News Intel 2Q25 Earnings

[Doug S]

Intel was under 80K employees in 2010, and they seemed to be able to execute just fine then. Why the anguish over them slimming down to that size? Yeah it sucks for the employees, and for the fact that the uncertainty will make some of the good ones who would not be laid off to leave on their own.

Intel has acquired companies and tried and failed to enter into other businesses in the years since 2010. Remember when they thought they could sell x86 mobile chips to go into phones and tablets? Remember when they bought Altera and thought they could make FPGAs a standard part of servers? Remember when they paid $15 billion for Mobileye? How many times have they claimed they would be making discrete GPUs - whether for gaming, HPC, or most recently, AI?

If they get rid of all those failed and folly lines of business and concentrate with laser focus on two things - matching or beating AMD in x86, and matching or beating TSMC on the leading edge process and manage to at least win a few midsize customers (they need to show success with those before they have a chance at landing whales like Apple and Qualcomm) why should they need more employees than they had in 2010?

Yes it looks ugly but it is necessary. They can no longer afford to waste time and effort on anything that doesn't drive them towards the two main goals. If they can turn things around only then should they worry about anything that isn't directly involved with reaching those goals.

 

[gdansk]

Lack of SMT isn't a problem, lack of execution is. SMT obviously does not help with that though, because validation penalties are in every generation.

I assume developing two different micro-architectures causes more validation problems?
But they're in a pickle here. Bread and butter Xeons are all P cores. Current x86 cores achieve higher performance per area with SMT. What's the cheapest way to put themselves on better competitive footing with Zens until Unified Core (Messiahmont) is ready?

 

[Geddagod]

In a different universe.
Different reality, perhaps?

I doubt DMR vs Venice Dense is any worse than GNR vs Turin Dense is rn

No.

They seem to be quickening their pace of innovation. Or at least trying to. Two tocks in a row.

Intel was under 80K employees in 2010, and they seemed to be able to execute just fine then. Why the anguish over them slimming down to that size? Yeah it sucks for the employees, and for the fact that the uncertainty will make some of the good ones who would not be laid off to leave on their own.

Intel has acquired companies and tried and failed to enter into other businesses in the years since 2010. Remember when they thought they could sell x86 mobile chips to go into phones and tablets? Remember when they bought Altera and thought they could make FPGAs a standard part of servers? Remember when they paid $15 billion for Mobileye? How many times have they claimed they would be making discrete GPUs - whether for gaming, HPC, or most recently, AI?

If they get rid of all those failed and folly lines of business and concentrate with laser focus on two things - matching or beating AMD in x86, and matching or beating TSMC on the leading edge process and manage to at least win a few midsize customers (they need to show success with those before they have a chance at landing whales like Apple and Qualcomm) why should they need more employees than they had in 2010?

Yes it looks ugly but it is necessary. They can no longer afford to waste time and effort on anything that doesn't drive them towards the two main goals. If they can turn things around only then should they worry about anything that isn't directly involved with reaching those goals.

Very much agree with you. Get rid of client discrete graphics, at the very least.

 

[jpiniero]

Intel was under 80K employees in 2010, and they seemed to be able to execute just fine then. Why the anguish over them slimming down to that size?

Need way more employees per product with more recent nodes.

 

[gdansk]

Why the anguish over them slimming down to that size?

The layoffs have been across the company. Based on what I could find it doesn't look like a laser focus to me.
But Intel has so many good candidates to cut/cancel that I think each new leader has faced a paralysis of sorts.

 

[csbin]

55%

https://twitter.com/x/status/1948566911748636897

 

[Joe NYC]

It is a problem in a world where AMD ships it.

So, Diamond Rapids will lose 10 to 30% MT performance just by losing SMT... And it will not surprise me if the core still contains SMT circuitry, just inactivated.

 

[adroc_thurston]

And it will not surprise me if the core still contains SMT circuitry, just inactivated.

Oh no it doesn't.

 

[Joe NYC]

Oh no it doesn't.

Then, the option of adding it back in probably does not exist for Diamond Rapids and Diamond Rapids will be a lost generation

Then, adding it to Diamond Rapids "next" will take extra time and in that time, Intel will be down to 33% server market share in ~2028 at the time it ships...

 

[Doug S]

Need way more employees per product with more recent nodes.

You talking about design teams? How many chip designers does AMD employ? Is there any reason Intel needs more of them than AMD has?

AMD has more than tripled its number of employees since its low point in 2016 - tripled to 28,000 as of last December. 8,200 employees as of Dec 2016 was all they needed to launch Zen 1 and put themselves back on course CPU wise, while continuing to iterate their far more successful than Intel's (even if they aren't as good as Nvidia) GPU line.

Yeah I get that fabs require a lot of people. TSMC has 83,000 employees, but their operations dwarf Intel's by so much it isn't even funny. So if AMD can do what they did with 8,200 people, and TSMC can do what they do with 83,000, I can't believe anyone is going to defend Intel needing more than the 70-75K people are worrying about Intel cutting down to. That's probably still a lot more than they really need. AMD showed you can do a lot without even cracking 10K.

 

[gdansk]

all they needed to launch Zen 1 and put themselves back on course CPU wise

Zen 1 was very minimum-viable-product.
To put together a system with working sleep states, reliable chipsets with fewer USB dropouts and so on I suspect you need more than AMD had at the time. AMD was lucky that people bought Zen 1. They survived by Polaris / console revenue. I don't think Intel won any consoles and their GPUs are made at TSMC so they can't undercut the competition. AMD had the GF wafer contract and cheap Samsung capacity which gave them a plan B with cheap Polaris spam. Intel won't survive if they regress that far.

 

[adroc_thurston]

that people bought it and survived by Polaris / console revenue.

Oh no that wasn't ever relevant (it's a fairytale for 30iq reddit goymers, by 30iq reddit goymers).
Graphics had never made real money before RDNA2.

 

[gdansk]

Oh no that wasn't ever relevant (it's a fairytale for 30iq reddit goymers, by 30iq reddit goymers).
Graphics had never made real money before RDNA2.

Reading AMD 2017 financial statements I get the opposite impression. But alas, you got the real numbers I'm sure Mr. 42%.

 

[adroc_thurston]

Reading AMD 2017 financial statements I get the opposite impression.

Graphics hasn't been the FCF driver, like, ever.

alas, you got the real numbers I'm sure Mr. 42%.

Indeed I do.

 

[gdansk]

Indeed I do.

Then you know that they reported

Our major customers, Sony Interactive Entertainment LLC and Microsoft Corporation, each accounted for more than 10% of our consolidated net revenue for the year ended December 30, 2017. Sales to Sony Interactive Entertainment LLC and Microsoft Corporation consisted of products from our Enterprise, Embedded and Semi-Custom segment.

Total it was more than 30%. And that wasn't on the strength of Zen. Polaris and Vega were in another segment altogether and were a large portion of it.

Not that it's material to an Intel thread but I do not recommend they get into a situation where they're shipping products as crappy as Zen 1. There's no other product at scale large enough to keep them alive as they iterate improvements. Well maybe OEMs ordering Raptor Lake forever could.

 

[adroc_thurston]

Total it was more than 30%

That doesn't mean it was actually driving profitability at any level required to sustain a proper roadmap.

 

[Timorous]

Another US company. Also Nvidia and Apple. No nation outside the US is going to do this.

Semi conductor supply and manufacturing is critical for more than just computing. There is a national security angle to the EU having fab capacity as well as it just being good for continuity.

Also ASML is in the EU and they are the No1 supplier of bleeding edge lithography machines.

The semiconductor shortage really hammered the car industry due to how reliant various modules are on ICs and that is a critical part of the EU economy.

 

[511]

Ian cuttress said 15% workforce to 75K btw for employees count they separated altera from the employee count + 15% employees laid off

 

[511]

This is karma. The sooner they rid themselves of their IDF legacy, the sooner they can rise anew. Until then, it's scary to think about a world without Intel since they have shaped everyday computing for more than 30 years of my life. Can AMD shoulder that burden alone? Do they actually BUY Intel for scraps, thus ensuring a less scary future of personal computing?

They have shaped the IC manufacturing since their inception lol

 

[511]

Another US company. Also Nvidia and Apple. No nation outside the US is going to do this.

They don't manufacturer anything in their product lineup

 

[adroc_thurston]

They have shaped the IC manufacturing since their inception lol

kids are *this* close to figuring out who Robert Noyce truly was.

 

[511]

kids are *this* close to figuring out who Robert Noyce truly was.

He is nicknamed "Mayor of silicone valley" for a reason lol

 

[MS_AT]

55%

https://twitter.com/x/status/1948566911748636897

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:

In addition, I have instituted a policy where every major chip design is reviewed and approved by me before tape-out. This discipline will improve our execution and reduce development costs.

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

 

[Fjodor2001]

Comments from Intel CEO, Lip-Bu Tan, after the recent earnings call:

Intel Confirms Nova Lake To Close The High-End Desktop CPU Gap With AMD In Late 2026, Bringing SMT Back To P-Cores With "Coral Rapids" On Servers By 2028-2029, Highlights Need To Consolidate On Xe GPUs, & 55% Server Share

Intel confirms that Nova Lake will close the high-end desktop gap with AMD in 2026, while Coral Rapids will bring SMT back to servers.

wccftech.com

Key Points:

• First Panther Lake Mobile SKU Launching In Late 2025
• Panther Lake Ramp & More SKUs Planned In Early 2026
• Nova Lake Aims To Bridge High-End Desktop CPU Gap With AMD
• Nova Lake Launches In Late 2026 for mobile and desktop platforms
• Intel 18A to be the main driver for at least three generations of client/enterprise products
• Diamond Rapids P-Core CPUs with up to 256 cores In 2H 2026
• Clearwater Forest E-Core CPUs with up to 288 cores In Mid 2026
• Coral Rapids P-Core To Replace Diamond Rapids By 2028-2029
• SMT Coming Back To P-Core With Coral Rapids Server Being The First To Reintroduce It
• Consolidation and Building Upon x86 CPU and Xe GPUs
• Intel 14A targets 2028-2029 timeframe, tackles TSMC's A14

 

[Quintessa]

Yep. Despite modest revenue uptick Q/Q, the gross margin erosion is brutal. Capex for 18A and foundry bets (like Tower, IMS) eating them alive. They're bleeding today for maybe winning tomorrow. But the roadmap is long, and patience is thin on Wall St.