News Intel 3Q21 Results

[Hitman928]

Intel EPS beats by $0.60, misses on nnon-gaap revenue

Intel (INTC): Q3 Non-GAAP EPS of $1.71 beats by $0.60; GAAP EPS of $1.67 beats by $0.61.Non-GAAP Revenue of $18.1B (-1.1% Y/Y) misses by $170M.Shares -3.6%.Press Release Achieved...

seekingalpha.com

• EPS beat by $0.60
• Revenue very slight miss by $170M
• Guidance: Raising full-year 2021 EPS and gross margin guidance. Now expecting GAAP EPS of $4.50 and non-GAAP EPS of $5.28 from prior guidance of $4.80 vs. $4.11 consensus. and GAAP gross margin of 55% and non-GAAP gross margin of 57%.
• Q4 Guidance: Revenue of $18.3B vs. $18.26B consensus, EPS of $0.90 vs. $0.94 consensus.
• Q4 gross margin guidance weak at 51.4% compared to 56% Q3.

Edit:
Revenues

	3Q21​	Relative 2Q21​	Relative 3Q20
CCG​	$9.7B	96%	98%
DCG​	$6.5B	100%	110%
IOTG​	$1.0B	101.6%	154%
Mobileye​	$326M	99.7%	139%​

 

[Hitman928]

Intel stock getting hit pretty hard in after hours, down 7% right now.

 

[Gideon]

Ian's take is interesting:

https://twitter.com/x/status/1451284670432301057

Some highlight from these:

• There is a huge cut to tax-rate 0.5% (!), down 14.7%
• 'performance per watt parity in 2024 and leadership in 2025' (So not sure what Pat was saying about AMD being done in 2022/23?)
• Gross margin in Q4 expected to be 51.4%, which is surely Intel's lowest quarterly GM in decades!
• Intel has wanted to be 'Data centric' with 50/50 between client and server for years. Q3 2021 is 53.4% client, 35.9% server, 10.6% other. usually server and other are combined, so 53.4 to 46.6
• Intel's numbers are all YoY, very few QoQ. At the last disclosure, they had both, now it's very one sided on YoY. I wonder what's hiding in those numbers

 

[jpiniero]

Hmm, Notebook volume went down -26% QoQ and 14% YoY but ASPs skyrocketed. Chromebook demand finally easing?

Also desktop volume was up quite a bit, 16% YoY.

 

[moinmoin]

'performance per watt parity in 2024 and leadership in 2025' (So not sure what Pat was saying about AMD being done in 2022/23?)

First AMD then Apple, now that Intel can't delude themselves anymore that Apple is a mere customer that's needs to be won over? Well...

 

[HurleyBird]

• 'performance per watt parity in 2024 and leadership in 2025' (So not sure what Pat was saying about AMD being done in 2022/23?)

Are they referring to AMD here, or Apple maybe?

In either case, I greatly dislike these statements about far-away moving targets. So, if you execute the way you expect (and the way things tend to work, the way you expect is pretty close to the best case scenario. Error bars are almost entirely in the direction of you executing worse), and your competitor doesn't do perform better than your projections, which is a big if, it's going to take you a full four years to surpass them?

Anything over a year-and-a-half feels more aspirational than predictive.

 

[Saylick]

Some pretty bold claims by Pat that Intel is "on track, if not ahead" of their roadmap targets. 4 nodes in 4 years? Pssshhh, give me a break.

 

[Doug S]

Anything over a year-and-a-half feels more aspirational than predictive.

Since when has Intel ever told something other than a rosy story for ANYTHING happening further than 18 months out?

I guess they figure the Wall Street analysts covering them also cover 30 other companies and have forgotten all the sunshine they blew up everyone's behind about 10nm with every call for years even as they had delay after delay? They even did a rigged release of ~100K units for a single SKU in 2018 so they could claim they met a target Wall Street was holding them to about "shipping 10nm for revenue".

They can aspire to deliver 20A in 2024 and 18A in 2025 all they want. Their record over the past 8 years should make everyone skeptical. TSMC isn't perfect either, as their clockwork release of a new process every two years timed for Apple's needs finally hit a speedbump with N3. Samsung overextended themselves trying to do EUV on 7nm to beat TSMC and it is looking like the same story all over again with GAA on 3nm. Intel has to do both EUV and GAA successfully to deliver 18A. If they ship it (a real 18A, i.e. something clearly superior in density to TSMC N2 which I'm guessing will be > 500Mtr / mm^2) in any real quantity before 2027 I'll be shocked.

 

[Joe NYC]

Another gem from the call is that Sapphire is not only not in production yet, it will not be in production in Q4, it will start production in Q1 2022.

Ramping in Q2 according to Pat. Based on this schedule, Intel will be nearly defenseless in the server market for nearly another year.

And when it finally starts to ship in volume, 56 core Sapphire Rapids will be competing against 96 core Genoa on TSMC N5 - with nearly double the cores. And a process node advantage.

 

[moinmoin]

Since when has Intel ever told something other than a rosy story for ANYTHING happening further than 18 months out?

Well, they got Pat now, it's different now.

What's actually different and by how much, this we still have to find out. At least Intel found a new way to insert "leadership" everywhere again.

 

[DrMrLordX]

Some pretty bold claims by Pat that Intel is "on track, if not ahead" of their roadmap targets. 4 nodes in 4 years? Pssshhh, give me a break.

If you count the nodes they're going to use from TSMC, then maybe. They're already taking N6, N5, and (eventually) N3. Wafer volume will be their biggest holdup until maybe Intel 20A.

 

[jpiniero]

https://twitter.com/x/status/1451309516537139206

According to Ian, on the earnings call Intel admitted that nobody is buying XCC Icelake Server.

 

[dullard]

Some pretty bold claims by Pat that Intel is "on track, if not ahead" of their roadmap targets. 4 nodes in 4 years? Pssshhh, give me a break.

1) They are just about to release Intel 7 products with Alder Lake.
2) Intel 4 is actually coming along. Loihi 2 is functional if not perfect https://www.anandtech.com/show/16960/intel-loihi-2-intel-4nm-4 , and Meteor Lake's Intel 4 CPU tile has powered on: https://images.anandtech.com/doci/17026/Q364.png . Meteor Lake is scheduled to launch 1H 2023.

That there is 2 nodes in just 1.5 years from now.

3) I haven't heard much about Intel 3 yet. Rumors are that it is just an optimization of Intel 4 using more EUV masks. If true, then that is not usually a difficult target to hit as long as Intel gets a few more EUV machines by then. https://wccftech.com/intel-process-node-roadmap-new-naming-scheme-intel-7-4-3-20a-beyond-2024/ They plan to release this small node bump in 2 years.

So there are 3 nodes in 2 years--which is probably doable.

4) Intel 20A is where I doubt that they can be so optimistic. It relies on Intel successfully completing multiple innovations, any of which could slip.

5) Intel 18A might be the easier one. Intel is at the front of the line for the high NA EUV equipment. If that ships on time, then Intel could very well have a much easier time than the other foundries.

 

[jpiniero]

2) Intel 4 is actually coming along. Loihi 2 is functional if not perfect https://www.anandtech.com/show/16960/intel-loihi-2-intel-4nm-4 , and Meteor Lake's Intel 4 CPU tile has powered on: https://images.anandtech.com/doci/17026/Q364.png . Meteor Lake is scheduled to launch 1H 2023.

It's one thing to get the product working, it's another to get the yield high enough to ship in any kind of volume. Even then it's questionable whether Intel has enough EUV equipment for anything real in 2023.

 

[dullard]

It's one thing to get the product working, it's another to get the yield high enough to ship in any kind of volume. Even then it's questionable whether Intel has enough EUV equipment for anything real in 2023.

I think that is why they separated Intel 4 from Intel 3: they couldn't do all the EUV that they wanted with Intel 4. But also, remember that they are not necessarily producing the SOC components or the GPU components on Intel 4 for Meteor Lake. That means the actual chip area that is being produced on Intel 4 is much smaller than any of their recent CPUs. This both increases yield and requires less EUV equipment.

 

[Asterox]

Well, they got Pat now, it's different now.

What's actually different and by how much, this we still have to find out. At least Intel found a new way to insert "leadership" everywhere again.

Yes, but the Pope still plays a crucial role.

https://twitter.com/x/status/1410614835151196169

 

[DrMrLordX]

as long as Intel gets a few more EUV machines by then.

Allegedly that won't be happening, at least not in great enough numbers.

 

[jpiniero]

This both increases yield and requires less EUV equipment.

The chiplet is very likely to be very small but the wpm is unlikely to be much either.

 

[NTMBK]

1) They are just about to release Intel 7 products with Alder Lake.
2) Intel 4 is actually coming along. Loihi 2 is functional if not perfect https://www.anandtech.com/show/16960/intel-loihi-2-intel-4nm-4 , and Meteor Lake's Intel 4 CPU tile has powered on: https://images.anandtech.com/doci/17026/Q364.png . Meteor Lake is scheduled to launch 1H 2023.

That there is 2 nodes in just 1.5 years from now.

3) I haven't heard much about Intel 3 yet. Rumors are that it is just an optimization of Intel 4 using more EUV masks. If true, then that is not usually a difficult target to hit as long as Intel gets a few more EUV machines by then. https://wccftech.com/intel-process-node-roadmap-new-naming-scheme-intel-7-4-3-20a-beyond-2024/ They plan to release this small node bump in 2 years.

So there are 3 nodes in 2 years--which is probably doable.

4) Intel 20A is where I doubt that they can be so optimistic. It relies on Intel successfully completing multiple innovations, any of which could slip.

5) Intel 18A might be the easier one. Intel is at the front of the line for the high NA EUV equipment. If that ships on time, then Intel could very well have a much easier time than the other foundries.

Look at the gap between Cannonlake and an actually shipping 10nm product. "Powered on some test chips" is pretty meaningless.

 

[Joe NYC]

https://twitter.com/x/status/1451309516537139206

According to Ian, on the earnings call Intel admitted that nobody is buying XCC Icelake Server.

High core count segment is shifting to AMD.

 

[Doug S]

2) Intel 4 is actually coming along. Loihi 2 is functional if not perfect https://www.anandtech.com/show/16960/intel-loihi-2-intel-4nm-4 , and Meteor Lake's Intel 4 CPU tile has powered on: https://images.anandtech.com/doci/17026/Q364.png . Meteor Lake is scheduled to launch 1H 2023.

Claiming functional chips in a future process is meaningless. I'm sure Apple has some functional N3 silicon back from TSMC already. TSMC probably has some working N2 silicon, if only something like an SRAM. Apple won't have a completed A* SoC yet since they won't have taped that out (and obviously neither has Meteor Lake or they wouldn't have talked about a "tile" powering up) but they surely give TSMC some stuff to run during process development before risk production begins for process characterization.

Consider my previous post about Intel and the sunshine they blow up everyone's ass for everything further than 18 months out. Intel can tell us they have "scheduled" something to launch 18+ months from now. They used to get the benefit of the doubt from their long history of success, but that's no more. The safe bet is on their schedule slipping.

For how many years was Intel telling us that mass production of 10nm was starting up in the near future - usually about 18 months out?

They had 10nm functional to the point they were able to produce simple stuff like they've now reported for Intel 4 back in 2015, and said everything was moving according to schedule for entering mass production. Then they couldn't get the kinks out to allow proper mass production for FIVE YEARS. Why should we believe this time they'll go from the early reports of success to true mass production in only 18 months like they did back in the day? Lots of things work in the development fab in small runs but fall flat when you try to scale that to tens of thousands of wafers.

The CEO isn't in the fab solving problems and getting the kinks out of mass production. I'm not sure why people expect a new CEO will fix Intel's problems, unless they were always non technical in nature and he could solve them by shuffling managers around on the production team.

 

[BorisTheBlade82]

Well, I give Pat the benefit that he has far more knowledge about the technical aspects than many of his predecessors. I mean, he was basically already a living legend back in the nineties and early 2000s. This definitely helps and he already shuffled some heads around.
But two things are really concerning to me: He brought a lot of old people back. So I am not sure if this is the right way to approach the future. And is public claims were very bold in the last weeks - a bit too bold for my liking with respect to the developments of the last couple of years.

 

[dullard]

Claiming functional chips in a future process is meaningless. I'm sure Apple has some functional N3 silicon back from TSMC already. TSMC probably has some working N2 silicon, if only something like an SRAM. Apple won't have a completed A* SoC yet since they won't have taped that out (and obviously neither has Meteor Lake or they wouldn't have talked about a "tile" powering up) but they surely give TSMC some stuff to run during process development before risk production begins for process characterization.

Consider my previous post about Intel and the sunshine they blow up everyone's ass for everything further than 18 months out. Intel can tell us they have "scheduled" something to launch 18+ months from now. They used to get the benefit of the doubt from their long history of success, but that's no more. The safe bet is on their schedule slipping.

For how many years was Intel telling us that mass production of 10nm was starting up in the near future - usually about 18 months out?

They had 10nm functional to the point they were able to produce simple stuff like they've now reported for Intel 4 back in 2015, and said everything was moving according to schedule for entering mass production. Then they couldn't get the kinks out to allow proper mass production for FIVE YEARS. Why should we believe this time they'll go from the early reports of success to true mass production in only 18 months like they did back in the day? Lots of things work in the development fab in small runs but fall flat when you try to scale that to tens of thousands of wafers.

The CEO isn't in the fab solving problems and getting the kinks out of mass production. I'm not sure why people expect a new CEO will fix Intel's problems, unless they were always non technical in nature and he could solve them by shuffling managers around on the production team.

Intel's main problem was they tried to do something that their technology could not do: too much area reduction in one step (while branching out into all kinds of different areas pulling their people in all directions). Intel failed for years at that. It was a bad management decision to try for too much and at the same time to let the best engineers go.

But, there are reasons to suspect that is over.
1) Intel isn't trying to push that much change any more in one node.
2) The design is transferrable between fabs and not Intel specific, meaning TSMC can produce Meteor Lake for Intel if needed to buy time for Intel to get more EUV up and running.
3) The chip sizes are far, far smaller now going with tiles instead of a giant monolith.
4) EUV means fewer patterns needed, which is easier to produce.
5) At least two chips have powered on and are functioning (not perfectly, but they have something to learn from and plenty of time). More info on Meteor Lake: https://www.tomshardware.com/news/intel-meteor-lake-powered-on
6) Intel has been hiring the better people back.
7) Intel has massive government spending to help.

All you have is speculation that the past will repeat itself despite the changes. Yes, Intel can still fail. I underlined where I think that is likely in my post above. But, the next three nodes are not likely to repeat the problems of Intel's recent past.

 

[Dayman1225]

Another gem from the call is that Sapphire is not only not in production yet, it will not be in production in Q4, it will start production in Q1 2022.

Ramping in Q2 according to Pat. Based on this schedule, Intel will be nearly defenseless in the server market for nearly another year.

And when it finally starts to ship in volume, 56 core Sapphire Rapids will be competing against 96 core Genoa on TSMC N5 - with nearly double the cores. And a process node advantage.

They literally said this months ago in June… not news.

Intel Delays Sapphire Rapids Xeon CPU Production To Q1 2022 | CRN

Intel announced that it’s delaying production the next generation of Xeon Scalable CPUs, code-named Sapphire Rapids, to the first quarter of 2022.

www.crn.com

 

[DrMrLordX]

They literally said this months ago in June… not news.

That actually surprises me a little since I had assumed Intel would at least have Sapphire Rapids ready for hyperscalars in Q1. Gelsinger's comments seem to cement the idea that they won't even have moved beyond ES by that point.