TSMC founder: Pat Gelsinger doesn't have enough time to take Intel to the top

[midian182]

What just happened? Pat Gelsinger has shaken things up throughout his brief time as CEO of Intel, placing a heavy focus on the company’s manufacturing operations and opening up its capacity to other chipmakers. He even proclaimed in October that “Intel is back.” But TSMC’s founder and former CEO thinks Gelsinger won’t turn Chipzilla into a global manufacturing leader for one simple reason: time is against him.

Morris Chang, the man who founded TSMC and served as its Chairman and CEO, made the comments at his lecture 'Cherish Taiwan's Advantages in Semiconductor Wafer Manufacturing,' reports UDN (via Tom’s Hardware). Chang wasn't being ageist. He noted that Intel, like many US companies, has a rule that its executives must retire at a certain age—65, in team blue’s case. Gelsinger turns 61 in March next year.

Update (12/14): An Intel spokesperson reached out to inform us that the company has updated the corporate officers' retirement age "for some time now." As such, corporate officers are no longer required to retire at age 65. The document (PDF) has been updated as recently as November 10, 2021, and reads that under the new policy, "non-employee directors may not stand for reelection after age 75."

Just over four years is no short span of time, but Intel's plans for dominance are long-term. However, the two Arizona chip plants that it broke ground on in September are expected to become fully operational in 2024. The plants are part of Intel's renewed IDM 2.0 strategy that involves its newly formed Intel Foundry Services (IFS) division manufacturing chips for others. Intel is talking to over 100 companies for foundry work, and two of the first high-profile customers will be Amazon and Qualcomm.

The factories are set to produce chips using Intel's 20A process, the first to use its "RibbonFET" version of Gate-All-Around (GAA) transistors and PowerVia interconnects.

Gelsinger himself has said that Intel’s IDM 2.0 initiative will help it compete against TSMC, Samsung, GlobalFoundries, UMC, and others by 2025, which would still be within his tenure as boss. And depending on his level of success, the Intel board could decide to retain its current CEO's services after he turns 65.

Some of what Chang says could be a response to Gelsinger publicly warning against the dangers of relying on Taiwanese chipmakers given China’s threatening activities in the region. "Taiwan is not a stable place," said Gelsinger, speaking at Fortune Brainstorm Tech, reports Nikkei. "Beijing sent 27 warplanes to Taiwan's air defense identification zone this week. Does that make you feel more comfortable or less?"

Gelsinger also said the US government should subsidize domestic companies exclusively when spending the $52 billion it has set aside for semiconductor funding. The comment didn’t go down well with current TSMC chairman Mark Liu. "It will be very negative for the United States to subsidize only American companies," Liu said. "Unlike Intel, TSMC is very positive about non-U.S. chipmakers expanding capacity in America. It is a great thing."

Permalink to story.

https://www.techspot.com/news/92558-tsmc-founder-thinks-pat-gelsinger-age-prevent-intel.html

 

[Nanochip]

Does the Taiwanese government subsidize American or other foreign chip companies too?

 

[bviktor]

"Intel, like many US companies, has a rule that its executives must retire at a certain age—65"

That's pretty shocking to me.

 

[seeprime]

Intel *currently* has a rule that requires retirement at age 65. With better health care, I suspect Intel will raise that to 75 or whatever age the board feels is best in order to let Gelsinger keep running the show, as long as Intel keeps on improving at a fast pace under his watch. Rules can be easily changed, especially if it is likely to improve stock prices.

 

[fps4ever]

TSMC/Taiwan can't promise future stability because of China. Why would/should Intel help fund that at all. Problem is Taiwan is so important to the world for tech manufacturing it could potentially lead to a military showdown if the spice stopped flowing due to a Chinese take over...

 

[Nobina]

If he's that good they should make an exception for him, no reason to gamble on someone else just because of Pats age.

 

[Aceseven]

"Intel, like many US companies, has a rule that its executives must retire at a certain age—65"

That's pretty shocking to me.

Wish the govt would make an age limit...

 

[hahahanoobs]

lol
How many thought AMD would lose a 3 year lead in 4 years? I know I felt good about them dropping off from what I seen from Zen thru Zen 2. AMD wasn't capitalizing as fast as positive Zen articles made you think they were.

On the server side, 0-10% is far more easier to obtain than 10-20%, 20-30% etc. The next few generations will show us who is really who.

TSMC could be right, but only time will tell.

 

[Sausagemeat]

lol
How many thought AMD would lose a 3 year lead in 4 years? I know I felt good about them dropping off from what I seen from Zen thru Zen 2. AMD wasn't capitalizing as fast as positive Zen articles made you think they were.

On the server side, 0-10% is far more easier to obtain than 10-20%, 20-30% etc. The next few generations will show us who is really who.

TSMC could be right, but only time will tell.

I thought they would lose the lead. They only had a lead because Intels 10nm was delayed so much, if that had been on time then Ryzen wouldn’t have been the massive success it was. Now they are relying on TSMC, if they struggle to get capacity at TSMC it could go downhill very quickly for AMD. But then again Intels new processes could suffer delays too. I don’t agree with Mr Chang, 4 years is a long time in this world, Gelsinger could do a lot.

 

[HardReset]

I thought they would lose the lead. They only had a lead because Intels 10nm was delayed so much, if that had been on time then Ryzen wouldn’t have been the massive success it was. Now they are relying on TSMC, if they struggle to get capacity at TSMC it could go downhill very quickly for AMD. But then again Intels new processes could suffer delays too. I don’t agree with Mr Chang, 4 years is a long time in this world, Gelsinger could do a lot.

Yeah right.

- Creating new CPU architecture from scratch and getting it into market takes at least 5 years.

- Developing new process node and building factory ready for manufacturing takes at least 4 years total.

There's basically nothing Gelsinger can do in 4 years. He can do but results are next CEO's problem or benefit.

 

[Dimitrios]

INTEL only has a few options left.

1.) Bring Steve Jobs back to life but you will need a shovel and gonna need to answer 3 riddles.

2.} Pour sugar in Lisa's gas tank every morning. ;-)

3.} Start mining.

 

[Irata]

Does the Taiwanese government subsidize American or other foreign chip companies too?

Do they build fabs in Taiwan ? If so, they might.

Btw: Intel (an American company) is also actively looking for subsidies to build a new fab in Europe, so that would be a case of foreign governments subsidizing an American company. That does seem ‚slightly‘ hypocritical in light of their statements.

 

[rmcrys]

Btw: Intel (an American company) is also actively looking for subsidies to build a new fab in Europe, so that would be a case of foreign governments subsidizing an American company.

The point is, not always there are reciprocity: if the EU says that all companies in EU soil get funds, then they cannot make exceptions; the USA may say otherwise. Then it is unfair, but that is EU's fault.

France, Holland, Germany etc still have a lot of interests (exportation and building cheaper parts) on Asia, so the EU rules are according what these few countries need; they clean China's a.s and they allow their products.

Intel will start building a chip factory in Germany as well in the USA to give supply and decrease Asia's dependency. The USA defends west/ country, sadly the EU doesn't make the same and is VERY Russia/ China dependent (even big car brands are producing most of the new electric cars parts in China and importing the parts...).
Also a big part of the European companies have Chinese or Russian investors, alone it wouldn't be bad if they were politically agnostic, but they aren't...

 

[yeeeeman]

That is a power move from the TSMC CEO haha
"You say TSMC is based in an unstable location? Then you're too old man to make a difference for intel" =))))

 

[StudioCardo]

Passive aggressive trash talking, TSMC style.

TSMC has been the standard for as long as I can remember, but she only became the cutting edge leader in the past few years primarily due to the fear of losing Apple to Samsung. So TSMC rose up against the challenge - they deserve the respect. Whether Intel can respond to this existential crisis, only time can tell.

It's essential to be able to fab sub-5nm chips in the US. And Taiwan has every reason to be concerned if TSMC becomes one of many choices. Very concerned.

 

[ACE76]

Wish the govt would make an age limit...

I thought they would lose the lead. They only had a lead because Intels 10nm was delayed so much, if that had been on time then Ryzen wouldn’t have been the massive success it was. Now they are relying on TSMC, if they struggle to get capacity at TSMC it could go downhill very quickly for AMD. But then again Intels new processes could suffer delays too. I don’t agree with Mr Chang, 4 years is a long time in this world, Gelsinger could do a lot.

Intel didn't just lose due to 10nm...Zen is superior to Intel's architecture and scales way better. Zen 4 will again hammer Intel back. Intel will need a new architecture and that ain't happening in 4 years.

 

[ACE76]

Passive aggressive trash talking, TSMC style.

TSMC has been the standard for as long as I can remember, but she only became the cutting edge leader in the past few years primarily due to the fear of losing Apple to Samsung. So TSMC rose up against the challenge - they deserve the respect. Whether Intel can respond to this existential crisis, only time can tell.

It's essential to be able to fab sub-5nm chips in the US. And Taiwan has every reason to be concerned if TSMC becomes one of many choices. Very concerned.

Fab process is only going to become harder... getting beyond 3nm is going to be extremely hard regardless of the fab. TSMC just leapfrogged the Industry with their awesome 7nm implementation and that's carrying them to 5nm...it'll be very interesting to see where the industry winds up at 3nm. Samsung is also rapidly catching up and their 5nm process is looking very promising. Their 8nm dies were underwhelming at best. I have to say I was very surprised to see TSMC beat Intel to 7nm. Intel's 10nm is nowhere near as good as they would like people to believe and their 7nm is not even close to being viable yet. With the amount of resources Intel has in their fabs, I thought they would be comfortably at 5nm by now...but that only further shows how insanely hard it has become to get these dies smaller and smaller. I have to wonder when silicon hits EOL status because it can't be that far away now.

 

[Ludak021]

Intel didn't just lose due to 10nm...Zen is superior to Intel's architecture and scales way better. Zen 4 will again hammer Intel back. Intel will need a new architecture and that ain't happening in 4 years.

Is Alder Lake a joke to you? If so, you aren't being serious. What do you think 13xxx will look like? I t will be refined ADL and ryzen will be behind (HEDT and server is not reachable by Intel at this point), it already is, but we are waiting for 3D cache. Zen 4 is already late.
Aside from that, Intel has much larger profit margins due to having own fabs. And, again, aside from that, CPU market is a fraction of Intel's profits, wheres for AMD on the other hand....
Competition is good.

 

[HardReset]

Is Alder Lake a joke to you? If so, you aren't being serious. What do you think 13xxx will look like? I t will be refined ADL and ryzen will be behind (HEDT and server is not reachable by Intel at this point), it already is, but we are waiting for 3D cache. Zen 4 is already late.
Aside from that, Intel has much larger profit margins due to having own fabs. And, again, aside from that, CPU market is a fraction of Intel's profits, wheres for AMD on the other hand....
Competition is good.

AMD is not behind. Alder Lake is too hot, it's pretty good for low thread games but for everything else AMD has better options. Zen4 is not late yet, it's supposed to come next year and that will make it on schedule.

I highly doubt Intel's margins are larger because of own fabs. Just looking how much 10nm fiasco has cost...

 

[Ludak021]

AMD is not behind. Alder Lake is too hot, it's pretty good for low thread games but for everything else AMD has better options. Zen4 is not late yet, it's supposed to come next year and that will make it on schedule.

I highly doubt Intel's margins are larger because of own fabs. Just looking how much 10nm fiasco has cost...

I am yet to find "too hot" outside of overcloking (a separate topic) and general clueless forum postings. In standard use, especially in gaming, it's not hotter than any other CPU from intel or AMD.
Now you probably go and look at 12900K running something "torture" that isn't used by anyone in any use. Like Cinebench, or Linx, maybe Prime95. Those are synthetic benchmarks. It is hot, but not too hot. Too hot is when PC shuts down. On the other hand, you can get 12700K and it's not even hot.

So please, move on from silly conclusions based on...I don't know, I've been reading reviews and I never saw what you apparently saw.

I highly doubt Intel's margins are larger because of own fabs. Just looking how much 10nm fiasco has cost...

..yea but they have been following Ryzen half a step behind all the time, using 14nm. Not paying to reserve slots or anything like that + re-using same tech for 5 years.
I could be wrong, but I think I made an educated guess.

 

[HardReset]

I am yet to find "too hot" outside of overcloking (a separate topic) and general clueless forum postings. In standard use, especially in gaming, it's not hotter than any other CPU from intel or AMD.
Now you probably go and look at 12900K running something "torture" that isn't used by anyone in any use. Like Cinebench, or Linx, maybe Prime95. Those are synthetic benchmarks. It is hot, but not too hot. Too hot is when PC shuts down. On the other hand, you can get 12700K and it's not even hot.

So please, move on from silly conclusions based on...I don't know, I've been reading reviews and I never saw what you apparently saw.

You do realize that with Alder Lake we are talking about CPU's that have only 8 "normal" cores and 8 gimped ones. Where is Alder Lake CPU with 16 or even 12 "performance" cores? Nowhere to be seen. Now make educated guess why that's the situation.

..yea but they have been following Ryzen half a step behind all the time, using 14nm. Not paying to reserve slots or anything like that + re-using same tech for 5 years.
I could be wrong, but I think I made an educated guess.

AMD gave up fabs because they considered them too expensive. That was 2009 when manufacturing costs on new processes were much lower. GlobalFoundries ditched 7nm production because it was too expensive.

Intel is building new fabs that cost about same amount than Intel's revenue on 2020. That's just building cost, not running costs.

Another way to put it: if AMD gets around 30K wafers per month from TSMC and wafer cost is around 12K, that makes around 4.3B per year. Constuction of Intel's latest plant (two factories) cost around 20B. Basically AMD could buy TSMC 5 years worth wafers (current rate) for cost Intel just builds factory. I know that Intel needs more capacity than AMD but still factories are ultra expensive.

 

[Ludak021]

I do realize that you switched from "too hot", to not enough cores, and gimped cores (each of those performs similar to cores in first ryzen while using a lot less power)
I also realize that this Intel's new arch will continue in the future and it's only going to get better and better. Like Ryzen did (first ryzen had more cores than any desktop cpu in its class ever, but overall, out of productivity, it wasn't very fast), Zen 3 is light years ahead of Zen 1 and 1+ So is 10900K (Skylake) compared to 6700K (Skylake).
It's naive to think that the same wont follow with ADL. AMD is on a good track with CPUs again, so, again, competition is good (for consumers).

edit: AMD gave up fabs because they were broke.

 

[HardReset]

I do realize that you switched from "too hot", to not enough cores, and gimped cores (each of those performs similar to cores in first ryzen while using a lot less power)

130 watts Over Ryzen 5950X consumption is way too hot (Techspot review). And that's exact reason why there are so few "big" cores (10900K had 10). You see, too hot and low amount of cores goes hand to hand. Too hot = less cores.

Perform similar vs Ryzen on certain situations, and with much bigger cache.

I also realize that this Intel's new arch will continue in the future and it's only going to get better and better. Like Ryzen did (first ryzen had more cores than any desktop cpu in its class ever, but overall, out of productivity, it wasn't very fast), Zen 3 is light years ahead of Zen 1 and 1+ So is 10900K (Skylake) compared to 6700K (Skylake).

This is not that much new arch, Alder Lake's Golden Cove cores are tweaked Sunny Lake cores that are tweaked Skylake cores that are tweaked Sandy Bridge cores from 2011.

As we already see, Alder Lake have huge problems with power consumption and developing it more means even higher power consumption. Unless Intel makes something radical.

Creating fast CPU is not problem. Creating fast CPU with low enough power consumption is. Alder Lake's Golden Cove is already too hot and that limits how much it can be improved.

 

[Ludak021]

Again , it's 130W more in something no one except reviewers will ever use. If you run games it's not hot at all, it's on par with AMD.
Clock that 5950X 100MHz more and se how much W it pulls. Clock it at 5.2GHz and it wont even work. Apples and oranges. My friend picked up 5800x that runs Cinebench 23 @~95C under 360AIO 10 minutes into the render...So, that can happen too.

Lets not speculate, let's just wait and see. I for one am not early adopter, and have no interest in purchasing ADL, or anything for that matter since I don't see any improvements over my 9700K that would warrant spending cash anyway.

Have a nice weekend!

 

[quadibloc]

Intel didn't just lose due to 10nm

AMD came back with the first generation of Ryzen long before 10nm was an issue.
But the first and second generations of Ryzen, although credible as alternatives, were still definitely behind Intel, even if not nearly as far behind as in the sad Bulldozer days before Ryzen.
It was the 10nm issue that allowed the third generation of Ryzen to be close enough to Intel for the differences to be insignificant, and for the fourth generation to actually get ahead of Intel.
And despite being in the wilderness for a while, Intel still has a lot more resources than AMD. Without a process disadvantage, Intel should be able to get back into the fray quickly.
I see no reason why Intel could not manage to achieve a clear lead over AMD with their next generation a year from now.
But while Intel's lead would be clearly visible, AMD would still remain a viable option unless one needed the absolute maximum of single-thread performance attainable - and AMD, of course, would, I expect, make great efforts to keep up.