• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Speculation: Intel will become fabless

Page 14 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

With the loss of its manufacturing lead, will Intel become fabless?

  • Yes, Intel is a product designer at heart, and they will seek a more flexible fabless model.

    Votes: 26 14.2%

  • No, manufacturing is integral to Intel, and they will continue to invest to stay competitive.

    Votes: 157 85.8%
  • Total voters
    183
Spartak said:
I think they might be a perfect match.
Click to expand...

That's an interesting take. On the one hand, their IP and base of engineering talent has 0 experience on any nodes smaller than their 12nm node (actually 12nm+, plus the FDX nodes fwiw). So I don't see them as being helpful in directly fixing any of Intel's node problems. But they did a). absorb IBM's old fabs + staff plus IBM's 32HP node and b). license/adapt 14nm from Samsung so they do have experience on how to adapt to survive. Maybe that experience combined with a licensed node from one of the active foundries (Samsung or TSMC) would be enough to keep Intel afloat for awhile yet.

I still don't see Intel surviving as an IDM of any kind without bending a knee to Samsung or TSMC.
 
DrMrLordX said:
That's an interesting take. On the one hand, their IP and base of engineering talent has 0 experience on any nodes smaller than their 12nm node (actually 12nm+, plus the FDX nodes fwiw). So I don't see them as being helpful in directly fixing any of Intel's node problems. But they did a). absorb IBM's old fabs + staff plus IBM's 32HP node and b). license/adapt 14nm from Samsung so they do have experience on how to adapt to survive. Maybe that experience combined with a licensed node from one of the active foundries (Samsung or TSMC) would be enough to keep Intel afloat for awhile yet.

I still don't see Intel surviving as an IDM of any kind without bending a knee to Samsung or TSMC.
Click to expand...
Seems like TSMC only sees their involvement with Intel as transitional. That leaves Samsung, which is having problems like Intel right now. I thought they might be a good match at first, but now I'm thinking the cultural differences (business culture and actual) will be a huge hurdle for a partnership. Samsung isn't doing well enough to just sell it's processes to Intel. Something more clever needs to be done. Intel has the volume, as best I can tell based on 14nm's high utilization, to remain an IDM from a financial standpoint.
 
DrMrLordX said:
That's an interesting take. On the one hand, their IP and base of engineering talent has 0 experience on any nodes smaller than their 12nm node (actually 12nm+, plus the FDX nodes fwiw). So I don't see them as being helpful in directly fixing any of Intel's node problems. But they did a). absorb IBM's old fabs + staff plus IBM's 32HP node and b). license/adapt 14nm from Samsung so they do have experience on how to adapt to survive. Maybe that experience combined with a licensed node from one of the active foundries (Samsung or TSMC) would be enough to keep Intel afloat for awhile yet.

I still don't see Intel surviving as an IDM of any kind without bending a knee to Samsung or TSMC.
Click to expand...
People seem to be forgetting that GF got a working 7nm 'gift' from IBM, and their process engineers. They chose however to not spend money building a new fab and/or getting the equipment to produce them.
 
scannall said:
People seem to be forgetting that GF got a working 7nm 'gift' from IBM, and their process engineers. They chose however to not spend money building a new fab and/or getting the equipment to produce them.
Click to expand...

A process that "works" for IBM mainframe might not be economically viable for mass market CPUs.
 
NTMBK said:
A process that "works" for IBM mainframe might not be economically viable for mass market CPUs.
Click to expand...

Economic viability of a process (once the fixed cost is sunk to build the fab and develop the process) is basically down to yield, and if you yield well enough for the giant chips IBM uses in their mainframes, you will yield just fine for the smaller chips most customers want.
 
I don't believe the IBM 7nm process would be ready to go after they install a bunch of EUV machines. Also: it would be similar to the TSMC/Samsung 7nm node, not the Intel 7nm that's ~50% denser.
 
Doug S said:
Economic viability of a process (once the fixed cost is sunk to build the fab and develop the process) is basically down to yield, and if you yield well enough for the giant chips IBM uses in their mainframes, you will yield just fine for the smaller chips most customers want.
Click to expand...

With the prices IBM charges for mainframes, I imagine that you could probably eat yields as bad as 1-2 viable chips per wafer.
 
NTMBK said:
With the prices IBM charges for mainframes, I imagine that you could probably eat yields as bad as 1-2 viable chips per wafer.
Click to expand...
The Power9 CPU's aren't as expensive as you might think. The 22 core model is roughly $3400. 22 cores and 88 threads isn't bad. 2.75 Ghz base, 3.8 Ghz boost. It's a 190 watt part as well.
 
scannall said:
People seem to be forgetting that GF got a working 7nm 'gift' from IBM, and their process engineers. They chose however to not spend money building a new fab and/or getting the equipment to produce them.
Click to expand...

Actually I wasn't even aware of that. Wasn't IBM forced to move POWER fabrication entirely away from GF due to their failure to advance in node tech?
 
This is a wake up call from the Street to (the GM of) Intel fabs. Do you want to keep your job? Show me some results in 7nm and they'd better be great and on time, or else ...

As of today, Intel is turning into a one Gig player after divesting in a few biz. This could be good news if the money generated goes to fund long term projects for the next computing wave after Intel lost the mobile computing opportunity.

The competitive advantage of having your own fab is huge, TSMC gross margin is still >50%, that means the 45 B$ profit would become 22.5 B$ if you don't count the cost saving once you don't run the fabs... not to mention that manufacturers may not be too gentle with pricing once the competition is gone.

I am very curious to see what happens the day TSMC gets a virtual monopoly on 5nm and 3 nm.
 
sfpietri said:
I am very curious to see what happens the day TSMC gets a virtual monopoly on 5nm and 3 nm.
Click to expand...
This is looks like more and more the case. Apparently Qualcomm is also shifting it's 5nm Samsung efforts to TSMC
https://twitter.com/x/status/1290939127878135809

I really hope Samsung's Nanosheet 3nm lives up to all the hype, but considering how they blew EUV after claiming of being faster than anyone else ... not that likely 🙁
 
Wow, losing a big customer like Qualcomm is really going to hurt Samsung. Their 5nm process must have been missing early development targets for QC to jump ship.
 
sfpietri said:
This is a wake up call from the Street to (the GM of) Intel fabs. Do you want to keep your job? Show me some results in 7nm and they'd better be great and on time, or else ...
Click to expand...

Maybe its not even possible to do that with the current situation. Then you'll just end up with another BK character because realistic people won't get hired. So they'll get some that just says everything is fine and waits for their stock to vest before being fired/retiring.

Recall the CEO search recently ended with them hiring one of their own bean counters. I wouldn't be surprised if most qualified candidates looked at Intel's situation, looked at the board that got it into this mess still existing, thought to themselves "there's 5-10 years of work here, and these goons will fire me when its not done in 3 because they're getting desperate" and then declined.
 
Gideon said:
This is looks like more and more the case. Apparently Qualcomm is also shifting it's 5nm Samsung efforts to TSMC
https://twitter.com/x/status/1290939127878135809

I really hope Samsung's Nanosheet 3nm lives up to all the hype, but considering how they blew EUV after claiming of being faster than anyone else ... not that likely 🙁
Click to expand...


Well, I guess we will find out when Qualcomm reveals Snapdragon 875G. China and Taiwanese media tends to overexaggerate Samsung's problem so I wouldn't take it as face value.
 
Taiwanese media are too biased. Samsung just made an statement that the rumours about their 5nm delay are false. Also it doesn't make any sense to me this rumour of switching the SD875 production in the middle of 2021 from SS to TSMC. By that time almost all the production is done and they are totally different processes to make a port easily.
 
Does Qualcomm ever dual source their SoCs? If for no other reason than Samsung probably wants the SoCs they put in their non-Exynos phones made in their own fabs.
 
Doug S said:
Does Qualcomm ever dual source their SoCs? If for no other reason than Samsung probably wants the SoCs they put in their non-Exynos phones made in their own fabs.
Click to expand...
Nope, they have never done that. This "rumour" makes no sense.
 
Looks like the huge TSMC 6nm order could partially be explained by the outsourcing of Xe-HPG dies for gaming GPUs :

videocardz.com

Intel to unveil Xe-HPG gaming architecture with hardware ray-tracing - VideoCardz.com

Intel has confirmed it is developing another Xe sub-architecture. Intel Xe-HPG for enthusiast gamers coming in 2021 Intel has been actively developing another GPU architecture since 2018. This is another Xe micro-architecture called Xe-HPG optimized for gaming. The XPG roadmap spans from...
videocardz.com videocardz.com

We will very likely hear more about this in the next couple of weeks (Hot chips).
 
Intel’s manufacturing crisis puts company at a crossroads

"Intel shows no sign of changing course, telling employees this month that owning its own fabs gives the company “the ability to control our own destiny.”"

"Keyvan Esfarjani, the newly appointed head of Intel’s manufacturing group, wrote to his organization Aug. 7 to affirm that integrated design and manufacturing “continues to be a crucial differentiator and advantage for Intel.” “We get a competitive advantage from the strong connection between products and process design and will continue to invest in our future process technology roadmap,” Esfarjani wrote. “Also, operating our fabs gives us the ability to control our own destiny in terms of capacity and flexibility for supply.” “Contingency plans” to outsource some manufacturing is a matter of giving Intel flexibility, Esfarjani wrote. He said Intel’s 10nm production is running 20% ahead of what it expected in January and he insisted the outlook is strong."

oregonlive.com

It seems their "contingency plan" is of the small mitigating kind, while the main plan is to get back to business-as-usual.

Will Intel stomach spending much on their contingency plans? The cost of taping out chips on leading-edge processes is becoming astronomical, so having chips ready at TSMC if internal manufacturing fails seems expensive. Can they do contingency planning on the cheap without affecting beloved margins? They are already taking a painful cut to their gross margins as they ramp up production for Tiger Lake and upcoming products on their expensive low-yielding 10nm process.

And how can they get a supply agreement of sufficient magnitude and longevity, if it is just a contingency?
 
Vattila said:
the main plan is to get back to business-as-usual.
Click to expand...
I have to wonder what they would consider as business-as-usual to get back to? Technically the business-as-usual they should strive for is the one from a decade ago, before their node cadence started to stutter more and more. And as we discussed already a lot of staff from back then is gone as well, either retired, let go or jumped ship. A contingency plan is fine if they are working on a feasible master plan. But such is indeed hard to see coming so far.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top