News AMD 3Q24 Earnings Results

[jpiniero]

AMD posted the 10-Q... Client sales increase was mainly due to volume... Up 25%. Only a 3% ASP Increase.

 

[randomhero]

AMD posted the 10-Q... Client sales increase was mainly due to volume... Up 25%. Only a 3% ASP Increase.

That is huge, actually. It means they have put one foot in corpo desktop

 

[OneEng2]

AMD's DC revenue has more than double compared to last year.

Unless Intel's DC revenue increases a lot, AMD will for the first time surpass Intel's DC revenue, another blow for Intel

... and this trend is the most alarming one if you are Intel. The high margin DC is the very best market to be gaining in at this time. The business market (laptops mostly) I think will be dampened in the next year or two because of Arrow Lake impact in this market where it is actually a quite good chip.

Intel still doesn't have a compelling answer to Turin, so I suspect that DC trend line to continue through 2025. Intel's next opportunity to turn things around are with Clearwater Forest on 18A. If Intel hits a home run on CWF, they will have about a year to get the slope of that trend line back in order, but will then be facing a Zen 6 competitor in DC (no rest for the weary).

 

[Markfw]

... and this trend is the most alarming one if you are Intel. The high margin DC is the very best market to be gaining in at this time. The business market (laptops mostly) I think will be dampened in the next year or two because of Arrow Lake impact in this market where it is actually a quite good chip.

Intel still doesn't have a compelling answer to Turin, so I suspect that DC trend line to continue through 2025. Intel's next opportunity to turn things around are with Clearwater Forest on 18A. If Intel hits a home run on CWF, they will have about a year to get the slope of that trend line back in order, but will then be facing a Zen 6 competitor in DC (no rest for the weary).

One good example I posted in The Turin builders thread. Supermicro can not provide motherboards that support Turin to virtually any reseller, like newegg, Amazon or wiredzone, since they are all going to assembly (or to be exact I think, to their own server chassis) They are selling so well, they are at max capacity. To repeat, The wording exactly is in the Turin thread.

Here:

Question - Zen 5 Turin builders thread.

OK, only CPU change on a Genoa build, but due to budget, thats first. The CPU (an ES) was not cheap. Stuck at 4800 on memory, but at least I have 12 channels. The final with be DDR5 ECC registered 6000.

forums.anandtech.com

 

[Hitman928]

AMD's revenue officially now at >50% of Intel's revenue.

 

[MangoX]

AMD's revenue officially now at >50% of Intel's revenue.

And with a quarter of their headcount, to boot.

 

[Markfw]

And with a quarter of their headcount, to boot.

And their data center is bigger in sales and profits.

 

[maddie]

And with a quarter of their headcount, to boot.

An Intel employee reading this is going to be very stressed going forward.

 

[DavidC1]

That is huge, actually. It means they have put one foot in corpo desktop

Let me guess. Maybe Raptorlake fiasco?

 

[Jan Olšan]

That is huge, actually. It means they have put one foot in corpo desktop

Or the "whole market is digesting excess stock of laptop CPUs so let's stop buying AMD chips in order to not having to cut Intel orders too much" chapter is finally over after 2 years, nothing more (nothing less)? Just in time for more glut and inventory snafu I guess.

 

[DrMrLordX]

Supermicro can not provide motherboards that support Turin to virtually any reseller, like newegg, Amazon or wiredzone, since they are all going to assembly (or to be exact I think, to their own server chassys) (sp, but no good guesses from the spelling AI)

Not trying to derail but SuperMicro may need their own thread eventually:

Super Micro Computer's shares plunge 33% after server maker's accounting firm quits

Shares of Super Micro Computer have plunged more than 33% after the server maker disclosed that Ernst & Young had resigned as its public accounting firm.

apnews.com

sigh

Let me guess. Maybe Raptorlake fiasco?

@Jan Olšan may be on to something here. That being said, the corpos may be looking at Hawk Point (cheap/readily available) and Strix Point (great js performance, among other things) and thinking to themselves that they sure beat the Intel junk clogging up the channel.

 

[DavidC1]

@Jan Olšan may be on to something here. That being said, the corpos may be looking at Hawk Point (cheap/readily available) and Strix Point (great js performance, among other things) and thinking to themselves that they sure beat the Intel junk clogging up the channel.

The thing is, if the products are worse, Intel was eventually bound to lose marketshare to AMD too. It's about trust. Do you trust that product? People don't really care about 20-30% differences, if they view the better product is from an unknown vendor. If it screws up on you, then even 50% is not worth it.

Sometimes it's all about getting your name out there, even if your product is mediocre. Intel had that trust for decades. Advertising and getting it out there is as important as the product itself.

These things are fractal and boil down to the smallest part, which is individuals. You have two potential hires. Person A is more qualified but you have this awkward feeling about him. Person B is less qualified but he's more charismatic, and easier to talk to. Person B will get hired.

This is more true for Enterprise segments, because it's even more important. You want your uptime for your servers to be as high as possible. What is their after sales service like? Intel had this.

 

[DrMrLordX]

This is more true for Enterprise segments, because it's even more important. You want your uptime for your servers to be as high as possible. What is their after sales service like? Intel had this.

AMD does not have a trust problem in enterprise/datacentre, and hasn't since maybe Rome. I was more thinking corpo laptop/desktop, which is where AMD has had problems cracking the nut. Until now!

 

[randomhero]

Let me guess. Maybe Raptorlake fiasco?

Or the "whole market is digesting excess stock of laptop CPUs so let's stop buying AMD chips in order to not having to cut Intel orders too much" chapter is finally over after 2 years, nothing more (nothing less)? Just in time for more glut and inventory snafu I guess.

Could be both, actually. Perfect storm sort of. Raptorlake problems, no more Intel money to stuff channel and AI desktop gimmick.

 

[Vattila]

It was disappointing that the outlook fell below market expectations.

On the other hand, even with the lower outlook and associated growth projection, AMD valuation is low compared to all-time-high — even compared to the 2021-11-30 intraday high ($164.46), which was well before the completion of the Xilinx acquisition. It's as if the valuation is actually negatively affected by the latter, with the still high amortisation numbers from the acquisition weighing on the GAAP numbers. If that is the case, then we should see better valuation going forward, as the Xilinx-related amortisation eases. It was down to $585M this quarter.

Client Segment: It was nice to see the operating margin recovering a little in the Client segment, but 15% is still dismal. It's as if they are going to all that trouble competing in that segment for fun only. Intel is still desperate to fill their fabs, I presume, so competition is really tough. The silver lining is that it keeps ARM at bay. But, at the end of the day, AMD is a high-performance computing player, so I don't think they should play in commodity markets for no profit — unless playing has some clear secondary benefits.

Gaming Segment: Similarly, the Gaming segment annoys me. I cannot understand why they accept these long console cycles from Microsoft and Sony. I would hope they could innovate in the high end and push the boundaries more. Hopefully, sales pick up with the Playstation refresh, but I think AMD needs to get into the driver seat in the gaming segment. In the PC space, Nvidia has shown that there is a premium segment in gaming. AMD needs to bring the premium segment into consoles as well. Have multiple models; entry, mainstream and premium. Push the boundaries at least every second year. The console game developers, who once insisted on a stable platform for 10 years, have presumably now all adopted the PC-like development approach and library abstractions. I would think they could deal with low to premium tier hardware models. If AMD doesn't innovate here, they risk losing the console segment to mediocrity. Again, they are supposed to be a player in high-performance computing. Other players can probably do cheap and cool better.

By the way, most of the operating expenses in the Gaming segment are related to the discrete gaming GPU developments and marketing, I presume. Operating expenses for the console contracts should be very low. The return on the discrete gaming GPU efforts is disappointing. I suspect this is a segment where Lisa Su has little input and little enthusiasm. And she is yet to find someone with the interest, talent and drive to make something happen here. Radeon looks pretty good to me, so I don't fault the engineers, but it doesn't sell for some reason.

Embedded Segment: On the other hand, I still love the operating margin in the Embedded space. Hopefully, revenue will return with the promised strong pipeline of product wins. This business is also wonderfully sticky, with AMD Xilinx having products that no one can match, deployed by customers for decades.

Data Centre Segment: The Data Centre segment is back to the 29% operating margin a year ago, and with the ramp of high margin GPUs, I would expect this to soon surpass the record 33% margin in 2022-Q1.

Here is my updated spreadsheet:

And here is the overview since 2021-Q2:

(My spreadsheet is available at OneDrive.)

I particularly like the nice upswing in operating income, which should continue to grow nicely as GPU sales ramp.

AI Prospects: Regarding AI, I am confident that AMD will execute well on their roadmap and become increasingly competitive with Nvidia. A nice recent development, regarding overcoming the Nvidia "moat", is that InfiniBand now looks to be supplanted by Ethernet in the large scale training clusters, as well as in the data centres where inference happens. On the software side, it also looks good. AI training is a narrow workload, and many players are working to open it up. AMD's own software stack ROCm is progressing fast and already has good support in the core frameworks.

This will only improve. Unlike the impression you get from the market today, there is little to no chance that Nvidia will be able to run further ahead of the competition. In the long run, Nvidias margins or market share will come down — likely both, as I see it.

Future Prospects: So, looking forward, what are AMD's prospects from here? The scenarios I posted back in 2021 (here) have served me well as a way of thinking about my investment, and the likelihoods I put on these scenarios, which perhaps seemed overly optimistic back then, seem pretty reasonable now. And here is my update from January this year, which puts the likelihood of staying below $200 at 45% and going above $200 at 55%. I think these numbers are reasonable. Let me know what you all think.

AMD longterm prospects (3 years)​	CAGR​	Revenue​	GM​	EPS​	Market cap.​	SP​	Likelihood​
Pessimistic: Underdog, meek or no growth, precarious future​	<10%​	<$26B​	<50%​	<$4​	<$160B​	<$100​	20%​
Stable: Respected competitor in growing markets​	10–30%​	$26B–44B​	50–55%​	$4–7​	$160B–320B​	$100–200​	25%​
Optimistic: Matches competitors, high margins and solid growth​	30–50%​	$44B–67B​	55–60%​	$7–9​	$320B–640B​	$200–400​	30%​
Pie-in-the-sky: Overtakes competitors, leading innovation and markets​	>50%​	>$67B​	>60%​	>$9​	>$640B​	>$400​	25%​

 

[gdansk]

Have multiple models; entry, mainstream and premium.

Consumers hate that. They want simplicity, where a PS5 is a PS5. The problem isn't needing more hardware models but the software. There simply doesn't exist new software to move new units.

 

[Vattila]

Consumers hate [choice].

Consumers without money, maybe. Those with money love buying premium, often just for the status.

 

[jpiniero]

Sony is doing the PS5 Pro. It's probably not going to sell all that great but the option is there.

 

[gdansk]

Consumers without money, maybe. Those with money love buying premium, often just for the status.

Maybe, but they find it confusing. Then their kid complains they bought them the wrong PlayStation apparently. Consoles are to be simple and remove choice. Adding choice makes it a worse experience for 90% of console buyers. Cost reduced updates make sense. Pro refreshes are just a flop in the making.

 

[adroc_thurston]

Pro refreshes are just a flop in the making.

They actually sell OK, but only assuming there are games to play. Alas, PS5 has not a single game on the platform.

 

[mmaenpaa]

AMD does not have a trust problem in enterprise/datacentre, and hasn't since maybe Rome. I was more thinking corpo laptop/desktop, which is where AMD has had problems cracking the nut. Until now!

I have been selling AMD based laptops for my *direct* corporate customers since HP ProBook 6465b which had first (affordable) 4c processor at the time (that model was/is a tank, lasted for ever, and those 4 cpus beat 2c/4t cpus in the end of laptop lifetime).

I also do subcontracting for bigger IT support companies. Yesterday was the *first time ever* I installed an AMD based laptop for enduser. It was LENOVO L14 G5 R5-7535U/14WUXGA/16GB/512SSD/W11P which is 82€ (~8%) cheaper compared to similar Intel model (LENOVO L14 G5 U5-125U/14WUXGA/16GB/512SSD/W11P).

So indeed from my point of view AMD is finally coming to corporation laptop space.

 

[Ghostsonplanets]

AMD needs to bring the premium segment into consoles as well.

? AMD is bound to what their semi-custom clients want. Consoles need to meet a balance between performance and pricing for consumers.

Premium is being explored, as Sony does Pro devices, but it's still a much smaller pie in the console space.

Also, AMD can't dictate what will be the console shelf cycle to their clients. They're bound to when MS and Sony want to introduce a new generation.

 

[moinmoin]

@Vattila Thanks for the write up as as always. The focus on margins disappoints me slightly (that part already destroyed classic Intel) but I guess it "makes sense" in a discussion about stock trading.

I'm wondering what the solution for that would be.

Console business is essentially a completely risk free service business that allows for easy guaranteed income, guaranteed externally paid for R&D, and tightening excellent relations with foundries by expanding their business. I don't see any downside to that at all, except those dreaded margins focused folks constantly putting that in a completely different light.

And with graphics business in the form of dGPUs there is a very strong interdependence with iGPUs, workstation GPUs and DC (nowadays AI) compute. Furthermore dGPU R&D tends to be on the forefront of packaging R&D which then gets used in other areas. How do you accurately reflect this kind of interdependence without killing it with cries of margins and as a result dragging down all the other connected areas with way better margins along it?

 

[DrMrLordX]

Yesterday was the *first time ever* I installed an AMD based laptop for enduser. It was LENOVO L14 G5 R5-7535U/14WUXGA/16GB/512SSD/W11P which is 82€ (~8%) cheaper compared to similar Intel model (LENOVO L14 G5 U5-125U/14WUXGA/16GB/512SSD/W11P).

That puts things into perspective in light of AMD's Q3 numbers. Thanks for the input!

edit: and it's a Rembrandt. Hmm! I wonder how many of those AMD still sells? Not surprised it was cheaper than a Meteor Lake system.

 

[OneEng2]

AMD's revenue officially now at >50% of Intel's revenue.

Not so long ago, that would have been unimaginable. AMD's server revenue is now greater than Intel's. Not so long ago many would have thought you daft to suggest such a thing.