Intel CEO admits 'I've bet the whole company on 18A'

Zarathustra[H]

Zarathustra[H]

Extremely [H]
Joined
Oct 29, 2000
Messages
38,878
"I’ve bet the whole company on 18A." So, says Intel CEO Pat Gelsinger in the most deadpan, matter-of-fact manner imaginable. He's not glib. He's not joking. Everything rides on Intel's 18A process according to Gelsinger.


View: https://www.youtube.com/watch?v=0PrmrMQ9gJU

As it stands, the U.S. cant afford to lose Intel's fab capacity from a national security perspective, so if it does fail, I expect bail outs.

And if there are going to be bail-outs if you lose, why not bet the whole company?
 
Probably a good bet anyway. Even if that specific process doesn't pan out, they need to get their ass in gear if they want to stay competitive in the fab space. This'll get them one step closer to the front, if nothing else.
 
18A will make or break Intel.
Intel's chips as they stand now are figuratively hot garbage. Not because the architecture is bad or their designs are shit but because Intel is almost 3 generations behind on their fabs.
Intel's best chips right now have their most complicated parts made by TSMC, and Intel is simply handling the packaging.
Intel can not continue to operate the way they are, and TSMC is not large enough to handle Intel's entire lineup.
Until Intel can reach some degree of parity with TSMCs process it will bleed customers and fall further behind.
 
Last edited:
Lakados said:
18A will make or break Intel.
Click to expand...
I really doubt it. Intel certainly needs to move towards better manufacturing but even Intel's current Meteor Lake lineup looks pretty compelling, and that's on their antiquated 7nm. If Intel wants to get back lost market share from AMD and Apple then they absolutely need 18A.
Intel can not continue to operate the way they are, and TSMC is not large enough to handle Intel's entire lineup.
Click to expand...
If TSMC isn't large enough to handle Intel then Intel must be selling a lot of chips. Also, wasn't Intel going after Samsung and IBM for manufacturing as well? Probably for their GPU's but Intel isn't afraid to buy up large amounts of manufacturing from competitors.
 
If you read their latest financials it says clearly it'll take 2-3 years to get back to parity with the competition, but all is on track it seems
 
DukenukemX said:
I really doubt it. Intel certainly needs to move towards better manufacturing but even Intel's current Meteor Lake lineup looks pretty compelling, and that's on their antiquated 7nm. If Intel wants to get back lost market share from AMD and Apple then they absolutely need 18A.

If TSMC isn't large enough to handle Intel then Intel must be selling a lot of chips. Also, wasn't Intel going after Samsung and IBM for manufacturing as well? Probably for their GPU's but Intel isn't afraid to buy up large amounts of manufacturing from competitors.
Click to expand...

Meteor Lake is one of the TSMC chips, Intel 4 does the computing, TSMC N5 does the GPU (which makes up the bulk of its performance), TSMC N6 does the SoC and IO processes.

Intel still sells an absolute crapload of CPUs, and TSMC can't currently make enough of them fast enough to feed AMD's demand, but TSMC also has 3 new plants coming online this year, 4 if some Miracle gets Arizona back on track that is a lot of new supply coming to market, that would only see AMD's presence grow and Intel's shrink more.

Intel's Foveros packaging lets them pick and choose, mix and match, which is unique to them and is really awesome, but buying all your components from other suppliers while your own manufacturing arm strangles you won't work for long, and IFS won't grow fast enough to cover that loss. Not to mention it would be terrible marketing, Come on everybody use IFS, I mean we can't but you should!
 
Lakados said:
Meteor Lake is one of the TSMC chips, Intel 4 does the computing, TSMC N5 does the GPU (which makes up the bulk of its performance), TSMC N6 does the SoC and IO processes.

Intel still sells an absolute crapload of CPUs, and TSMC can't currently make enough of them fast enough to feed AMD's demand, but TSMC also has 3 new plants coming online this year, 4 if some Miracle gets Arizona back on track that is a lot of new supply coming to market, that would only see AMD's presence grow and Intel's shrink more.

Intel's Foveros packaging lets them pick and choose, mix and match, which is unique to them and is really awesome, but buying all your components from other suppliers while your own manufacturing arm strangles you won't work for long, and IFS won't grow fast enough to cover that loss. Not to mention it would be terrible marketing, Come on everybody use IFS, I mean we can't but you should!
Click to expand...
Chip shortage will be gone this quarter. All that demand will be filled
 
Zarathustra[H] said:
"I’ve bet the whole company on 18A." So, says Intel CEO Pat Gelsinger in the most deadpan, matter-of-fact manner imaginable. He's not glib. He's not joking. Everything rides on Intel's 18A process according to Gelsinger.


View: https://www.youtube.com/watch?v=0PrmrMQ9gJU

As it stands, the U.S. cant afford to lose Intel's fab capacity from a national security perspective, so if it does fail, I expect bail outs.

And if there are going to be bail-outs if you lose, why not bet the whole company?
Click to expand...

Most of Intel's fabrication still happens in Israel, as far as I'm aware. I'm not aware of any major CPUs that are made in their US fabs, but that may have changed in the past few years.
 
Lakados said:
Meteor Lake is one of the TSMC chips, Intel 4 does the computing, TSMC N5 does the GPU (which makes up the bulk of its performance), TSMC N6 does the SoC and IO processes.
Click to expand...
That's fine and shows that Intel is willing to go beyond their own manufacturing.
Intel still sells an absolute crapload of CPUs, and TSMC can't currently make enough of them fast enough to feed AMD's demand, but TSMC also has 3 new plants coming online this year, 4 if some Miracle gets Arizona back on track that is a lot of new supply coming to market, that would only see AMD's presence grow and Intel's shrink more.
Click to expand...
I think that just means AMD will be even more competitive against Intel, which is a win for us consumers. Moore's Law Is Dead even said that AMD was bidding [H]ard for the Switch 2 over Nvidia. Right now AMD is in the heart of almost all Switch 2 clones, so AMD has a lot of business that Intel could jump in with their Meteor Lake chips. We're post crypto crash and mid AI boom, so as someone who just wants a faster computer will find it kinda sucks when all these tech giants can claim chip shortage and crank up prices because AI go brrrr. As it is right now I'm not particular to AMD or Intel when it comes to CPU, but with GPU I still lean towards AMD.
Intel's Foveros packaging lets them pick and choose, mix and match, which is unique to them and is really awesome, but buying all your components from other suppliers while your own manufacturing arm strangles you won't work for long, and IFS won't grow fast enough to cover that loss. Not to mention it would be terrible marketing, Come on everybody use IFS, I mean we can't but you should!
Click to expand...
The key thing here is that Intel is choosing to limit their CPU manufacturing to themselves, and something they could always change in a heart beat. It's just Intel being greedy and prideful that's pushing them to continue to use their antiquated manufacturing. Hopefully their 18A comes into action and we can start seeing faster and more power efficient chips from Intel.
 
DukenukemX said:
That's fine and shows that Intel is willing to go beyond their own manufacturing.

I think that just means AMD will be even more competitive against Intel, which is a win for us consumers. Moore's Law Is Dead even said that AMD was bidding [H]ard for the Switch 2 over Nvidia. Right now AMD is in the heart of almost all Switch 2 clones, so AMD has a lot of business that Intel could jump in with their Meteor Lake chips. We're post crypto crash and mid AI boom, so as someone who just wants a faster computer will find it kinda sucks when all these tech giants can claim chip shortage and crank up prices because AI go brrrr. As it is right now I'm not particular to AMD or Intel when it comes to CPU, but with GPU I still lean towards AMD.

The key thing here is that Intel is choosing to limit their CPU manufacturing to themselves, and something they could always change in a heart beat. It's just Intel being greedy and prideful that's pushing them to continue to use their antiquated manufacturing. Hopefully their 18A comes into action and we can start seeing faster and more power efficient chips from Intel.
Click to expand...

Consoles and mobile devices are usually large volume low margin stuff. Sales to consoles helped save AMD before Ryzen launched, but it is still not the market chipmakers like AMD and Intel really want to be in.
 
Zarathustra[H] said:
Consoles and mobile devices are usually large volume low margin stuff. Sales to consoles helped save AMD before Ryzen launched, but it is still not the market chipmakers like AMD and Intel really want to be in.
Click to expand...
Nvidia has now higher margin than both I think and I am sure AMD very much so want to be in the next PS6-Xbox.

You do not want to make less Epyc server chips, those less than 400mm chip going for over $5000 because of console APUs, but you take the hit sometime because you need to agree to volume with those clients and Sony can want near best tech by watt you can get.

Specially here, Switch will be on a cheap node (samsung 8m), very small chips with superbe yield, does not have to hurt much your other sales as virtually all your other product will not be made on it and once the console is old and still sell well you are probably getting good money for cheap by now to make old tech (and fully amortized R&D, everything a long time ago).
 
Last edited:
LukeTbk said:
Nvidia has now higher margin than both I think and I am sure AMD very much so want to be in the next PS6-Xbox.

You do not want to make less Epyc server chips, those less than 400mm chip going for over $5000, but you take the hit sometime because you need to agree to volume with those clients and Sony can want near best tech by watt you can get.

Specially here, Switch will be on a cheap node (samsung 8m), very small chips with superbe yield, does not have to hurt much your other sales as virtually all your other product will not be made on it and once the console is old and still sell well you are probably getting good money for cheap by now to make old tech (and fully amortized R&D, everything a long time ago).
Click to expand...

That is a good point. If you can pick up these large volume low margin contracts using architectures that have already been designed, so they're is low cost, and older process nodes so they dont pull fab capacity away from higher margin products, then they can be a huge win.

I didn't realize that the likes of Nintendo were OK with two gens old process nodes.
 
Zarathustra[H] said:
I didn't realize that the likes of Nintendo were OK with two gens old process nodes.
Click to expand...
Apparently (and maybe in a no covid world it would have happened) it could have been a Christmas 2022 product from the Nvidia SOC point of view.

The original 2017 Switch was on a Maxwell (2014) type gpu with a 2012 ARM Cortex platform CPU on TSMC 20nm, the first TSMC 7 product launched in 2018 I think to give some notion of how mature TSMC 14 was by then let alone 20, this new one could still end up more up to date and non Nintendo exclusive AAA able than the original switch, but they like making good money by console sold, need giant volume and at that price point it can mean old nodes and tech only.
 
Last edited:
Zarathustra[H] said:
I hope you are right. Dealing with overpriced chips due to the chip shortage has been getting old.
Click to expand...
All the top guns say so in their quarterly financials so probably, maybe?

Either way it looks like a good time to buy a bunch of Intel stock while the new fab ripens.
 
Zarathustra[H] said:
I hope you are right. Dealing with overpriced chips due to the chip shortage has been getting old.
Click to expand...
Sadly everyone is just rolling back production to cut supply and drive up costs on components to match. Now we’ll just have overpriced because margins.

Just about every component inside an electrolytic capacitor is expected to double in price this year, same with most electronic components as mining and refining output cuts back because of changes in carbon emission guidelines and blah blah blah energy credit brokerage fee gobbilty gook.

So to add woe to our misery we can also look forward to components that have a shorter MTBF. The expensive stuff will remain expensive because they are using the good parts while the others will stay about the same but use significantly cheaper quality components to keep the costs relatively flat. So I’m expecting 4-5 years down the road we get to relive the good old days when lead was being phased out and everything was exploding after 3-4 years of usage.
 
Last edited:
It's like I say it and then it comes true. Probably means that 18A isn't as nearby as we thought but it also means that Intel isn't as dependent on 18A's success.
https://www.tomshardware.com/news/intel-will-spend-14-billion-on-manufacturing-chips-at-tsmc-report

"Intel will allegedly place $4 billion worth of orders with TSMC in 2024 to fab 3nm CPU tiles, per a report from semiconductor analyst Andrew Lu (via eeNews). 2025 will also see a large number of Intel chips produced at TSMC, with orders totaling $10 billion. Lunar Lake is allegedly Intel's first processor to have its CPU cores fabbed at an external foundry, and Lu believes Intel will become increasingly dependent on TSMC in the future."

**EDIT**
Never mind the article is old.
 
DukenukemX said:
I really doubt it. Intel certainly needs to move towards better manufacturing but even Intel's current Meteor Lake lineup looks pretty compelling, and that's on their antiquated 7nm. If Intel wants to get back lost market share from AMD and Apple then they absolutely need 18A.

If TSMC isn't large enough to handle Intel then Intel must be selling a lot of chips. Also, wasn't Intel going after Samsung and IBM for manufacturing as well? Probably for their GPU's but Intel isn't afraid to buy up large amounts of manufacturing from competitors.
Click to expand...

Samsung perhaps, IBM sold its fabs to Global Foundries back when GloFo was still trying to keep up with cutting edge process nodes.
 
DukenukemX said:
It's like I say it and then it comes true. Probably means that 18A isn't as nearby as we thought but it also means that Intel isn't as dependent on 18A's success.
https://www.tomshardware.com/news/intel-will-spend-14-billion-on-manufacturing-chips-at-tsmc-report

"Intel will allegedly place $4 billion worth of orders with TSMC in 2024 to fab 3nm CPU tiles, per a report from semiconductor analyst Andrew Lu (via eeNews). 2025 will also see a large number of Intel chips produced at TSMC, with orders totaling $10 billion. Lunar Lake is allegedly Intel's first processor to have its CPU cores fabbed at an external foundry, and Lu believes Intel will become increasingly dependent on TSMC in the future."

**EDIT**
Never mind the article is old.
Click to expand...

Apparently slightly off mark. More recent articles I've read are that Intel will continue making it's GPU tiles on TSMC for those generations of CPU. As long as they're using the same architecture for discrete and IGPs, fabbing them all on the same process makes sense.
 
DanNeely said:
Apparently slightly off mark. More recent articles I've read are that Intel will continue making it's GPU tiles on TSMC for those generations of CPU. As long as they're using the same architecture for discrete and IGPs, fabbing them all on the same process makes sense.
Click to expand...
Intel has 3 generations of GPU’s planned at TSMC, I doubt they are going to spend the money to redesign them for 18A.

I really hope Intel follows up 18A with R it would just make sense.
 
I hope their 18A works as expected, and they get back into the HEDT processor market. I'm still on 10900x because I need the lanes.
 
GoodBoy said:
I hope their 18A works as expected, and they get back into the HEDT processor market. I'm still on 10900x because I need the lanes.
Click to expand...
If only to give AMD some competition, so they don't sit on their well-deserved laurels.
 
Drastic node leaps seems like it's destined for failure. Didn't they try that a few years ago? Seems like you would learn something important at each level, as electrons get closer together.
intel needs to work their employees 24/7 like TSMC.
 
sleepeeg3 said:
Drastic node leaps seems like it's destined for failure. Didn't they try that a few years ago? Seems like you would learn something important at each level, as electrons get closer together.
intel needs to work their employees 24/7 like TSMC.
Click to expand...

Intel is trying to avoid 10nm fiasco with going baby steps (Intel 7 > 4 > 3> 20A > 18A) instead of jumping to a new node in a single shot. 5 Nodes in 4 years is more like 2.5 nodes in 4 years in Intel's old nm terms.
 
what happened to the other Intel 3 like intel 20 that were planned before intel 18a ?

I think they are still in part still on:
https://www.techpowerup.com/319345/...wafer-pictured-first-silicon-built-on-intel-3

first commercial silicon built on the new Intel 3 foundry node, which is expected to be the company's final silicon fabrication node to implement FinFET technology; before the company switches to Nanosheets with the next-generation Intel 20A.

I could be missing the point here too
 
sleepeeg3 said:
Drastic node leaps seems like it's destined for failure. Didn't they try that a few years ago? Seems like you would learn something important at each level, as electrons get closer together.
intel needs to work their employees 24/7 like TSMC.
Click to expand...

No. Intel got into trouble trying to jam more than 2 years of upgrades into their every 2 year process update.

Doing smaller upgrades once a year or so is how TSMC got where it is.
 
Verge said:
I dunno, at this point intel should focus more on 18a and just make ARM chips. x86 needs to go away.
Click to expand...
But then Intel is competing with a lot of other ARM providers, and if they make a socketed ARM board then what’s to stop Qualcomm, Broadcom, etc from making chips for those boards. It’s the Intel, AMD, VIA, Cyrix, days all over again, nah Intel would be better off putting research into RISCV.
Imagine this if you will, Intel builds an abstraction layer that processes the incoming commands and breaks them down assigning them to the most appropriate core completely and independently of the OS. The abstraction layer can easily be x86 outwardly and be x86, GPU, ARM, RISC, NPU,… internally and that would just work. Intel is working on making it x86, GPU, and NPU for now, but nothing stopping them from expanding on that.
 
You must log in or register to reply here.
Back
Top