Info Intel 4Q23 Earnings

[Hitman928]

Don't have much time right now but here are the highlights:

• Intel press release (NASDAQ:INTC): Q4 Non-GAAP EPS of $0.54 beats by $0.09.
• Revenue of $15.4B (+10.0% Y/Y) beats by $230M.
• Shares -5%.
• Forecasting first-quarter 2024 revenue of $12.2 billion to $13.2 billion vs $14.16B consensus; expecting first-quarter EPS attributable to Intel of $(0.25) (non-GAAP EPS attributable to Intel of $0.13 vs $0.32 consensus.

Net sales by category: Client Computing: $8.8B (33.0 Y/Y %); Data Center and AI: $4B (-10.0 Y/Y %); Network and Edge: $1.5B (-24.0 Y/Y %); Mobileye: $637M (13.0 Y/Y %); Intel Foundry Services: $291M (63.0 Y/Y %);

Pulled from:

Intel Non-GAAP EPS of $0.54 beats by $0.09, revenue of $15.4B beats by $230M (NASDAQ:INTC)

Intel beats expectations with Q4 earnings, reporting non-GAAP EPS of $0.54 and revenue of $15.4B. However, shares drop by 5%.

seekingalpha.com

Client segment up big Y/Y but data center and network down. Mobileye recovered from previous quarter and IFS up Y/Y but (I believe, someone can double check) down Q/Q.

 

[adroc_thurston]

Forecasting first-quarter 2024 revenue of $12.2 billion to $13.2 billion vs $14.16B consensus; expecting first-quarter EPS attributable to Intel of $(0.25) (non-GAAP EPS attributable to Intel of $0.13 vs $0.32 consensus.

ouch

 

[Hitman928]

ouch

Yeah, big drop in Q1 from Q4 and back to losing money. Q4 results are OK-ish in light of recent performance, but Q1 forecast is not good, especially when we are in the middle of a huge AI boom and Intel is trying to make itself out to be a significant AI player. These numbers tell otherwise.

 

[adroc_thurston]

Yeah, big drop in Q1 from Q4 and back to losing money. Q4 results are OK-ish in light of recent performance, but Q1 forecast is not good

CCG and brandon tax breaks are carrying them hard.
NEX is still eating glue.

especially when we are in the middle of a huge AI boom and Intel is trying to make itself out to be a significant AI player. These numbers tell otherwise.

Total Gaudi 2 units shipped are equivalent to like <1m of MI300X sales and probably <4 days of H100 sales.
They're not a player.

 

[Saylick]

Pretty much every tech company has to say they are leveraging AI or getting money from AI to uphold their stock price. The moment analysts realize the AI boom isn't materializing, the stock will take a hit. Intel is no exception.

We'll see how AMD fares in comparison when they have their earnings report. Lisa mentioned 2B in AI sales (MI300 if I'm not mistaken), which was likely a sandbagged number in the usual Lisa Su fashion. Question is how much are analysts pricing in...

 

[adroc_thurston]

Intel is not exception.

Semis stocks like NVDA or AMD or SMCI are getting AI pumps due to their share in training/inference farms.
That's real money, which Intel gets none of.

 

[Saylick]

Semis stocks like NVDA or AMD or SMCI are getting AI pumps due to their share in training/inference farms.
That's real money, which Intel gets none of.

Right... they are behind the 8 ball on the AI boom and it's going to be increasingly difficult to compete if they are already behind. AI is a blue ocean and first mover advantage carries so much weight; if you cannot be first mover, you must move faster than those who got there before you. Intel is not doing that at all.

 

[adroc_thurston]

if you cannot be first mover, you must move faster than those who got there before you. Intel is not doing that at all.

You just need strategy and good hardware and unfortunately Intel showcased neither so far.
Gotta remember that Intel had a headstart in ML market on everyone but Nvidia basically, since they bought Nervana in 2016.

 

[trivik12]

Horrendous Guide for sure. Its going to be horrible year for Intel again. DC spend has moved more to GPU end at this point and so NVidia cannot make their GPUs fast enough. Musk mentioned the same yesterday in Tesla call. If NVidia can supply sufficient GPU they will be good. Intel is still having just Ponte Vechio which is a 2021 product for Argonne that was delayed for ever. Gaudi will be niche as they are stopping after Gaudi 3. Falcon Shores is 2025 product which means it wont be on volume until 2026. There is no hope for them at this point.

Let us see how much of GPU share can AMD get once their earnings are out. MI300 is a competitive product but its all about H/W plus S/W.

 

[Saylick]

You just need strategy and good hardware and unfortunately Intel showcased neither so far.
Gotta remember that Intel had a headstart in ML market on everyone but Nvidia basically, since they bought Nervana in 2016.

I mean... Intel could've had a headstart w/ GPU computing over a decade ago if they played their cards right. Intel needs to re-focus as a company if they want to success. It feels like they've got too many markets they are trying desperately to keep hold of but are spread too thin and are too unfocused to make an impact on most of them.

 

[adroc_thurston]

I mean... Intel could've had a headstart w/ GPU computing over a decade ago if they played their cards right.

Well no, their GPU uArch sucks.
Nervana actually made interesting h/w but alas.

It feels like they've got too many markets they are trying desperately to keep hold of but are spread too thin and are too unfocused to make an impact on most of them.

The opposite, Pat is too focused on core Intel stuff like fabs to have a strategy for emergent markets.

DC spend has moved more to GPU end at this point and so NVidia cannot make their GPUs fast enough

they can, but they actually lost share at hyperscale to AMD.

MI300 is a competitive product but its all about H/W plus S/W.

H/W is the hard part really.
Current ML toolchains are built upon so many layers of abstraction that inserting additional h/w support is nowhere near as hard as it used to be.

 

[Saylick]

Well no, their GPU uArch sucks.
Nervana actually made interesting h/w but alas.

Again, they had a headstart.

The opposite, Pat is too focused on core Intel stuff like fabs to have a strategy for emergent markets.

Again, spread too thin.

 

[adroc_thurston]

Again, they had a headstart.

Half the industry had GPU IP once.
That never gave anyone any headstart.
It's only a headstart when your core hardware is good and Intel gfx is anything but.

Again, spread too thin.

The opposite, they focus too much on legacy stuff.
fabs-fabs-fabs, CCG-CCG, Xeon.
That's all they do.
Everything else is lip service or already dead (like their networking portfolio).

 

[Saylick]

Half the industry had GPU IP once.
That never gave anyone any headstart.
It's only a headstart when your core hardware is good and Intel gfx is anything but.

I wasn't talking about graphics. I was talking about GPU computing. They could have pivoted to being purely HPC with their GPUs but they tossed it all into the trash.

The opposite, they focus too much on legacy stuff.
fabs-fabs-fabs, CCG-CCG, Xeon.
That's all they do.
Everything else is lip service or already dead (like their networking portfolio).

Focusing on legacy stuff doesn't allow them to focus on new stuff, hence spread too thin. They aren't even successful in their core stuff, so how can you say they focus too much on it?

 

[adroc_thurston]

I was talking about GPU computing

That's only possible if your core GPU IP is good.
Intel's isn't.

They could have pivoted to being purely HPC with their GPUs but they tossed it all into the trash.

You can't.
Need to copy GCN first.

Focusing on legacy stuff doesn't allow them to focus on new stuff, hence spread too thin.

The opposite, they've been chopping product lines all the time since a while to focus on those core 3 precepts.

They aren't even successful in their core stuff

Yeah skill issue.

 

[Khato]

I'll certainly be a bit amused that in their conference call Intel claimed the same "well above $2 billion" 2024 AI accelerator pipeline as AMD previously did.

 

[Joe NYC]

Horrendous Guide for sure. Its going to be horrible year for Intel again. DC spend has moved more to GPU end at this point and so NVidia cannot make their GPUs fast enough. Musk mentioned the same yesterday in Tesla call. If NVidia can supply sufficient GPU they will be good. Intel is still having just Ponte Vechio which is a 2021 product for Argonne that was delayed for ever. Gaudi will be niche as they are stopping after Gaudi 3. Falcon Shores is 2025 product which means it wont be on volume until 2026. There is no hope for them at this point.

Let us see how much of GPU share can AMD get once their earnings are out. MI300 is a competitive product but its all about H/W plus S/W.

I think only Mi300a from El Capitan will be counted in Q4 for AMD, which should be ~ $400m in sales. Which is still meaningful.

Mi300x only kicks-in in Q1 2024. AMD estimated $2 billion in sales for 2024. Patrick Moorhead just predicted $10 billion for 2024. So that's a pretty big range.

 

[adroc_thurston]

I think only Mi300a from El Capitan will be counted in Q4 for AMD, which should be ~ $400m in sales. Which is still meaningful.

Mi300x only kicks-in in Q1 2024

Let's just say it's not quite right.

Patrick Moorhead just predicted $10 billion for 2024. So that's a pretty big range.

No that's accurate, I too expect $8-10B of MI300 sales.
They're genuinely taking GPGPU share from NV which is:
a) wild
b) imagine that 5 years ago

 

[Ajay]

CCG and brandon tax breaks are carrying them hard.

What's a brandon tax break?

They're genuinely taking GPGPU share from NV which is:

Taking? I thought that AMD was benefitting from lack of supply on NV's part. Or, are you talking about HPC and ML?

 

[adroc_thurston]

What's a brandon tax break?

Biden given Intel tax breaks.

Taking? I thought that AMD was benefitting from lack of supply on NV's part.

Yea that was the idea until hyperscalers (quite some really) switched orders to MI300X and they definitely don't have H100 in a short supply.
You can only imagine how hard everyone at NV is fuming right now.

 

[Tigerick]

IFS up Y/Y but (I believe, someone can double check) down Q/Q.

Yes, Q to Q's revenue of IFS has dropped:- And here is full year revenue from IFS, less than $1 billions with almost $500 millions in operating loss...

IFS's 2023 Total Revenue	Q1	Q2	Q3	Q4	Total
Revenue	$118 m	$232 m	$311 m	$291 m	$952 m
Operating Income	- $140 m	- $143 m	- $86 m	- $113 m	- $482 m

 

[coercitiv]

What's a brandon tax break?

Brandon is special and specific hot sauce. Combine this with some ER missiles for good measure and you'll get an idea of how polarized the hero scientist really is. When this forum wakes up from under his spell, it will do so with a nasty hangover.

 

[gdansk]

Brandon is special and specific hot sauce. Combine this with some ER missiles for good measure and you'll get an idea of how polarized the hero scientist really is. When this forum wakes up from under his spell, it will do so with a nasty hangover.

His tone is very foreign here. We're not gonna have a hangover because someone disappears after making big claims for months without proof.

 

[adroc_thurston]

His tone is very foreign here. We're not gonna have a hangover because someone disappears after making big claims for months without proof.

No need to get upset.
Get some street cred and you'll have all the proofs you need.

 

[Vattila]

Intel's resilience in the client segment is impressive, but overall it looks stark for Intel, with fierce competition from all directions, and no obvious winning formula.

While they may see some AI windfall in their product divisions, Intel's prospects in the long-term relies on Intel Foundry Services (IFS), I surmise. Apparently, establishing Intel as a leading foundry is CEO Pat Gelsinger's main focus.

Dr. Ann Kelleher, EVP and GM of Technology Development, looks to have delivered brilliantly on Gelsinger's "5-in-4" process roadmap mantra:

1. Rebranded and polished the underperforming bread-and-butter "10nm" FinFET process, now known as Intel_7.

2. Rescued and delivered their delayed "7nm" EUV FinFET process, with pre-release version Intel_4 used for the lead-vehicle "Meteor Lake" notebook CPU and the full-release version Intel_3 ready for the "Sierra Forest" dense server CPU and external IFS customers this year.

3. Developed and looks poised to deliver their next-generation process based on backside power delivery ("PowerVia") and gate-all-around transistors ("RibbonFET"), targeting process leadership, with pre-release version Intel_20A used for the lead-vehicle "Arrow Lake" client CPU by the end of this year and the full-release version Intel_18A ready for the "Clearwater Forest" dense server CPU and external IFS customers next year.

Of these achievements, the Intel_18A process with PowerVia and RibbonFET is the really crucial milestone for Pat Gelsinger's turnaround plan, as this process is planned to regain industry leadership. I look forward to seeing how Intel_18A fares in the market. As far as I understand, interest in Intel_3 is muted. So IFS take-off now looks to all ride on Intel_18A. Hopefully for Intel, and for the global semiconductor supply chain, they will ramp meaningful production volumes for external customers in 2026 and beyond.

The IFS Direct Connect 2024 event next month will be interesting.