News Intel 2Q25 Earnings

[DrMrLordX]

Yeah. No margin for Tejas experimentation. At least not for the next 5 years

IN SOVIET RUSSIA, there is no margin for Tejas to experiment upon YOU!

da tovarisch

Wtaf is that.

Intel 17th Gen CPUs to Get Rentable Units: Why Hyper-Threading is Going Away | Hardware Times

Intel’s adoption of the hybrid core architecture has significantly changed the roadmap of the PC chipmaking industry. More and more applications are now taking advantage of the “secondary” low-power E-cores to boost performance and efficiency. This approach has its shortcomings, which Intel...

hardwaretimes.com

(you had to ask)

 

[Saylick]

I posted this not too long ago, but given the recent ER, I am holding steady to it:

At this point, I see Intel as descending for a few years with the trough being sometime in 2028, before slowing rebuilding back up in the first half of the 2030s. A lot of talent is going to be shed in the next few years in the interest of keeping the company afloat, and it takes an incredibly long time to rebuild talent. Then, you have to wait for the talent to come back before their products come to market, which itself takes a few years. So I have to ask myself: why would I, the consumer, gamble with Intel products for the next 10 years?

What I did not expect was LBT's comment about axing Intel 14A if they don't have any external customers, which goes completely against Intel's vertical integration approach since it implies 14A is DOA even if Intel were to somehow shift all its own products on it. It confirms that a ridiculous amount of volume is required to sustain a bleeding edge node these days that even Chipzilla itself couldn't produce enough volume alone to make it work.

 

[jpiniero]

I posted this not too long ago, but given the recent ER, I am holding steady to it:

What I did not expect was LBT's comment about axing Intel 14A if they don't have any external customers, which goes completely against Intel's vertical integration approach since it implies 14A is DOA even if Intel were to somehow shift all its own products on it. It confirms that a ridiculous amount of volume is required to sustain a bleeding edge node these days that even Chipzilla itself couldn't produce enough volume alone to make it work.

It's both volume and that Server doesn't generate the profits it used to.

 

[madtronik]

Well, I hope this is a transcription error, but I would at least hope that the CEO knows what SMT is Especially if he wants to have a final say according to the letter:

Surprising they don't refer to SMT by their own HT name.

Well, I guess he is just repeating what feature their customers tell him that makes them buy another guy's cpus because they don't have it.

 

[Jan Olšan]

Well, I hope this is a transcription error, but I would at least hope that the CEO knows what SMT is Especially if he wants to have a final say according to the letter:

Surprising they don't refer to SMT by their own HT name.

Likely just said the wrong term despite thinking the right concept. You are likely quite under pressure during these calls and presentations.

That 55% share is likely correct, or at least, correct wrt available data.

Gotta remember that the data we see from Mercury Research are just for AMD and ignoring ARM so AMD having like 28 % doesn't mean Intel has 72% of server. There's like 15 % for ARM or therebouts courtesy of all the cloud hyperscalers running Neoverse chips. So 55 % may be what Intel is currently at.

 

[511]

55% share includes ARM as well

 

[Panino Manino]

How many years was AMD's decline until recovery?
Will Intel spend more time in the gutter?

Intel discontinued Clear Linux.
No more x86 optimisations?
Would this imply that Intel is thinking about a future without x86?

 

[511]

How many years was AMD's decline until recovery?
Will Intel spend more time in the gutter?

Intel discontinued Clear Linux.
No more x86 optimisations?
Would this imply that Intel is thinking about a future without x86?

ahh discontinuing clear Linux doesn't mean x86 is dead they will just contribute directly to the popular distros they will focus those resources there

 

[Darkmont]

https://www.crn.com/news/networking/2025/intel-reveals-plan-to-spin-off-networking-business-in-memo NEX will be spun off

 

[adroc_thurston]

https://www.crn.com/news/networking/2025/intel-reveals-plan-to-spin-off-networking-business-in-memo NEX will be spun off

It's really funny how they hatcheted both FPGAs and networking after AMD got into both businesses.

 

[Joe NYC]

Likely just said the wrong term despite thinking the right concept. You are likely quite under pressure during these calls and presentations.

That 55% share is likely correct, or at least, correct wrt available data.

Gotta remember that the data we see from Mercury Research are just for AMD and ignoring ARM so AMD having like 28 % doesn't mean Intel has 72% of server. There's like 15 % for ARM or therebouts courtesy of all the cloud hyperscalers running Neoverse chips. So 55 % may be what Intel is currently at.

There may be more than one survey, but Mercury releases one that includes only x86 CPUs, so just AMD vs. Intel.

There are 2 metrics: Unit and Revenue share. When the CEO speaks, they are more likely to speak about dollar market share. According to Mercury Research, as of March 31, AMD was at 40%, Intel was at 60%

There is a lot of talk online that AMD is rapidly approaching the 50%. So when LBT says 55%, these are the same metrics: dollar share of x86 server CPU.

 

[Joe NYC]

https://www.crn.com/news/networking/2025/intel-reveals-plan-to-spin-off-networking-business-in-memo NEX will be spun off

What is kind of interesting is that maybe this has been the plan for some time, but for the last couple of quarters, Intel "reorganized" its reporting, stopped reporting NEX as a standalone category and split the revenue into server and client (padding the revenue of these divisions).

 

[Saylick]

https://www.crn.com/news/networking/2025/intel-reveals-plan-to-spin-off-networking-business-in-memo NEX will be spun off

I guess that seals the deal for a fully Intel-based AI rack-scale solution...

 

[Joe NYC]

It's really funny how they hatcheted both FPGAs and networking after AMD got into both businesses.

BTW, one glaring gap in AMD portfolio is in networking on the switch side. My opinion is that AMD needs to make an acquisition here, and buy one (or both) of the companies that are working on UA Link switches. Which appear to be Astera Labs (15 billion market cap) and XCom (privately held). Anybody else?

BTW, and time for AMD to wake up and recognize that Broadcom is not their friend...

 

[Kepler_L2]

I guess that seals the deal for a fully Intel-based AI rack-scale solution...

AMD is like 5 years ahead of them and barely anyone bothers because "muh CUDA" so they have no chance.

 

[Joe NYC]

I guess that seals the deal for a fully Intel-based AI rack-scale solution...

Yup. It almost sounds like the Bible story where St. Peter denied knowing Jesus Christ 3 times.

When talking about Intel's AI strategy, 3 times in the investor call, LBT did not bring up Jaguar Shores once.

 

[adroc_thurston]

I guess that seals the deal for a fully Intel-based AI rack-scale solution...

Intel never had any chances there.
They quite literally sold their in-house server biz.

BTW, one glaring gap in AMD portfolio is in networking on the switch side

Nope.

My opinion is that AMD needs to make an acquisition here, and buy one (or both) of the companies that are working on UA Link switches. Which appear to be Astera Labs (15 billion market cap) and XCom (privately held). Anybody else?

I think you should stop having opinions and instead listen to what Forrest Norrod says.

 

[Joe NYC]

Nope.

I think you should stop having opinions and instead listen to what Forrest Norrod says.

I don't know if it was Forest Norrod or Papermaster, who said they don't want to anger Hock Tan, but either way, in retrospect, it looks like it was a mistake.

Which, I hope AMD corrects pronto...

 

[adroc_thurston]

I don't know if it was Forest Norrod or Papermaster, who said they don't want to anger Hock Tan, but either way, in retrospect, it looks like it was a mistake.

It was Norrod and it is the right decision.

Which, I hope AMD corrects pronto...

That's your opinion that no single exec at AMD shares.

 

[Joe NYC]

It was Norrod and it is the right decision.

Probably the right decision at the time.

But a lot has changed in very short time, and Broadcom went from partner to competitor - to some extent eating AMD lunch in AI deployments.

That's your opinion that no single exec at AMD shares.

We will see. The one piece of puzzle (*) that AMD is missing badly (UA Link switch) is the the segment that Broadcom exited (stabbed AMD in the back), so buying one of the other vendors with promising solution would not even be "angering Hock Tan"

(*) AMD would lo like to have a full rack AI solution with all of the crucial parts being AMD.

Ultra Ethernet Switches - a lot of companies are doing, some even in collaboration with AMD Pensando, but, whoever has a good UA Link switch, AMD will have to buy them eventually anyway, so why not do it sooner (for less money) than later (for a lot more money).

 

[adroc_thurston]

Probably the right decision at the time.

It's the right decision period.

But a lot has changed in very short time, and Broadcom went from partner to competitor - to some extent eating AMD lunch in AI deployments.

Uh, no? no.

We will see. The one piece of puzzle (*) that AMD is missing badly (UA Link switch) is the the segment that Broadcom exited (stabbed AMD in the back), so buying one of the other vendors with promising solution would not even be "angering Hock Tan"

(*) AMD would lo like to have a full rack AI solution with all of the crucial parts being AMD.

Ultra Ethernet Switches - a lot of companies are doing, some even in collaboration with AMD Pensando, but, whoever has a good UA Link switch, AMD will have to buy them eventually anyway, so why not do it sooner (for less money) than later (for a lot more money).

again you should just stop.

 

[Doug S]

What is kind of interesting is that maybe this has been the plan for some time, but for the last couple of quarters, Intel "reorganized" its reporting, stopped reporting NEX as a standalone category and split the revenue into server and client (padding the revenue of these divisions).

I was saying a couple years ago that Intel needed to dump Altera, dump networking, dump everything that wasn't part of their sales of x86 PC & server SKUs and the foundry. Under Gelsinger they were pussy footing around with this stuff, like he was afraid of upsetting senior managers he'd known for years. No more pussy footing around, time to cut loose from all the barnacles that have attached themselves to Intel when they were overly fat and happy.

They've been in networking for a LONG time, I remember Intel NICs in the 90s, Centrino wireless solutions in the 00s so it has been everywhere not just the HPC level stuff. Problem is, other than at the really high end that's all commodity. Why does Intel still design and sell wifi chips? OK yes the answer is obvious - back when they were a near monopoly they wanted to extract more money out of every PC sale so they tied sales of Intel mobile CPUs to Intel wireless chips. Maybe it even made sense beyond that back then since there probably were some fairly crappy wireless chips.

But again this is all commodity stuff today. No one is creating any value add with wifi chips anymore. No one even looks at what wifi version a device supports, they know whatever it is its more than "good enough".

 

[gdansk]

No one is creating any value add with wifi chips anymore.

Intel doesn't need it, but saying it isn't a value add is an overstatement.
Google "site:reddit.com mediatek wireless".
Or "site:reddit.com apple broadcom wireless bootcamp"
Or "site:reddit.com atheros linux"

Intel wireless chipsets are a pretty safe recommendation. That Intel laptops almost always come with a good wireless chip is some value.

 

[511]

not to mention it's well supported in Linux

 

[QuickyDuck]

Intel is trying to pivot to inference which might have some opportunity. Training is a lost war.