News Intel 3Q25 Earnings

[DrMrLordX]

I know the average trader is a monkey with lackluster impulse control and delayed reasoning; I still can't explain the rally.

If you're going tobe a monkey, it's better to be a wealthy one!

All silliness aside, it certainly looks like a bad report for Intel (overall) since future forecasts show lower output, lower margins, and lower revenue. Cost-cutting while also selling off entire business units is only good if you're able to expand elsewhere on higher-margin product lines.

 

[Joe NYC]

Hey I think I figured out who you are! Charlie from Semiaccurate, am I right? 🤣

Charlie's brain was fried during Covid.

I know some people personally who came out of Covid completely Zombified.

 

[adroc_thurston]

All silliness aside, it certainly looks like a bad report for Intel (overall)

Nah it's fine actually.
Good enough quarter and non-total disaster guidance.
The stinky part is all wrt capacity constraints.

Charlie's brain was fried during Covid.

no he's very much alright.

 

[DrMrLordX]

The stinky part is all wrt capacity constraints.

The question is, how long is that going to be a problem for them? Capacity constraints = limited growth potential/investment potential. They're contracting out of necessity.

 

[adroc_thurston]

The question is, how long is that going to be a problem for them?

Until PTL ramps pmuch.
So mid-26.

 

[Joe NYC]

Well that's just dumb.
Intel is capacity constrained because their product mix is mostly legacy 10nm stuff.

That's where they say they are most capacity constrained. They did not say Intel 3 or Intel 4. They said it is 10nm / Intel 7.

They also said that Intel is re-allocating some of the Intel 7 wafers from client (Raptor Lake) to server (Saphire Rapids / Emerald Rapids).

Which then creates even a bigger hole in low end client.

Both of these development point to opportunity for AMD to sell out their inventories and deliver upside to quarterly earnings. So, stock market reaction was for Intel to go down from After Hours irrational exuberance, and AMD stock to rise.

All the new parts Intel has negative incentive to sell since margins there are Pretty Bad.

BTW, there were some interesting comments on by Zinsner on that, especially costs - paraphrasing:

In client, we have good competitiveness but our costs suck
In server, we suck a both competitiveness and costs.

Eh, only in theory.
We'll see on AMD Q4 guide.

After guiding $8.7B for Q3, I am pretty sure Q4 guide will be above $9B. My question is if it will be closer to $9B or $10B.

 

[Joe NYC]

If you're going tobe a monkey, it's better to be a wealthy one!

All silliness aside, it certainly looks like a bad report for Intel (overall) since future forecasts show lower output, lower margins, and lower revenue. Cost-cutting while also selling off entire business units is only good if you're able to expand elsewhere on higher-margin product lines.

Part of the Q4 forecast will be loss of ~$400m revenue from Altera. If you subtract it from Q3 earnings, then Q4 guide = Q3 actual. In other words flat.

But flat is not good for Q4. Q4 is seasonally the strongest quarter. The most puzzling part of the report was Intel guiding client "moderately" down for Q4, which, especially for client, is the strongest quarter.

So, something is going on with the capacity...

 

[adroc_thurston]

They did not say Intel 3 or Intel 4

Because i3/i4 contain products no one cares about.

Both of these development point to opportunity for AMD to sell out their inventories and deliver upside to quarterly earnings.

AMD does not have fat inventories of client stuff.

In client, we have good competitiveness but our costs suck
In server, we suck a both competitiveness and costs.

yeah water's wet.

After guiding $8.7B for Q3, I am pretty sure Q4 guide will be above $9B. My question is if it will be closer to $9B or $10B.

Any upside will be pmuch GPGPU-driven.

 

[Joe NYC]

The question is, how long is that going to be a problem for them? Capacity constraints = limited growth potential/investment potential. They're contracting out of necessity.

I have a feeling that TSMC is not going to bend over backwards to find capacity for Intel for 3 quarters during highest demand, given Intel's stated goal to ditch TSMC wherever possible.

Makes more sense for TSMC to find that missing capacity and offer it to AMD.

 

[adroc_thurston]

I have a feeling that TSMC is not going to bend over backwards to find capacity for Intel for 3 quarters during highest demand, given Intel's stated goal to ditch TSMC wherever possible.

again, Intel *does not want* to ship their TSM-fabbed parts.
They're margin drains.

 

[Joe NYC]

Because i3/i4 contain products no one cares about.

These would be Meteor Lake, Sierra Forest and Granite Rapids.

If Intel is re-allocating capacity from client (Raptor Lake) to server (Saphire Rapids, Emerald Rapids), it does not look like demand for Granite Rapids is that hot... Which is interesting, given Granite Rapids is more competitive... Maybe just slow ramp and not enough design wins.

AMD does not have fat inventories of client stuff.

We will see. Lisa said in either Q1 or Q2 call that as far as inventories, AMD is building / preparing "for success"

We will find out soon what exactly that meant.

Any upside will be pmuch GPGPU-driven.

Seems like everything will be up, except gaming, which will probably be dead in the water. But really big gains can only come from GPGPU.

 

[Khato]

For those interested in the actual quote:

We're in reasonably decent shape on clients in terms of product performance and competitiveness. You know, with a few exceptions, but we're not where we need to be on a cost basis. And so we've got to make improvements there, and we have that on the roadmap team recognizes it, but that's a multiyear process to get there. But it's more pronounced on the data center side. Not only do we not have the right cost structure, but we also don't have the right competitiveness to really get the right margins from our customers.

Which can easily be paraphrased as client is okay, but we can't hit our target >50% gross margin buying silicon from TSMC. Meanwhile server is at least keeping all the fabs busy, but being second tier performance on parts intended to be priced as top-tier similarly means the gross margins are bad. (Though they have been notably improving from the very bad state of a year ago.)

 

[Joe NYC]

again, Intel *does not want* to ship their TSM-fabbed parts.
They're margin drains.

Intel's overarching goal (up until the current CEO LBT) was to not lose market share. Even at cost of margins. So, if something has to give, it would be the margin.

We don't know what happened, maybe Intel could not obtain more capacity from TSMC to hold the market share.

 

[branch_suggestion]

again, Intel *does not want* to ship their TSM-fabbed parts.
They're margin drains.

People gotta understand that NVL/DMR will be the low point for Intel, that is the truly painful period where margin and market share are doomed to regress unless there is a super high demand market.

 

[Khato]

One peculiar point to me remains Intel blaming the capacity-constrained Intel 7 fabs for lowering margin. I guess that's indicative of just how many layers they had to use multi-patterning on? Since normal legacy nodes are the money makers. I guess this could be an artifact of the prior Intel model of designing process nodes for leading edge usage only. Regardless, still struck me as odd given that those fabs are destined for conversion to Intel 12 once Intel 7 is no longer needed, correct? Or were any of them built with the ability to accommodate EUV scanners? Wonder if Intel 12 is going to be adequate for them to go back to manufacturing their own desktop chipsets or if that'll continue to be outsourced to Samsung.

 

[Tarkin77]

That's where they say they are most capacity constrained. They did not say Intel 3 or Intel 4. They said it is 10nm / Intel 7.

They also said that Intel is re-allocating some of the Intel 7 wafers from client (Raptor Lake) to server (Saphire Rapids / Emerald Rapids).

Which then creates even a bigger hole in low end client.

Both of these development point to opportunity for AMD to sell out their inventories and deliver upside to quarterly earnings. So, stock market reaction was for Intel to go down from After Hours irrational exuberance, and AMD stock to rise.

Very true. The situation reminds me of 2018/19 when Intel struggled with insufficient 14nm capacity, and AMD gained significant market share.

 

[coercitiv]

IMO, it is a reaction to Intel's earnings report, where Intel announced that the demand is so strong that Intel is capacity constrained. And Intel will continue to be capacity constrained well into Q1 and Q2 of 2026. In other words, full 12 months of shortages for Intel that started somewhat in Q2 2025, but became more material in Q3 2025/

This makes the recent rumor of price hikes for 7nm products in Asia seem genuine.

If Intel is re-allocating capacity from client (Raptor Lake) to server (Saphire Rapids, Emerald Rapids), it does not look like demand for Granite Rapids is that hot... Which is interesting, given Granite Rapids is more competitive... Maybe just slow ramp and not enough design wins.

We also need to keep in mind the last few years of 7nm client manufacturing were partially lost by making chips which deteriorate and require replacements from current production output. In the context of these shortages, every defective Raptor Lake is a double loss for Intel.

Which can easily be paraphrased as client is okay, but we can't hit our target >50% gross margin buying silicon from TSMC.

Intel is also having trouble with gross margins because they're selling Arrow Lake at a discount. The 265K went from $400 in late 2024 to $260-300 in late 2025.The cheaper 245K started at $330 and gradually worked it's way down to $200-220 in late 2025.

 

[Khato]

Intel is also having trouble with gross margins because they're selling Arrow Lake at a discount. The 265K went from $400 in late 2024 to $260-300 in late 2025.The cheaper 245K started at $330 and gradually worked it's way down to $200-220 in late 2025.

No question that the retail market pricing on ARL has been creeping downward. I'd guess the same is true for OEM pricing, though likely to a lesser extent since their pricing didn't start as high.

Intel ends up with quite a few more of the lower SKUs than they'd prefer due to the poor frequency binsplit on the 285k. Will be interesting to see if the 270k maintains better pricing.

 

[jpiniero]

The problem with Arrow Lake in DIY is the crap gaming performance.

 

[adroc_thurston]

We also need to keep in mind the last few years of 7nm client manufacturing were partially lost by making chips which deteriorate and require replacements from current production output. In the context of these shortages, every defective Raptor Lake is a double loss for Intel.

Exploding RPLs make a tiny subset of their SKU stack.
The damage was in brand rep, not money.

 

[KompuKare]

Exploding RPLs make a tiny subset of their SKU stack.
The damage was in brand rep, not money.

On the other hand:
isn't it a clear illustration of penny-wise, pound-foolish?

Be too stingy to do right for RPL buyers, save a few pennies* but in the end loose big time in brand reputation.

*of course, pennies for a full recall of many RPL SKUs might have been tens of $billions. Brand reputational loss might be worth a lot more but hey: that's long-term while sorting RPL would be big short-term hit.

 

[adroc_thurston]

Be too stingy to do right for RPL buyers, save a few pennies* but in the end loose big time in brand reputation.

Oh no that's not what I mean.
x900k SKUs were a relatively tiny subset of RPL volume.
The damage was not in RMA counts, but brand rep.

 

[coercitiv]

Oh no that's not what I mean.
x900k SKUs were a relatively tiny subset of RPL volume.

Based on SKUs they covered with extended warranty, many were affected. Even if we consider this as "only" indirect proof, we already know all K SKUs have had cases of degradation.

Spoiler

Anyway, my point was this issue is compounding with the rest. Every little bit hurts.

 

[coercitiv]

Based on Intel's luck with acquisitions, they will buy a company like this just before the bubble bursts and they're left with a multi-billion hot potato on their hands:

https://archive.ph/20251030201806/https://www.bloomberg.com/news/articles/2025-10-30/intel-in-talks-to-acquire-ai-chip-startup-sambanova

Intel Corp. is in preliminary talks to buy artificial intelligence chip startup SambaNova Systems Inc., according to people familiar with the matter.

Any deal would likely value SambaNova at below the $5 billion it fetched in a 2021 funding round, the people said, asking not to be identified because the information is private.

For Intel, an acquisition of SambaNova would bring into its fold a company that Chief Executive Officer Lip-Bu Tan knows well.

Tan has served as SambaNova’s executive chairman since May 2024 and his venture capital firm Walden International was one of the company’s founding investors, having led a $56 million Series A round in 2018. SoftBank Group Corp. led a $676 million funding round in 2021 that valued the startup at $5 billion.

I'm not convinced the rumored acquisition is true, just that I fully expect Intel to try something similar even when there's overwhelming evidence that this "AI economy" will require a correction soon.

 

[DrMrLordX]

I'm not convinced the rumored acquisition is true, just that I fully expect Intel to try something similar even when there's overwhelming evidence that this "AI economy" will require a correction soon.

How many AI companies will Intel buy