Still stuck at 12nm? Are they really not going to spend any money on improving that? Tsms/samsung just need to undercut them a bit and they'll start losing customers.
Maybe they will license Samsung's 7nm or 5nm, just like they licensed the 14nm, but that would be expensive, especially with the need for EUV hardware.
Maybe there's a lot of money still in 12nm of various forms. Not everybody needs cutting edge, but you need to follow your competition at some point.
Every Ryzen chip sold has TSMC chiplet(s) and an GF I/O chip. Looking at the layout of the chip packages, the I/O chip isn't limiting the design--you need a lot of surface area for the I/O, so going denser with the logic doesn't buy you much more. Possilby, they could put more L3 on there, but they're surface area/IO bound on those chips. Why pay for a process that doesn't buy you anything?
Possibly it would make sense to make a more power efficient chip for mobile Ryzen parts? But at that point, might as well integrate the I/O onto the main die and be done with it.
Currently, every fab everywhere is at capacity--all the way back to micron sized features. Pay for capacity not process improvement right now. A year or two from now that may change, but there's always going to be a use for those older fabs.
AMD is about to ditch 12nm and use TSMC's 6nm for Rembrandt and I/O dies. But they could also use GloFo's newer 12LP+, e.g. for "Monet" quad-core Zen 3 APUs. And AMD will probably continue making various 12nm parts for the embedded market.
GF's i/o chips take up quite a bit of power. In chips with massive amounts of cores like milan it can take up as much power as the cores themselves. I doubt they will be able to continue scaling IPC and/or core count if they stick with global foundries, at least at the enterprise level.
The I/O chips use the power they do mostly because of the interfaces they provide and that will not vary with process geometry. It takes a certain amount of power to run a PCI-E 4.0 bus, same for SDRAM, etc. The only amount that varies a bit is the logic behind those protocols and the cache. Cache is mostly leakage and that doesn't improve with smaller processes.
There's not much to gain with a smaller process geometry and it will increase the cost of the I/O dies.
If that were the case the I/O wouldn't take so much power compared to similar core count icelake chips.
Anandtech already did review on them and the I/O uses a significant amount more power than chips with the same platform features, and another revision where they were able to improve performance with more power delivery, so even current cores have performance headroom but are held back by package power.
That's definitely not the case for the high volume desktop processors. 95W is easily achievable, so the mere extr 5W IOD cost can't be limiting anything.
Similarly for higher performance pieces 115W isn't even an issue. Definitely not limited by the IOD. Not to mention next gen 12LP+ is improving upon 12LP.
Mobile Ryzen is already a monolithic die for power reasons, but I think AMD have now hit the point where a shrink of the IO die would benefit them, in the Server arena at the very least.
The I/O die may be the reason for the problems getting the Infinity Fabric speed up to 2GHz or above. The APUs(4000G and 5000G) are rumored to be able to hit 2.2-2.4GHz for the IF speed, and those are monolithic processors without the I/O die, so are all TSMC without Global Foundries.
I appreciate the need to save TSMC fab process for the CPU core manufacturing, but even using Global for the I/O die is holding AMD performance back.
How much of that is conservative management that is afraid of doing an INVESTMENT in R&D. Sure, there is a lot of expense, but once they get it working, the payoff would be huge.
GF invested in a 7nm process but had few customers so they cancelled it. Not surprising, no amount of investment is going to make designing around a niche logic node that might not exist in a year a good business decision vs playing it safe with TSMC.
Perhaps the hot market now will allow them to pick that up, but they'd be behind in the game. The rumored buyout by Intel might benefit them both in that regard, with Intel getting more capacity, and GloFo getting back in on modern process technology (despite Intel's massive struggles, they at least have things in the pipeline).
A lot of money to be made in that size, you don't need to be cutting edge for most uses- cars, stoves, etc. The world doesn't have enough capacity right now, so this is just trying to fill that hole.
"Surely there's economies of scale, though? Smaller process node means more chips to sell?"
therein lies the Capitalist's Conundrum: the Tyranny of Fixed Cost. the more capital intensive the production process, the higher the fixed cost of production, which means lower flexibility of output. the higher the fixed cost, the only way to lower average cost is to boost output, spreading the fixed cost over more units. if demand slackens, even a teeny bit, average cost jumps, and then what do you do? firing employees buys you little to nothing. the notion that *all* production, globally, needs to be at Xnm is false. most embedded chips still run on double digit nodes, and likely always will. because the capital is largely, if not entirely, amortized, producers' average cost is much lower. cutting edge gaming and whatever-edge AI are the outliers, not the norm.
"Surely there's economies of scale, though? Smaller process node means more chips to sell?"
Around 22nm, cost/transistor started to go back up as process scales continued to shrink. If you;re building a chip that is not operating right at the bleeding edge of performance, then shrinking your design to a newer smaller process means cost per chip goes UP, not down. And there are a LOT of ICs that aren't CPUs from disc drive controllers to the humble 555 timer.
The other factor here is that 450mm wafers have been "a few years out" for over a decade now. It's possible they've been holding out capital for a 450mm upgrade that never got to play out.
Half of my tools are 450 capable. The others are 300 only.
There is obviously 100% demand for everything they make, but one strategy would be to let the node scaling run out, wait for beyond-EUV lithography to be deployed, etc. and then jump down to the lowest nodes at a fraction of the costs paid by TSMC/Samsung. Could happen in a decade.
Also, if Intel buys GlobalFoundries, technically they will no longer be stuck at 12nm.
I don't think GF is going to target the bleeding edge performance silicon market. Vehicle ECU's are still 20nm designs and that is where the chip shortage is.
Plenty of chips being made in greater than 14nm size, so GloFo is making money.
Now, if they truly are trying to capitalize on the current insane demand & move their IPO up, then that is spelling out "not available right now" to Intel buying them.
Unfortunately for those of us in the US of A, that confused old man is currently the US Senate's Majority Leader, "Chuck" Schumer (D), NY. Not that his Republican predecessor (McConnell), Senate Majority Leader until December 2020, was any brighter or otherwise preferable. The Chinese Leadership can rest easy knowing that this is the current state of politics in these United States.
Politicians are where they are because they are good at political intrigues and excel in their little bubble. Never underestimate their incompetence in reality.
Career criminal Schumer likely doesn't actually know the difference between a potato chip and a semi-conductor chip. After ignoring the exportation of millions of U.S. jobs to China and allowing Foxconn/Apple to use slave and child labor to produce iPhones that sell for $1000+ in the U.S., now these DC parasites like drunks are doling our tax payer money to all. It's easy to spend someone elses money when you're living on the dole all your adult life.
Building or expanding U.S. fabs doesn't result in most of the revenue staying in the U.S. economy either so tax payers should not be endowing multi-national companies who already reap fortunes from industry monopolies. These companies should be required to manufacture the goods they sell in the country they sell it in, pay fair labor prices, provide decent employee working conditions and be environmentally responsible. Anything less is exploitation and it's been ongoing for decades.
> now these DC parasites like drunks are doling our tax payer money to all.
It's a lot of money for the industry, but not actually that much in terms of federal spending. There's tons of crap the federal government subsidizes. Do you get this worked up over tax breaks for oil exploration?
More to the point: how do you expect the government to get more domestic semiconductor production? Do you really want an import ban on chips to crash the economy for however long it takes someone to build fabs? An approach involving subsidies is the only one that's politically feasible. I doubt you'll find many successful examples of other models, anywhere else in the world. Like it or not, subsidies is how this sort of thing gets done. The far bigger problem is that big corporations pay virtually no tax on their profits.
> allowing Foxconn/Apple to use slave and child labor to produce iPhones
It's not only iPhones. Most manufacturing is in China, or somewhere with similar labor & environmental practices. Don't act like you haven't benefitted from it, too.
> It's easy to spend someone elses money when you're living on the dole all your adult life.
All politicians spend other people's money. Whether it's through tax breaks or subsidies, it amounts to the same thing. The only difference is which constituency gets the benefits.
> Building or expanding U.S. fabs doesn't result in most of the revenue staying in the U.S.
It's not only about revenue, but also for strategic reasons. Otherwise, when China takes back Taiwan and threatens to cut off the US from TSMC, we're completely at their mercy.
> companies should be required to manufacture the goods they sell in the country they sell it in
That's insane. You can't just unwind centuries of globalization. If you even tried, it would certainly crater the economy like nothing we've ever seen.
> pay fair labor prices, provide decent employee working conditions > and be environmentally responsible.
I always get confused when people blame the politicians in capitalist states for the things capitalism does (e.g. outsourcing labour when local labour gets basic rights, bilking the state for subsidies). It usually accompanies some other form of pseudo-Libertarian rhetoric, though, which entails a kind of specific blindness for the problems inherent to capitalist systems.
With all these manufacturers (TSMC, Samsung, Intel, some other non-bleeding edge fabs whom I forget, and now GF) announcing huge expansion plans, I'm just wating for the semiconductor glut in 3 years and all the fabs complaining about too much supply driving prices down too much so they start losing money ...
Soon it'll look like the memory industry that's on a 4-year oversupply (low prices)/undersupply (high prices) cycle.
it could be sooner. it's going to dawn on folks that 5G is the end of the line for cellular. everyone, even Verizon, "cannea change the laws of physics, Capn!" it makes no difference that ever smaller nodes could fit some hypothetical 6G radio in a frame, there would have to be a 'cell' every line-of-sight 30 feet away, using bandwidth that may not even exist. phones have been the growth driver for a couple decades. some may say that Moore's Law will always pertain, but using such devices in practical devices may not.
The big wildcard in the semiconductor supply/demand situation is when China starts to approach parity with the leading fabs.
Otherwise, I think semiconductor demand has gotten very difficult to predict, in recent years. I remember there was supposed to be a cooling-off sometime around 2017-2019 that never happened. Blame it on mining or AI, but I don't think there's any one cause.
What is that utter trash logo damn. Old one made so much perfect sense, A globe with silicon circuitry under microscope. Now it's all gone thanks to this bs flat design.
The Intel rumor crashed and burned, as predicted...;) People need to learn how to separate the wheat from the chaff insofar as rumor is concerned. If something sounds silly and nonsensical it probably, very likely, is.
$30B is less than the cost of GF's fabs themsleves. Intel's offer is a ridiculous lowball and Mubadala would be idiotic to accept it. They'll easily get twice that in the IPO - Mubadala's target should be $100B, their lowest acceptable offer ~$70B.
How exactly is it not? I guess the only thing that doesn't wash is that the State should get some rights to the IP, were it a real partner. That's usually what happens in private partnerships.
> The ‘bottom 90%’ of the public is completely invisible to Congress
Non-sequitur. And certainly not true of the pandemic-related benefits passed in both last year and this.
Also, if you look at the tax code, the lower half of the US population does quite well. Even the middle-income folks get benefits like child tax-credits and mortgage interest deductions.
> Public/private partnership is a euphemism
As long as the public gets more benefits than it would otherwise accrue, it's not. It's in the long-term strategic interest of the US public to maintain a domestic semiconductor manufacturing capability. That's the public interest. Another is to benefit other US industries, such as automotive manufacturing.
As I said, I'd like for the public to get a greater return on its investment, either in financial or other terms, but there's indeed a legitimate public interest.
It's almost like you're challenging yourself to see how cynical you can be. My sympathies for whatever misfortune has embittered you so. That's a very sad and one-dimensional way of looking at the world. There's not such a deficit of sadness in the world that we need to search it out so fastidiously.
I don't think so. I am not strictly partisan. I just try to get past the reflexive cynicism that seems to be in vogue, and actually engage my brain.
You call yourself "Oxford", but I sense a degree of anti-intellectualism that reminds me of the sportsball players I knew in high-school. It's like you've forgotten how to look at issues on their merits and challenge extreme positions by searching for counter-examples. That's not a refusal to take a position on things, but I try to take *measured* positions and not get too calcified in my views.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
74 Comments
Back to Article
shabby - Monday, July 19, 2021 - link
Still stuck at 12nm? Are they really not going to spend any money on improving that? Tsms/samsung just need to undercut them a bit and they'll start losing customers.psychobriggsy - Monday, July 19, 2021 - link
Maybe they will license Samsung's 7nm or 5nm, just like they licensed the 14nm, but that would be expensive, especially with the need for EUV hardware.Maybe there's a lot of money still in 12nm of various forms. Not everybody needs cutting edge, but you need to follow your competition at some point.
dwillmore - Monday, July 19, 2021 - link
Every Ryzen chip sold has TSMC chiplet(s) and an GF I/O chip. Looking at the layout of the chip packages, the I/O chip isn't limiting the design--you need a lot of surface area for the I/O, so going denser with the logic doesn't buy you much more. Possilby, they could put more L3 on there, but they're surface area/IO bound on those chips. Why pay for a process that doesn't buy you anything?Possibly it would make sense to make a more power efficient chip for mobile Ryzen parts? But at that point, might as well integrate the I/O onto the main die and be done with it.
Currently, every fab everywhere is at capacity--all the way back to micron sized features. Pay for capacity not process improvement right now. A year or two from now that may change, but there's always going to be a use for those older fabs.
nandnandnand - Monday, July 19, 2021 - link
AMD is about to ditch 12nm and use TSMC's 6nm for Rembrandt and I/O dies. But they could also use GloFo's newer 12LP+, e.g. for "Monet" quad-core Zen 3 APUs. And AMD will probably continue making various 12nm parts for the embedded market.whatthe123 - Tuesday, July 20, 2021 - link
GF's i/o chips take up quite a bit of power. In chips with massive amounts of cores like milan it can take up as much power as the cores themselves. I doubt they will be able to continue scaling IPC and/or core count if they stick with global foundries, at least at the enterprise level.dwillmore - Tuesday, July 20, 2021 - link
The I/O chips use the power they do mostly because of the interfaces they provide and that will not vary with process geometry. It takes a certain amount of power to run a PCI-E 4.0 bus, same for SDRAM, etc. The only amount that varies a bit is the logic behind those protocols and the cache. Cache is mostly leakage and that doesn't improve with smaller processes.There's not much to gain with a smaller process geometry and it will increase the cost of the I/O dies.
FreckledTrout - Tuesday, July 20, 2021 - link
One could integrate the chipset into the IO die to keep the surface area high.dwillmore - Tuesday, July 20, 2021 - link
These chips are SoCs. The chipset *is* integrated into the I/O die. At least on the desktop parts. I don't know about the server ones.whatthe123 - Tuesday, July 20, 2021 - link
If that were the case the I/O wouldn't take so much power compared to similar core count icelake chips.Anandtech already did review on them and the I/O uses a significant amount more power than chips with the same platform features, and another revision where they were able to improve performance with more power delivery, so even current cores have performance headroom but are held back by package power.
dotjaz - Wednesday, July 21, 2021 - link
That's definitely not the case for the high volume desktop processors. 95W is easily achievable, so the mere extr 5W IOD cost can't be limiting anything.dotjaz - Wednesday, July 21, 2021 - link
Similarly for higher performance pieces 115W isn't even an issue. Definitely not limited by the IOD. Not to mention next gen 12LP+ is improving upon 12LP.Spunjji - Tuesday, July 20, 2021 - link
Mobile Ryzen is already a monolithic die for power reasons, but I think AMD have now hit the point where a shrink of the IO die would benefit them, in the Server arena at the very least.Targon - Tuesday, July 20, 2021 - link
The I/O die may be the reason for the problems getting the Infinity Fabric speed up to 2GHz or above. The APUs(4000G and 5000G) are rumored to be able to hit 2.2-2.4GHz for the IF speed, and those are monolithic processors without the I/O die, so are all TSMC without Global Foundries.I appreciate the need to save TSMC fab process for the CPU core manufacturing, but even using Global for the I/O die is holding AMD performance back.
haukionkannel - Tuesday, July 20, 2021 - link
Too expensive to go 7nm… that is why not.Targon - Tuesday, July 20, 2021 - link
How much of that is conservative management that is afraid of doing an INVESTMENT in R&D. Sure, there is a lot of expense, but once they get it working, the payoff would be huge.saratoga4 - Wednesday, July 21, 2021 - link
GF invested in a 7nm process but had few customers so they cancelled it. Not surprising, no amount of investment is going to make designing around a niche logic node that might not exist in a year a good business decision vs playing it safe with TSMC.fcth - Monday, July 19, 2021 - link
They basically gave up on getting smaller, killing their 7nm process and abandoning their 5nm and 3nm research: https://www.anandtech.com/show/13277/globalfoundri...Perhaps the hot market now will allow them to pick that up, but they'd be behind in the game. The rumored buyout by Intel might benefit them both in that regard, with Intel getting more capacity, and GloFo getting back in on modern process technology (despite Intel's massive struggles, they at least have things in the pipeline).
Papaspud - Monday, July 19, 2021 - link
A lot of money to be made in that size, you don't need to be cutting edge for most uses- cars, stoves, etc. The world doesn't have enough capacity right now, so this is just trying to fill that hole.Kushan - Monday, July 19, 2021 - link
Surely there's economies of scale, though? Smaller process node means more chips to sell?I agree it does make business sense, it just seems weird that they're not investing at all.
FunBunny2 - Monday, July 19, 2021 - link
"Surely there's economies of scale, though? Smaller process node means more chips to sell?"therein lies the Capitalist's Conundrum: the Tyranny of Fixed Cost. the more capital intensive the production process, the higher the fixed cost of production, which means lower flexibility of output. the higher the fixed cost, the only way to lower average cost is to boost output, spreading the fixed cost over more units. if demand slackens, even a teeny bit, average cost jumps, and then what do you do? firing employees buys you little to nothing. the notion that *all* production, globally, needs to be at Xnm is false. most embedded chips still run on double digit nodes, and likely always will. because the capital is largely, if not entirely, amortized, producers' average cost is much lower. cutting edge gaming and whatever-edge AI are the outliers, not the norm.
edzieba - Tuesday, July 20, 2021 - link
"Surely there's economies of scale, though? Smaller process node means more chips to sell?"Around 22nm, cost/transistor started to go back up as process scales continued to shrink. If you;re building a chip that is not operating right at the bleeding edge of performance, then shrinking your design to a newer smaller process means cost per chip goes UP, not down. And there are a LOT of ICs that aren't CPUs from disc drive controllers to the humble 555 timer.
JKflipflop98 - Wednesday, July 21, 2021 - link
The other factor here is that 450mm wafers have been "a few years out" for over a decade now. It's possible they've been holding out capital for a 450mm upgrade that never got to play out.Half of my tools are 450 capable. The others are 300 only.
nandnandnand - Monday, July 19, 2021 - link
12LP+ sounds like a big improvement.There is obviously 100% demand for everything they make, but one strategy would be to let the node scaling run out, wait for beyond-EUV lithography to be deployed, etc. and then jump down to the lowest nodes at a fraction of the costs paid by TSMC/Samsung. Could happen in a decade.
Also, if Intel buys GlobalFoundries, technically they will no longer be stuck at 12nm.
Hyper72 - Tuesday, July 20, 2021 - link
Probably half the market is still 28-68nm. 12/14nm have a long future ahead of it.Zingam - Wednesday, July 21, 2021 - link
Some make shovels others excavators. Both dig holes.Samus - Wednesday, July 21, 2021 - link
I don't think GF is going to target the bleeding edge performance silicon market. Vehicle ECU's are still 20nm designs and that is where the chip shortage is.romrunning - Monday, July 19, 2021 - link
Plenty of chips being made in greater than 14nm size, so GloFo is making money.Now, if they truly are trying to capitalize on the current insane demand & move their IPO up, then that is spelling out "not available right now" to Intel buying them.
Mobile-Dom - Monday, July 19, 2021 - link
TIL the GloFo fab in Malta is not, in fact in the country of Malta, but in NY state. ngl, I'm kinda disappointedWereweeb - Tuesday, July 20, 2021 - link
Hahaha, yeah, yank town names are sillycoburn_c - Monday, July 19, 2021 - link
Someone tell that confused old man these are not chips for eating.The Von Matrices - Monday, July 19, 2021 - link
I didn't know that Ian was that old.DaTanMan - Monday, July 19, 2021 - link
Dr. Cuttress would argue otherwise. ;)Shaunathan - Monday, July 19, 2021 - link
man are you telling me you never ate a processor before come on manphilehidiot - Monday, July 19, 2021 - link
My brother once swallowed a battery and I'm a fan of making my own oven chips? That count?eastcoast_pete - Monday, July 19, 2021 - link
Unfortunately for those of us in the US of A, that confused old man is currently the US Senate's Majority Leader, "Chuck" Schumer (D), NY. Not that his Republican predecessor (McConnell), Senate Majority Leader until December 2020, was any brighter or otherwise preferable. The Chinese Leadership can rest easy knowing that this is the current state of politics in these United States.mode_13h - Tuesday, July 20, 2021 - link
They didn't get where they are by accident. You underestimate them at your peril.And I didn't hear what he actually said, but the potato chips are clearly there as some sort of pun or rhetorical pivot.
smalM - Tuesday, July 20, 2021 - link
Politicians are where they are because they are good at political intrigues and excel in their little bubble. Never underestimate their incompetence in reality.Oxford Guy - Wednesday, July 21, 2021 - link
Not bubbles. Series of tubes.bobby Valentino - Tuesday, July 20, 2021 - link
The potato chip was invented in the town next door. It looks silly but it makes sense if you see the video. He understands the difference lol.waterdog - Monday, July 19, 2021 - link
"adjacency technologies", really?Ian Cutress - Monday, July 19, 2021 - link
Everything not on the bleeding edge logic is usually called 'adjacency', yes. That includes RF, High Voltage, Analog, MEMs, eNVM, photonic.webdoctors - Monday, July 19, 2021 - link
Must be some pretty sweet incentives to do another one in NY, can't wait to see them published.Arsenica - Monday, July 19, 2021 - link
As Cuomo wasn't in this event it most likely mean that GloFo is using the recently approved Federal incentives.eastcoast_pete - Monday, July 19, 2021 - link
+1. You couldn't have kept him away otherwise.bobby Valentino - Tuesday, July 20, 2021 - link
I would guess the Commerce Secretary who was there that she prob showed up with a gigantic check.Techie2 - Monday, July 19, 2021 - link
Career criminal Schumer likely doesn't actually know the difference between a potato chip and a semi-conductor chip. After ignoring the exportation of millions of U.S. jobs to China and allowing Foxconn/Apple to use slave and child labor to produce iPhones that sell for $1000+ in the U.S., now these DC parasites like drunks are doling our tax payer money to all. It's easy to spend someone elses money when you're living on the dole all your adult life.Building or expanding U.S. fabs doesn't result in most of the revenue staying in the U.S. economy either so tax payers should not be endowing multi-national companies who already reap fortunes from industry monopolies. These companies should be required to manufacture the goods they sell in the country they sell it in, pay fair labor prices, provide decent employee working conditions and be environmentally responsible. Anything less is exploitation and it's been ongoing for decades.
Tilmitt - Monday, July 19, 2021 - link
Welfare for communist dictatorship enabling multinational corporations loyal to nothing.mode_13h - Tuesday, July 20, 2021 - link
> now these DC parasites like drunks are doling our tax payer money to all.It's a lot of money for the industry, but not actually that much in terms of federal spending. There's tons of crap the federal government subsidizes. Do you get this worked up over tax breaks for oil exploration?
More to the point: how do you expect the government to get more domestic semiconductor production? Do you really want an import ban on chips to crash the economy for however long it takes someone to build fabs? An approach involving subsidies is the only one that's politically feasible. I doubt you'll find many successful examples of other models, anywhere else in the world. Like it or not, subsidies is how this sort of thing gets done. The far bigger problem is that big corporations pay virtually no tax on their profits.
> allowing Foxconn/Apple to use slave and child labor to produce iPhones
It's not only iPhones. Most manufacturing is in China, or somewhere with similar labor & environmental practices. Don't act like you haven't benefitted from it, too.
> It's easy to spend someone elses money when you're living on the dole all your adult life.
All politicians spend other people's money. Whether it's through tax breaks or subsidies, it amounts to the same thing. The only difference is which constituency gets the benefits.
> Building or expanding U.S. fabs doesn't result in most of the revenue staying in the U.S.
It's not only about revenue, but also for strategic reasons. Otherwise, when China takes back Taiwan and threatens to cut off the US from TSMC, we're completely at their mercy.
> companies should be required to manufacture the goods they sell in the country they sell it in
That's insane. You can't just unwind centuries of globalization. If you even tried, it would certainly crater the economy like nothing we've ever seen.
> pay fair labor prices, provide decent employee working conditions
> and be environmentally responsible.
Agreed.
Spunjji - Tuesday, July 20, 2021 - link
I always get confused when people blame the politicians in capitalist states for the things capitalism does (e.g. outsourcing labour when local labour gets basic rights, bilking the state for subsidies). It usually accompanies some other form of pseudo-Libertarian rhetoric, though, which entails a kind of specific blindness for the problems inherent to capitalist systems.Oxford Guy - Wednesday, July 21, 2021 - link
What capitalism?mode_13h - Thursday, July 22, 2021 - link
I know you love playing the cynic, but even with some degree of state-capture, it's still capitalism.Oxford Guy - Monday, July 26, 2021 - link
You forgot to insert word ‘called’ in there.mode_13h - Tuesday, July 27, 2021 - link
No, I said what I meant.There's free and open competition, for the most part. That's capitalism.
Oxford Guy - Wednesday, July 28, 2021 - link
Uh huh.bobby Valentino - Tuesday, July 20, 2021 - link
Lol, the potato chip was invented in the town next door, it was a worthy point he made. Thanks try again.boozed - Monday, July 19, 2021 - link
"Senator Schumer spoke about the need to pass grow semiconductors"Is this a typo? If not what does it mean?
Thunder 57 - Monday, July 19, 2021 - link
I was wondering the same thing. Nice name, BTW.eldakka - Monday, July 19, 2021 - link
With all these manufacturers (TSMC, Samsung, Intel, some other non-bleeding edge fabs whom I forget, and now GF) announcing huge expansion plans, I'm just wating for the semiconductor glut in 3 years and all the fabs complaining about too much supply driving prices down too much so they start losing money ...Soon it'll look like the memory industry that's on a 4-year oversupply (low prices)/undersupply (high prices) cycle.
Wereweeb - Tuesday, July 20, 2021 - link
Naturally, and then they'll start asking for tax cuts, subsidies, etc...They're expanding right now because governments are giving them the money to expand. Something something socialized losses, privatized profits.
FunBunny2 - Tuesday, July 20, 2021 - link
"the semiconductor glut in 3 years"it could be sooner. it's going to dawn on folks that 5G is the end of the line for cellular. everyone, even Verizon, "cannea change the laws of physics, Capn!" it makes no difference that ever smaller nodes could fit some hypothetical 6G radio in a frame, there would have to be a 'cell' every line-of-sight 30 feet away, using bandwidth that may not even exist. phones have been the growth driver for a couple decades. some may say that Moore's Law will always pertain, but using such devices in practical devices may not.
mode_13h - Thursday, July 22, 2021 - link
The big wildcard in the semiconductor supply/demand situation is when China starts to approach parity with the leading fabs.Otherwise, I think semiconductor demand has gotten very difficult to predict, in recent years. I remember there was supposed to be a cooling-off sometime around 2017-2019 that never happened. Blame it on mining or AI, but I don't think there's any one cause.
Silver5urfer - Tuesday, July 20, 2021 - link
What is that utter trash logo damn. Old one made so much perfect sense, A globe with silicon circuitry under microscope. Now it's all gone thanks to this bs flat design.GeoffreyA - Tuesday, July 20, 2021 - link
I think the text looks nice, but the squares and flowery GF are unsightly.WaltC - Tuesday, July 20, 2021 - link
The Intel rumor crashed and burned, as predicted...;) People need to learn how to separate the wheat from the chaff insofar as rumor is concerned. If something sounds silly and nonsensical it probably, very likely, is.Sahrin - Tuesday, July 20, 2021 - link
$30B is less than the cost of GF's fabs themsleves. Intel's offer is a ridiculous lowball and Mubadala would be idiotic to accept it. They'll easily get twice that in the IPO - Mubadala's target should be $100B, their lowest acceptable offer ~$70B.Oxford Guy - Wednesday, July 21, 2021 - link
‘a private-public partnership’Haha.
mode_13h - Thursday, July 22, 2021 - link
How exactly is it not? I guess the only thing that doesn't wash is that the State should get some rights to the IP, were it a real partner. That's usually what happens in private partnerships.Oxford Guy - Monday, July 26, 2021 - link
The ‘bottom 90%’ of the public is completely invisible to Congress according to the famous Harvard/Northwestern study.Public/private partnership is a euphemism, one gleefully deployed in pseudo-capitalism.
mode_13h - Tuesday, July 27, 2021 - link
> The ‘bottom 90%’ of the public is completely invisible to CongressNon-sequitur. And certainly not true of the pandemic-related benefits passed in both last year and this.
Also, if you look at the tax code, the lower half of the US population does quite well. Even the middle-income folks get benefits like child tax-credits and mortgage interest deductions.
> Public/private partnership is a euphemism
As long as the public gets more benefits than it would otherwise accrue, it's not. It's in the long-term strategic interest of the US public to maintain a domestic semiconductor manufacturing capability. That's the public interest. Another is to benefit other US industries, such as automotive manufacturing.
As I said, I'd like for the public to get a greater return on its investment, either in financial or other terms, but there's indeed a legitimate public interest.
It's almost like you're challenging yourself to see how cynical you can be. My sympathies for whatever misfortune has embittered you so. That's a very sad and one-dimensional way of looking at the world. There's not such a deficit of sadness in the world that we need to search it out so fastidiously.
Oxford Guy - Wednesday, July 28, 2021 - link
‘Non-sequitur.’Spoken like a member of Congress.
mode_13h - Saturday, July 31, 2021 - link
I don't think so. I am not strictly partisan. I just try to get past the reflexive cynicism that seems to be in vogue, and actually engage my brain.You call yourself "Oxford", but I sense a degree of anti-intellectualism that reminds me of the sportsball players I knew in high-school. It's like you've forgotten how to look at issues on their merits and challenge extreme positions by searching for counter-examples. That's not a refusal to take a position on things, but I try to take *measured* positions and not get too calcified in my views.
Oxford Guy - Wednesday, July 21, 2021 - link
Just out of curiosity... is GF still making chips with the 32nm SOI used for Bulldozer/Piledriver?ashima02 - Thursday, July 29, 2021 - link
Good <a href="https://airaeuroautomation.webgarden.com/">...mode_13h - Saturday, July 31, 2021 - link
spammer