News AMD 3Q23 Results

[Hitman928]

Overall back to growth but not very significant. Will do more breakdown as I have time.



For the fourth quarter of 2023, AMD expects revenue to be approximately $6.1 billion, plus or minus $300 million. At the mid-point of the revenue range, this represents year-over-year growth of approximately 9% and sequential growth of approximately 5%. Non-GAAP gross margin is expected to be approximately 51.5%.

 

[sandorski]

Nice.

 

[Hitman928]

So AMD didn't have a strong beat over guidance like Intel, but they also didn't have as much of a fall to recover from. Similar to Intel, they are expecting a seasonal uptick in revenue going into 4Q. Consumer segment had 42% y/y growth so strong recovery in consumer, much like Intel reported.

Biggest downside for the quarter and guidance for next quarter is gaming revenue is down due to semi-custom sales being down. Data center revenue was flat y/y but up 21% q/q, so they are seeing good recovery in this segment. They had record sales of Epyc processors but Xilinx product sales were down. Embedded segment also down for the quarter.

AMD previously had been projecting annual revenue that would mean much higher revenues in 4Q, so they will miss that projection, but are still on track for near pandemic level (record level) revenue to close the year.

 

[Hitman928]

Segment table:





Link to full slides:

https://d1io3yog0oux5.cloudfront.net/_bb564a866bc41f45b16c329e8171eb02/amd/db/778/6935/file/AMD+Q3%2723+Earnings+Slides.pdf

 

[trivik12]

AMD should be taking more market share while intel is down. Otherwise its not intel who will come back but ARM and may be Risc-V which will win the war. I am expecting both client and DC to be bloody in next 2 years. Qualcomm launching Oryon and many DC chips with ARM and AI chips with ARM/Risc-V.

 

[adroc_thurston]

AMD should be taking more market share while intel is down

mss is worthless unless you can command satisfactory margins.

but ARM and may be Risc-V which will win the war

Both are meme.

I am expecting both client and DC to be bloody in next 2 years.

No, it's a stomp, in DC in particular.

Qualcomm launching Oryon and many DC chips with ARM and AI chips with ARM/Risc-V.

8cx g4 is middling and server SoCs are just woefully inadequate versus the Turin duo.

 

[tajoh111]

Disappointing outlook. Amd outlook should be higher since this is where some of mi300 revenue should be coming in.

What is the limiting factor for mi300 revenue? Just for comparison, Nvidia data center revenue is going to be 12 to 14 billion in q3(not total revenue).

The total addressable market center for ai is huge right now and may shrink in the perhaps not too distant future(When the bubble bursts). Amd is missing the boat.

 

[Hitman928]

Disappointing outlook. Amd outlook should be higher since this is where some of mi300 revenue should be coming in.

What is the limiting factor for mi300 revenue? Just for comparison, Nvidia data center revenue is going to be 12 to 14 billion in q3(not total revenue).

The total addressable market center for ai is huge right now and may shrink in the perhaps not too distant future(When the bubble bursts). Amd is missing the boat.

MI300 production starts towards the end of the year but doesn't ramp until first quarter next year. I'm guessing it will hit full production probably around end of March or April 2024.

 

[trivik12]

Snapdragon X Elite is looking really good. Apple can expand their market share if they release cheaper macbook which is probably happening next year. NVidia Grace is looking really good and we have custom ARM CPUs from Amazon and I am sure other Cloud OEM. I think x86 is in trouble in both Client and Server.

 

[itsmydamnation]

Snapdragon X Elite is looking really good. Apple can expand their market share if they release cheaper macbook which is probably happening next year. NVidia Grace is looking really good and we have custom ARM CPUs from Amazon and I am sure other Cloud OEM. I think x86 is in trouble in both Client and Server.

how many times have we heard this song and dance ? about every time an arm entrant releases something that might be up to par.......

 

[tajoh111]

MI300 production starts towards the end of the year but doesn't ramp until first quarter next year. I'm guessing it will hit full production probably around end of March or April 2024.

Su is reporting 2024 revenue from mi300 to be 2 billion for the year. Probably conservative but not terribly exciting for growth for Amds overall revenue for such a huge market. E.g If all else stays the same Amd becomes a 24 billion dollar company vs the 22 billion revenue company it is today.

Amd needs more revenue growth for r and d considering the manpower and engineering hours for smaller nodes and all the markets AMD is involved in. E.g ai and supposedly arm chips. Otherwise cuts like the recent ones to RTG will happen.

 

[Abwx]

AMD should be taking more market share while intel is down. Otherwise its not intel who will come back but ARM and may be Risc-V which will win the war. I am expecting both client and DC to be bloody in next 2 years. Qualcomm launching Oryon and many DC chips with ARM and AI chips with ARM/Risc-V.

As i pointed in the other thread you can hardly gain marketshare in DC if your competitor is selling its chips at a loss, last quarter Intel sold them at zero net margin, so AMD captured a meager 300M revenue out from 500M lost by Intel, remaining 200M are surely due to both the X86 DC market slightly shrinking and the ARM camp taking a few sales.

 

[Joe NYC]

mss is worthless unless you can command satisfactory margins.

It is true(ish) between two fabless companies, such as AMD and NVDA.

But for Intel, manufacturing scale is a weapon that Intel can wield against competitors - or it could be own sword to fall on, below certain sales threshold.

So, any additional sales AMD achieves are disarming Intel.

 

[Joe NYC]

Disappointing outlook. Amd outlook should be higher since this is where some of mi300 revenue should be coming in.

What is the limiting factor for mi300 revenue? Just for comparison, Nvidia data center revenue is going to be 12 to 14 billion in q3(not total revenue).

The total addressable market center for ai is huge right now and may shrink in the perhaps not too distant future(When the bubble bursts). Amd is missing the boat.

For Q4, it is the size of the market for Mi300a (El Capitan mostly) and a trickle of Mi300x that can be brought to market.

Going forward, I think it is mainly limited by demand.

I think AMD mentioned that they have supply secured for 2 billion worth of revenue, but there is an upside.

My interpretation is that AMD is mainly limiting the outlook to sales that have already been secured. But there could be additional sales...

 

[Joe NYC]

Snapdragon X Elite is looking really good. Apple can expand their market share if they release cheaper macbook which is probably happening next year. NVidia Grace is looking really good and we have custom ARM CPUs from Amazon and I am sure other Cloud OEM. I think x86 is in trouble in both Client and Server.

If Mac sales and market share are going anywhere, they are going down...

 

[adroc_thurston]

If Mac sales and market share are going anywhere, they are going down...

Yea lol.
Granted, Macs are cyclical devices and this product cycle has been bad for Mac (new MBPs were a nothingburger and the 15" MBA was outright a flop).

 

[Markfw]

I will ask a 2 sided question. How long can Intel keep losing money on their server products before they become bankrupt/ how long can AMD put up with Intel losing money on server chips before they sue Intel for anti-competitive behavior ?

 

[Doug S]

how many times have we heard this song and dance ? about every time an arm entrant releases something that might be up to par.......

Yep, dunno why people assume ARM is going to start eating market share all of a sudden.

The reason for previous market moves like mainframes to minis, minis to RISC workstation/servers, RISC workstation/servers to x86 PC/servers is because the replacement was CHEAPER, not because it performed better.

I don't see any evidence that Qualcomm's ARM PCs are going to be sold for less than the equivalent Intel/AMD PC. Knowing Qualcomm they'll price them for hundreds more, and try to claim having built in cellular is an advantage. They'll probably be priced like Apple, with no low end and little midrange. Even if Apple does add a cheaper Mac it will be functionally like the iPhone SE (though it won't be priced nearly that low) and provide a limited midrange but still nothing on the low end.

Custom ARM CPUs from Amazon etc. will likely remain for their internal use and only be sold as a cloud product, so other than maybe limiting the worse excesses of Intel/AMD hyperscale CPU pricing (Intel/AMD have a LOT of room to come down on those prices in response to competition) they won't make much difference.

 

[yuri69]

Disappointing outlook. Amd outlook should be higher since this is where some of mi300 revenue should be coming in.

What is the limiting factor for mi300 revenue? Just for comparison, Nvidia data center revenue is going to be 12 to 14 billion in q3(not total revenue).

The total addressable market center for ai is huge right now and may shrink in the perhaps not too distant future(When the bubble bursts). Amd is missing the boat.

MI300 is a 2024 product.

It seems MI300's primary design target was the El Capitan deal. Just like MI250's was the Frontier. Just look how MI250 sells outside the super-computers... Let's hope MI300 design target was not so narrow.

 

[controlflow]

I will ask a 2 sided question. How long can Intel keep losing money on their server products before they become bankrupt/ how long can AMD put up with Intel losing money on server chips before they sue Intel for anti-competitive behavior ?

Intel Data Center group had an operating margin of 2% and made an operating profit of $0.1B in Q3. The guidance for next quarter is better. I don't think they will be going bankrupt from that...
This is a pretty good outcome for them given how bad Sapphire Rapids is in comparison to AMD offerings.

As far as anti-competitive behavior, can you elaborate more on this?

 

[lightisgood]

MI300 is a 2024 product.

It seems MI300's primary design target was the El Capitan deal. Just like MI250's was the Frontier. Just look how MI250 sells outside the super-computers... Let's hope MI300 design target was not so narrow.

Lisa said that MI300 is only $0.5B / quarter product.
This is very poor performance.

Intel Gaudi2 achieved 2~3x perf in 3Q23.
Gaudi3 would achieve more good perf in next year, too.

NVIDIA is, as everyone knows, the most heavyweight.

I think AMD is completely missing gen AI opportunity because was so occupied with HPC (e.g. Frontier, El Capitan).
This blunder may cut Lisa's CEO life.

 

[Joe NYC]

Lisa said that MI300 is only $0.5B / quarter product.
This is very poor performance.

$0.5B would be about 30% increase in Datacenter revenue for AMD, that is recurring, unlike supercomputer sales. I don't think it is poor performance.

I think it is a conservative projection based on sales / contracts AMD has visibility on as of now.

Intel Gaudi2 achieved 2~3x perf in 3Q23.
Gaudi3 would achieve more good perf in next year, too.

What do you mean 2-3x performance in Q323? I don't think Gaudi2 had revenue anywhere close to $500m in Q3.

NVIDIA is, as everyone knows, the most heavyweight.

I think AMD is completely missing gen AI opportunity because was so occupied with HPC (e.g. Frontier, El Capitan).
This blunder may cut Lisa's CEO life.

So is everyone else (missing the AI opportunity).

While the Mi300a for El Capitan seems like a decent product, it may not have been a 100% perfect fit for the LLMs of today, so AMD tweaked the design to make the Mi300x.

Mi300x is likely going to end up being the most powerful datacenter GPU for multiple workload categories. Which is a positive testimony to the flexibility of the basic Mi300 design, and ability to pivot to market demands.

And I think for that, Lisa Su will give himself a raise.

 

[lightisgood]

What do you mean 2-3x performance in Q323? I don't think Gaudi2 had revenue anywhere close to $500m in Q3.

You can see Intel's earnings report.

While the Mi300a for El Capitan seems like a decent product, it may not have been a 100% perfect fit for the LLMs of today, so AMD tweaked the design to make the Mi300x.

A tweaking is just being tweaking...
It is not optimized architecture.

Mi300x is likely going to end up being the most powerful datacenter GPU for multiple workload categories. Which is a positive testimony to the flexibility of the basic Mi300 design, and ability to pivot to market demands.

If you run after two hares, you will catch neither.
yuri69 is concerned about this.

 

[Gideon]

Semianalysis has a very bullish article on MI300 ramp (in the upcoming quarters):

AMD MI300 Ramp, GPT-4 Performance, ASP & Volumes

Order Volumes From Microsoft, Meta, Oracle, Google, Supermicro/Quanta direct, Amazon

www.semianalysis.com

 

[Joe NYC]

You can see Intel's earnings report.

Intel earnings report does not break out Gaudi revenue. So, I have no idea what you are referring to.

But in a separate interview, Intel CFO said that Intel hopes Gaudi will achieve its first $100 million revenue sometimes in 2024.

Intel is about as far behind AMD as AMD is behind NVidia in datacenter GPU.

If you run after two hares, you will catch neither.
yuri69 is concerned about this.

I think you have a misconception about Intel's participation in datacenter AI accelerator market.

NVidia is leading, AMD is far behind NVidia. Intel is part of a pack that is far behind AMD.