Intel, Nvidia, TSMC Execs Agree: Chip Shortage Could Last Into 2023 (arstechnica.com) 52
How many years will the ongoing chip shortage affect technology firms across the world? This week, multiple tech executives offered their own dismal estimates as part of their usual public financial disclosures, with the worst one coming in at "a couple of years." From a report: That nasty estimate comes from Intel CEO Pat Gelsinger, who offered that vague timeframe to The Washington Post in an interview on Tuesday. He clarified that was an estimate for how long it would take the company to "build capacity" to potentially address supply shortages. The conversation came as Intel offered to step up for two supply chains particularly pinched by the silicon drought: medical supplies and in-car computer systems.
In previous statements, Gelsinger pointed to Intel's current $20 billion plan to build a pair of factories in Arizona, and this week's interview added praise for President Joe Biden's proposed $50 billion chip-production infrastructure plan -- though Gelsinger indicated that Biden should be ready to spend more than that. TSMC CEO C.C. Wei offered a similarly dire estimate to investors on Thursday, saying that the Taiwan-based company hoped to "offer more capacity" for meeting retail and manufacturing demand "in 2023." TSMC, coincidentally, is moving forward with a manufacturing plant of its own in Arizona, which Bloomberg claims could cost "up to $12 billion," despite the company clarifying that it intends to prioritize research, development, and production in its home nation.
In previous statements, Gelsinger pointed to Intel's current $20 billion plan to build a pair of factories in Arizona, and this week's interview added praise for President Joe Biden's proposed $50 billion chip-production infrastructure plan -- though Gelsinger indicated that Biden should be ready to spend more than that. TSMC CEO C.C. Wei offered a similarly dire estimate to investors on Thursday, saying that the Taiwan-based company hoped to "offer more capacity" for meeting retail and manufacturing demand "in 2023." TSMC, coincidentally, is moving forward with a manufacturing plant of its own in Arizona, which Bloomberg claims could cost "up to $12 billion," despite the company clarifying that it intends to prioritize research, development, and production in its home nation.
So WTF happened? (Score:2)
Did the demand skyrocket or what? Where'd this multi-year backlog come from?
So WTF happened?-555. (Score:5, Funny)
Shortage of 555 timers for vibrators.
Re: (Score:2)
Available in astable or monostable multivibrator mode.
Re:So WTF happened?-555. (Score:4, Informative)
Shortage of 555 timers for vibrators.
As somebody who's built several vibrators and other sex toys over the last year, and in contact with others doing the same, I can tell you that's bullshit. We're mostly using ESP32s. :)
Re: (Score:1)
The 555 timer is really a piece of crap. There is not anything important you can do with it that you can't do better with the cheapest dual opamp like a 1458. However, that means you have to know what you're doing, which is optional with the 555.
Re: (Score:2)
Re: (Score:2)
I don't think you understand how supply/demand works [vgcats.com]. (SFW)
Re: (Score:3)
Cellphones.
Re: (Score:2)
Re:So WTF happened? (Score:4, Funny)
My money is on the plague of frogs that hit the Willamette Valley last year.
Re: (Score:2)
Re: (Score:3)
The problem isn't the big chips. It's the little jellybean ones that everyone uses but doesn't give a second thought to supply. I mean, normally those things are stocked by the millions so if you need them, anyone can send them to you overnight.
It's like say, 1K or 10K ohm resistors. Everyone expects to be able to use them anywhere freely - I mean, you don't ever expect the world
Re: (Score:2)
Politics superseding science?
Of course I'm basing this on the constant stream of pictures from such manufacturers showing the workers in clean-room 'bunny suits' and wearing face-masks.
Is that fake?
Re: (Score:2)
The best place to invest the money in the USA.... would be to pay for Global Foundries to build 7, 5 and 3 NM fabs... Intel has enough money to do whatever they want, TSMC has no intention of building anything state of the art in the USA.
Re:So WTF happened? (Score:5, Informative)
TSMC has no intention of building anything state of the art in the USA.
TSMC is building a 5 nm fab in Arizona [tomshardware.com].
JIT Happened (Score:5, Insightful)
Moreover nobody wants to risk building out to meet demand until they're absolutely 111% positive that demand will be there in a few years, because if you mess up and lose money Wall Street will eat you alive next quarter and you'll have a Vulture Capitalist looking to buy you up to extract anything of value from your corpse for themselves.
TL;DR; it's a symptom of a deeply dysfunctional economy.
Re: (Score:2)
Yes demand skyrocketed among other factors. If we are talking about PCs and components, CoVID quarantine meant more people were spending time at home. But now they needed computers that could do video conferencing while doing other work. I would speculate that while people had PCs before CoVID, many of them may have been older and not adequate. Also their PCs may not have been adequate for gaming as entertainment.
Other factors mean that the supply has been disrupted. Global shipping is still a mess. There w
Re: (Score:3)
The biggest factor is that the car companies decided to stop buying those chips because people weren't buying cars, and because they stopped buying them, the chip vendors didn't feel the need to keep allocating manufacturing capacity for building them. And when they later decided that they needed parts... whoops! We'll build a new plant to meet your needs. It should be online in a year or two.
This isn't a shortage caused by excess demand. This is an entirely artificial shortage caused largely by the ab
Re: (Score:1)
Re: (Score:3)
If zero inventory were really that bad there would be a fortune to be made in warehousing and selling components.
You're making the mistake of assuming that what's best for consumers is what's best for companies' bottom line. Companies do zero-inventory because otherwise they get stuck with excess inventory that they have to dump at a loss. Consumers benefit from that excess inventory by being able to get what they want in a timely manner, and by being able to purchase the closeout inventory at a cheap price if they don't care about getting the latest and greatest version of a product. But it's hard to make more mon
Re: (Score:2)
Re: (Score:2)
Yes there could be a fortune in warehousing parts of someone had the premonition that a pandemic was about to occur. However that company would have inventory and warehouse costs for years so they could maybe make a fortune in a small window of time.
The thing is, there are small shortages of things all the time. Most people don't notice it too much, but I've gone for weeks without being able to buy certain food products just because the stores never have them in stock. It's not that there necessarily needs to be one company taking on the burden of huge amounts of warehousing, so much as that there's a need for just a bit more buffering at every level.
Also, for this specific case, the car companies knew that the downturn was temporary, and that they w
Re: (Score:1)
Excessive stock piling due to POTUS causing issues with China and compainies like Huewei, trade, use of tech, etc...
Now everyone wants 3 to 6 months stock, rather than just-in-time.
Re: (Score:2)
Samsung (in S. Korea) handles about a fifth of the world's advanced ICs. Taiwan well over half. The scraps for everyone else amount to less than a quarter of the market.
The US and EU are quite capable of advanced fabrication. But currently not in the volumes or price that Taiwan and S. Korea can handle. Mainland China is still working on their foundries, but it would be much faster for China to invade Taiwan and claim in practice what they've already asserted is theirs in theory.
Re: (Score:2)
The US and EU are quite capable of advanced fabrication.
Citation needed.
Intel is the only western company that is even close, and Intel is way behind the transistor density that Samsung and TSMC routinely produce.
Re: (Score:2)
Re: (Score:2)
Depends on your definition of advanced.
I TOTALLY agree. If you are willing to redefine "advanced" to mean years behind the completion in both technology and cost-effectiveness with no plausible plan to catch up, then the EU is certainly "advanced". Outer Mongolia, even more so.
Re: (Score:2)
Re: (Score:2)
IBM, for example, was producing 5nm GAAFET test chips in 2015.
The "US and EU" are not behind in fabrication technology any more than they're behind in rare earth metal mining technology.
Re:Lawlz (Score:4, Insightful)
STMicro is producing 28nm and 32nm in France.
The educational NanoFab 300 in New York is also down to 22nm.
GlobalFoundries recently started 12nm and 14nm production in New York.
Samsung's S2-Line is at 14nm or 11nm in Texas, but they're producing a fairly wide range of node sizes at that facility.
The above is not an exhaustive list, I picked ones that are easy to look up. And of course "advanced" means more than simply smallest process node. There are some niche applications in communication, aerospace, and defense that the smaller domestic foundries handle that is difficult to do elsewhere.
To answer the unstated question: Is there any place in the US or EU that can produce the equivalent of TSMC's 4nm node by 2022?
No. There's not way this can happen. Not in any volume. Even if we invested billions right now into building it, it won't be there by 2022. Without TSMC and Samsung, mobile chip developers like Qualcomm and Apple would have to either hold off and continue down the roughly 14nm for mobile chips, and on top of that take a huge hit on volume and price.
On the other hand you can build CPUs and GPUs and super computers with almost any process node. Architecture matters more. Of course if you based your new super computing cluster around AMD Epyc 2, but they can't deliver them to you because 7 nm isn't available. Then you either wait for it to become available, or they would have to redesign for a different process node (which they'd not do, unless something catastrophic happened to all of TSMC). Either way: long way & shortages.
Re: (Score:2)
No. There's not way this can happen. Not in any volume. Even if we invested billions right now into building it, it won't be there by 2022. Without TSMC and Samsung, mobile chip developers like Qualcomm and Apple would have to either hold off and continue down the roughly 14nm for mobile chips, and on top of that take a huge hit on volume and price.
I disagree entirely.
Fabrication is a tough business to get into. The dominant market players are quite good. US firms are however at the very forefront of silicon fabrication technologies.
They're just not in the mass chip making business.
I see no reason to believe they couldn't be if they had to.
Re: (Score:2)
Takes 18 months minimum to build a new facility. It's 2021.
If US or EU firms wanted to be domestically mass producing 5nm or smaller by some later date, such as 2025, I think with money is the biggest barrier. If the facilities can technically produce the product, but not produce it below the market price, then why would they ever turn on the foundry on let alone build it?
Re: (Score:2)
Takes 18 months minimum to build a new facility. It's 2021.
Na. You just can't make that claim.
It may "normally" take that, in the normal course of business for established foundries (who are in no great hurry)
If US or EU firms wanted to be domestically mass producing 5nm or smaller by some later date, such as 2025, I think with money is the biggest barrier.
I don't think money is any barrier at all should a loss of Asian supply occur.
If the facilities can technically produce the product, but not produce it below the market price, then why would they ever turn on the foundry on let alone build it?
Because with the loss of Taiwanese and Korean production, the "market price" is no longer relevant to the equation.
If we're to limit this to what the "US and EU" can do under the normal market conditions, then ya, nobody could (more to the point- would) do that.
But the conversa
Re: (Score:2)
This isn't to say that Intel's 10nm node doesn't have yield problems, but it's in most ways superior to TSMCs and Samsung's 7nm node.
Comment (Score:2)
Make some interventions in reducing demand (Score:4, Insightful)
Re: (Score:3)
Not just phones, computers too. That includes smartphones, tablets, laptops, desktops.
Taiwan is 1/2 of the capacity? (Score:1)
As a poster on Ars pointed out (Score:3, Insightful)
We are going to have to subsidize them because these shortages are a national security problem, but Enough is Enough.
Raise Corporate Taxes to 50%. Give regulators statutory leeway so they can close loopholes in real time. Hire economists to write the regulations. And *no* Goldman Sachs people. Draw your economists from the Universities. Carefully (they've been packing the Universities too lately)...
That way the money they're getting in subsidies comes out of their hides like it should.
They agree? (Score:1)
Yes, snd whenever you share your doubt guess what. (Score:1)
Golden opportunity to inflate prices (Score:3)
The same happened about 10 years ago when the floods in Thailand affected hard disk manufacturing.
https://www.zdnet.com/article/... [zdnet.com]
Re: (Score:2)
One of these three things is not like the others (Score:3)
As sessame street said.
Two of those people represent companies that actually MAKE semiconductors, while the third one represent a company which DESIGNS semiconductors, and contract the manufacturing to others...
Is up to you to decide who to listen to...
Maybe not such a bad thing? (Score:2, Interesting)
Perhaps this will encourage people to hold onto existing hardware longer, or seek out second hand hardware?
I was looking for an RX560 for my 2010 mac pro 5,1 - absolutely nothing available 'new'.
I paid just under the equivalent new price for a 2nd hand one on eBay - the amount of bidders was insane, so I just jumped in on the 'buy now' price.
The reality is, for most people, if we take video cards as an example, the 'life span' of a video card tends to be ridiculously short in terms of how long until you buy
Re: Maybe not such a bad thing? (Score:1)