies were still on buying spree. The shrewd middleman everywhere, not just in China, saw shortage coming, and they stock piling chips too.
Yeah, most semi-industry people in Taiwan think that US sanctions and the threat to impose more sanctions definitely contribute to the current shortage.
It will raise prices and hurt the economy, probably, so I can understand the reluctance. Nothing is easy.
What's better: A $1000 iPhone made in China that Americans can't afford because their jobs were globalized to China, or a $2000 iPhone made in USA that Americans can afford because they work in the iPhone factory or live in communities that are supported by it?
Page 5 alludes to this, but it does seem to be underrepresented in the report. Erratic policymaking on the US's part incentivized China to hoard semiconductors while they could still get them. For all they knew, their whole country was about to end up on the entity list.
How a fab that didn't exist yet can affect current demand?
US companies sold the US out when they moved our manufacturing base to china and handed them the future.
Now we should give up our negotiation leverage so the chip shortage lasts a bit shorter?
Chip shortage, or, better say, capacity shortage for affordable nodes been there for 5-7 years, and nobody really gave it much notice.
Lead times for average 180nm project were 12 months+ for last 5 years.
There is still a stratospheric gap in between "common man fab service," and the realm of multi-million dollar semi companies, who can afford service on relatively new, (and not decades old) equipment.
All significant chips which went into shortage were using that "common man fab process" on eighties, and nineties era equipment.
I don't think I agree with that. A LOT of small microcontrollers went into allocation and they were on 120nm, 90nm, and 65nm processes.
However, I do agree that there really was a shortage even before Covid and nobody cared. Effectively--nobody builds a line for an older node, ever. And there has been a lot of consolidation and sales of facilities such that very few people have been building fabs below 90nm.
The number of fabs built in the 2010's that aren't TSMC, Intel, or Samsung is pretty minimal.
65nm? 90nm? 180nm? I'd argue that the processes can be improved and cost can probably come down compared to the older machines.
I'm old enough to remember when 180nm chips were the fastest ones.
This is really dubious for two reasons. First, you can do a die shrink and fit many, many more equivalent chips onto a single wafer at a modern node. This is somewhat mitigated by the high cost of mask sets at modern nodes -- but there's just no way it would ever make economic sense to build capacity at 180nm.
Second, the equipment for old nodes is not made anymore. It's cheap, but that's only because there's ample supply on the secondary market. If fabs began to get restarted in earnest at old nodes, that supply would quickly evaporate. Then you would have to buy modern (expensive!) equipment -- at which point why not take advantage of its full potential? Wafer throughput is the same regardless of whether you're making 100 chips per wafer at 180nm or 10,000 at 22nm.
So, if your market is <1M dies a year, you have to price your chip >$1 just to break even. And in the car chip market, even 1M dies is a giant amount.
Car market was all about decades old chips, made to stringent automotive specs, manufactured in tiny batches, and sold on long term contracts at sub-10 cent per packaged, tested, and binned die (and low bin chips were usually simply thrown away.)
No, this is not the case. If you have a small run, the foundry will "consolidate" your chip with other customers to make a single mask. This makes sense all around, since the foundry would rather not have to deal with a mask set that has only a few wafer starts anyway.
Certainly the 14nm fab that Intel promised lately[1] for automotive demand would not entertain that sort of thing.
[1] https://www.electronicdesign.com/markets/automotive/article/...
Yes, I agree, but what I'm saying is that there's a cost optimum that might not necessarily be at 22nm (you get more chips per wafer but if the machines cost 10x as much, let's say, then it's not necessarily worth it)
I'm not even sure everything can be shrunk that easily. Power ICs for example.
That's sort of the point I'm making here: there is fundamentally no real cost differential unless you go really far back to when the equipment was really rudimentary. The old machines are cheap because they're used. If there was someone still manufacturing them, their cost would not be greatly different from modern equipment.
The best analogy I can make is an old car. Being older doesn't mean it's a whole lot cheaper to manufacture. Maybe a little -- no fancy computers or direct fuel injection -- but you lose that benefit really fast because of the shitty fuel economy.
Most of the modern equipment is "backward compatible"; modern CVD or PVD or CMP or etch equipment would work fine at any trailing node, with the exception of the fact that they would work only with 300mm wafers (which is actually an added cost efficiency).
The biggest difference I could imagine is with litho equipment. 193nm immersion is expensive and EUV is even more outrageous. You could probably use old non-immersion litho technology with modern multiple patterning instead.
Small relative increases in fuel economy don't save you that much money, either
UMC, Taiwan's No. 2 fab, offered to build a new 64nm line in an existing fab in exchange for a $1B firm order commitment. Nobody bit, they preferred instead to lean on the Biden administration to pressure the Taiwanese government to allow them to skip the line.
A new fab on old process nodes is not viable if it has to compete against fully amortized hand-me-down fabs that used to be cutting-edge 10-20 years ago.
No, not it isn't. That assertion is fundamentally wrong at many levels. It seems you're confusing the intro to economics definition of equilibrium between supply and demand with your own definition of shortage. However, in the process you're making the mistake of believing somehow that pricing out demand from the market and being unable to scale up production to meet demand does not involve any shortages of any kind.
Think about it for a second: I'd suddenly food prices skyrocketed to the point a single 1kg bag of rice sold for over €100, and most of the world was thus unable to buy food for being priced out of the market, would that represent a shortage even if some people could still afford it?
For microchips, it's likely the prices have not increased enough to kickstart this increase in supply.
And the time scale to establish new capacity from scratch is many years to a decade or more. Look at the capacity build up in China. It has taken decades, and that is with all the power of the nation, and in some cases loose on regulations regarding IP rights.
It simply isn't worth it to manufacture these chips without selling billions of units. Not to mention the development, even the small chips today are extremely advanced from a technical perspective.
Opening a lemonade stand is a far more solid business plan. Since owning know how is somehow not worth anything despite heavy handed patent laws, I think the business case is just bad. China seems to work against this with specifically funding establishment of fabs.
You gave a right example, to prove the current problem... the prices are high now... go build a nuclear energy heater to grow rice, I'll wait.
The prices for chips are high now, the process of building a production facility is very very expensive, and takes a lot of time, and the people building them (and quite a few are being built right now), are still unsure if the high prices will hold, to make it worthwhile.
GPU shortages are driven by a number of factors, and almost none of them are to do with component production (or wafers). There's a big scalper problem for GPUs, where some assholes will buy every GPU they can find and then resell them at a huge markup to either (increasingly desperate) end users and system integrators, or, increasingly, to cryptofuckers who will pay much more than legitimate users to ride their bubble. This means that even if there's plenty of production, unless the production can outspend the scalpers' total capital, it's all going to be swallowed up.
There's no general wafer shortage, but the highest grade wafers that get used in the highest end production have long lead times (because the fabs that use them reserve them way ahead of time). Wafer availability is not a bottleneck since the smallest nodes have such low capacity to start with. What has happened is that wafers have gone up in price in response to the general stockpiling behavior, but the high end stuff (including high-end GPUs) don't care as the wafer cost is basically negligible to them compared to all the process costs.
Business is good otherwise we are not in danger of going out of business. But for people in production this is a cut of their income. We did get some supplies though and are expected to restart full production the week after next. Still crazy it happened at all. We never had a supply problem like this. We certainly will try to eliminate the bottleneck though.
I assume that this would allow the advanced manufacturers to concentrate on the newest tech unless the shortages are only at the cutting edge?
I once heard a major blockage was also the number of companies who had the skill to build machines that build semiconductors was low and therefore prices and leadtimes are high but again, someone who already makes high precision stuff building the lower size fabs must be doable?
The thing is, these things used to be built all over the place. However, consolidation, environmental concerns (the Silicon Valley is one of the US' most polluted areas due to decades of semiconductor manufacture) and market forces have destroyed lots of the capacity that used to exist, and now everyone is stuck and waiting. Not to mention that the people who operated the old fabs are gone for good, either in pension or in completely different careers.
Bringing the old fabs back or constructing new ones will take years. Capitalism is great for driving down costs in peace times, but usually the cost-saving comes at the cost (sic) of losing resilience in crisis times, and forget about "perfect storm" events like the one we're currently seeing with corona, coin miners, the shift towards electric mobility and the Chinese saber-rattling against Taiwan.
A government has an incentive to avoid shortages of specific products and resources. In some industries, this leads to, essentially, a promise to purchase the supply the market does not (agriculture).
Does a similar model need to apply hereafter to certain technological inputs?