Intel's Manufacturing Outlook for the Future Doesn't Inspire Confidence in Successful Competition, According to Susquehanna Call

[bencrutz]

I'm still staggered Intel's stock hasn't fallen into the sub atomically level yet.

intel bought back their own stock in 2019, spent $7.6 billion on it
intel had allocated another $10 billion this year on buybacks, basically at the cost of their NAND business.

 

[thesmokingman]

intel bought back their own stock in 2019, spent $7.6 billion on it
intel had allocated another $10 billion this year on buybacks, basically at the cost of their NAND business.

Yeap, and they are bleeding cash these last few quarters.

 

[R0H1T]

intel bought back their own stock in 2019, spent $7.6 billion on it
intel had allocated another $10 billion this year on buybacks, basically at the cost of their NAND business.

Well how do you suppose you can pay yourself a big fat bonus & afford multiple platinum parachutes before you jump the (T)itanic

 

[Valantar]

I don't suppose Intel would tell us if they were having problems with industrial espionage.

It seems they should under stand the physics and the engineering required to make these denser products.

It would also add clarity to articles like this if we had a definition of what a 10nm node is.

We know it's not gate size. The "size" of an electron in silicon is about 5nm so it is unclear what TSMC or Intel mean on 7nm or smaller nodes.

Are you suggesting the low yields are due to actual sabotage? I mean, I guess it's possible, but I have a hard time imagining something like that would go by unnoticed. It's not like the engineering teams involved in this are small, and they no doubt check each others' work in various ways.

As for what a 10nm node is, that is essentially entirely up to the fab owner. Node names are marketing, and it's been a long, long time since the number reflected any specific feature size. In general, what one can say is that a 10nm node is denser and has smaller features than a >10nm node from the same fab, which typically also means lower power consumption and higher performance. There are no specifics beyond this that can be gleaned from the name alone, and node names across fabs are generally not comparable - for example Intel 10nm is supposedly comparable to TSMC 7nm and Samsung 7nm in physical feature sizes.

 

[renz496]

so going forward CPU market going to be suck? intel are having issues for years now and will not going to get better for another few years (even that is an IF). as a result will continue to fall behind AMD as time goes by although they probably still able to make some money because they got the capacity to sell to OEMs. AMD have the best CPU but capacity will continue to be an issue to them since there is dozens if not hundreds of other chip maker also fab their chip at TSMC. with TSMC keep things tight to ensure they reap the benefit for being the best fab in the world price will also going to reflect this. we supposed to get faster performance for more cheaper price but when i saw AMD price their 5600X at $300 i was uhh, to be honest i was expecting with Ryzen 5k we probably able to get new 8 core 16 thread CPU for under $300. not the existing model end up keeping it's price and the new model end up being more expensive because it was faster. when will brand new 8 core 16 thread CPU based on the newest architecture design will reach $150? i sure hope it happens before 2025.

well maybe if nvidia failed to get ARM for themselves intel will let themselves to be bought by nvidia so they can have JHH as their new CEO.

 

[Valantar]

so going forward CPU market going to be suck? intel are having issues for years now and will not going to get better for another few years (even that is an IF). as a result will continue to fall behind AMD as time goes by although they probably still able to make some money because they got the capacity to sell to OEMs. AMD have the best CPU but capacity will continue to be an issue to them since there is dozens if not hundreds of other chip maker also fab their chip at TSMC. with TSMC keep things tight to ensure they reap the benefit for being the best fab in the world price will also going to reflect this. we supposed to get faster performance for more cheaper price but when i saw AMD price their 5600X at $300 i was uhh, to be honest i was expecting with Ryzen 5k we probably able to get new 8 core 16 thread CPU for under $300. not the existing model end up keeping it's price and the new model end up being more expensive because it was faster. when will brand new 8 core 16 thread CPU based on the newest architecture design will reach $150? i sure hope it happens before 2025.

well maybe if nvidia failed to get ARM for themselves intel will let themselves to be bought by nvidia so they can have JHH as their new CEO.

We might be looking at something like 2011-2012 in reverse - one company with a process and arch advantage leading to far better efficiency, with the other company pushing clocks and throwing power efficiency out the window to compete. There will of course be differences; Intel is eminently competitive in IPC still, but on the other hand have to go with huge die sizes to keep up on 14nm. How that will evolve once Zen 4 arrives... who knows? I don't think we'll see another 14nm backport, so Rocket Lake might be peak IPC on 14nm. The question then becomes whether they'll keep pushing clocks higher or if they will have a 10nm/7nm alternative at that point.

I dont think we're going to see 8c16t below $150 in the next decade. Why would they do that? Nobody would have any reason to buy a CPU more expensive than that then, tanking average sales prices and causing major damage to the companies. Core counts have already increased enough to not be a limiting factor - in early 2017 high end MSDT was 4c8t, now its 16c32t. Going beyond that for consumers is meaningless. We might see 16c at ~$500 in a generation or two, and 8c at ~$300, but given the increasing costs of silicon production, that doesn't really seem likely.

 

[xenocide]

Yeap, and they are bleeding cash these last few quarters.

Not really? Their Earnings Per Share figure was still 2.5x AMD's, and their revenue YoY was pretty consistent with last year--last year was $19.2b and this year was $18.33b, but that's consistent with their struggles. Hardly a huge drop indicative of them "bleeding cash". Their balance sheet claims their CoH is actually up from a year ago--not quite as high as it was throughout the beginning of the year, but not in bad shape either--they reported $18.3b in Cash/Cash Equivalents in Quarter ending 9/30/20, compared to $12b in Quarter ending 9/30/19. But really, there's no indication they're in serious trouble from a financial perspective.

 

[R0H1T]

Yeah, you're not really paying attention then are you?

The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

www.anandtech.com

Their margins are going down because of the volume discounts they're having to dole out to make sure their server chips are accepted, now this only works for so long. Then there's the issue of 10nm still not yielding enough for Intel to transition en masse to 10nm, no indication that they ever can in fact! Let's not even get into the white whale that's 7nm, as it stands right now it's another 10nm repeat ~ except much more expensive

 

[Vayra86]

I'm still staggered Intel's stock hasn't fallen into the sub atomically level yet.



You do know that makes no sense, right ? Part of what makes these nodes and their multi-billion dollar investments profitable is capacity. Until you start moving hundreds of thousands or millions of units for several consecutive quarters, you are in the red.

Intel has fabs and therefore its own guaranteed production plus distribution channels. All competitors are waiting in line somewhere and high demand pressures their margins directly. So even if Intel hasnt the best chips they are still fast and more importantly not surrounded by risk and uncertainty. Its why their 14nm is so important. They can deliver in volume and a shareholder knows they will, because even shortages on smaller nodes will keep 14nm in business.

If you think of it, Intel could be having the most cost effective node right now on 14nm. Performance capped, but cheap and well refined evenfor bigger chips.