AMD Q423 grew Revenue at 10 percent

Stocks A to Z
1

I was going over the AMD report and talking to another investor. I said that AMD was underwhelming as an investment and in relationship to NVDA. He said it was apple and oranges. He didn’t expound on it or give his reasoning why so I thought I would discuss it with you all and see what you think.

Lisa Su keeps calling out inferencing and the data center on her call. I know some of it has to do with servers but she is also talking about gpu’s. She was calling for GPU revenue in 2024 to be around 2 billion dollars and now she is calling for 3.5 billion dollars.
That seems pretty underwhelming when NVDA is calling for 20 billion dollars from 14 billion dollars in their Data center.
SMCI just reported the other day and they are a server company that makes servers in the Data Centers and they had 3.6 billion in Revenue up 103 percent YoY and guiding for 4.1 billion next quarter up 219 percent YoY.
AMD is guiding for Revenue to be flat next quarter YoY. The PC refresh isn’t expected to happen yet, Gaming and Embedded are supposed to be down. I just do not understand why AMD stock has gone up at all. It all seems to be built around hope. Hope that they can get to 20 percent Revenue growth while NVDA is growing over 200 percent. So getting back to apple and oranges.

While I get it that AMD plays in different markets than NVDA they are calling out Inference, where they compete directly with NVDA, as one of their high growth areas. NVDA has been taking them to the wood shed in that sector. Would you all consider AMD and NVDA to be apples and oranges? Also, why would you give AMD a higher P/E multiple to NVDA? What makes AMD a great investment going forward?

Andy

1 Like
2

It took AMD several years to get into double digit marketshare against INTC in servers, if will take just 1 year for it to get into double digits against NVDA. The conference call validates that MI300 is a legitimate contender based on customer references. So the key is to see AMD aspirations in this space taking wings and the momentum to continue to grow. Since market is forward looking it sees that, hence the premium

3 Likes
3

I am not so sure of that. They are guiding flat next quarter. Where are the big wins? NVDA is coming with the B100 chip could it be that the customers are waiting for it? I am wondering if the market has AMD to far over their ski’s. Even by the end of next year they are only projecting 20 percent revenue growth. It seems underwhelming with the other companies doing so well in the Data center.

Andy

4

They lower guide is influenced by significant drop in embedded and gaming
Embedded – Inventory Correction while overall business healthy and gaining marketshare
Gaming - In the 5th year and drop is typical.
The other 2 segments Client and Datacenter are healthy and growing. The datacenter piece is where everyone is focused on. They upped the year target from 2B to 3.5B. I expect them to up the target again and end the year with $6 to $7B and try to get to close to 10% of NVIDIA. A stupendous achievment. Post that, the silicon engineering advantage of AMD (in chiplets, interconnects and packaging) will see the momentum grow further. In the big scheme of things the next Q guide is a nothing burger

3 Likes
5

Except the new chip that NVDA is putting out, the B100 is based on chiplets.

Andy

6

Silicon engineering (chiplets) is not an on off switch. AMD has several generations of chiplet experience under its belt (plus 3D packaging,…) . NVDA cannot be expected to match that expertise in Round 1. For a failed chiplet into, there is a datapoint of Intel Ponte Vecchio. I fully expect AMD to gain market share from here on out. They have a team already deep into development of M1400 which they will announce at some point. AMD is in a good place. They are the only company in the world which is #1 or #2 in (CPU, GPU and FPGA)

3 Likes
7

Well it isn’t showing up now and it isn’t showing up this year either. They are only projecting 20 percent growth this coming year and NVDA is already projecting over 200 percent for the next quarter.

Andy

8

Hope is the motivation for all investments, isn’t it?

Whether you prefer Nvidia or AMD as your investment, depends on what you are looking for, I guess. Both are strong capable companies smack in the middle of a technology revolution driven by the hugely promising magic we call artificial intelligence. However, if you invest in Nvidia now, you are probably not looking for more than a 2x return, as that would require Nvidia to be valued more than the current most valuable company in the world. With AMD, you are looking at a potential 5x+ just to match Nvidia’s current valuation. There is simply more room to grow.

Also, for Nvidia to hold its value it has to successfully defend an already astronomical growth and market share in the long term. Meanwhile, AMD starts from a very low base, but has to prove they are capable of taking share.

From a perspective of capability, there is nothing preventing AMD from doing what Nvidia is doing in the AI market. Much is made of the “CUDA moat”, but in the AI field, CUDA is not much of a moat. The AI frameworks are highly abstracted, mostly using the Python scripting language, not low-level C++ dialects like CUDA. What these AI frameworks need is a powerful backend for doing all the matrix mathematics required for neural networks. Backends have so far relied on CUDA, but ROCm is now a potent alternative.

In particular, PyTorch, the leading framework for AI developments, now has support for the ROCm backend out of the box. Note that this means that all changes to the framework API and implementation code, as part of the further development of PyTorch, now need to run validation tests to ensure they work well with the ROCm backend. Likewise, key frameworks from Microsoft (ref. DeepSpeed) and OpenAI (ref. Triton) for optimising large-scale AI now have ROCm support. HuggingFace has partnered with AMD on providing ROCm support in all their countless open-source AI models. Hence, the CUDA moat has pretty much been overcome in AI.

In the wider HPC market, Nvidia is doing great things reliant on their CUDA software, such as Omniverse and traditional supercomputing. Here their software offerings are much broader than AMD’s. But this is not the big growth driver compared to the AI wonder. And, by the way, AMD is doing pretty well in HPC as well, having won the world’s current most powerful supercomputers. Some of the largest language models in the world are being trained on AMD-powered supercomputers today (ref. LUMI).

Nvidia has been tremendously innovative in GPGPU, CUDA and AI in particular. Can they continue to innovate and stay ahead? And can they do so without giving up market share? I think they will do very well, but I think it is inevitable that AMD will get a sizable piece of the pie.

4 Likes
9

I think this is the crux of the question Vattila, and thanks for your response. I think AMD has missed this cycle and their hope has to be that they can get more marketshare next cycle. Nothing is written in stone and NVDA isn’t any slouch in competing. So I think NVDA will do really great this cycle and we will just have to wait 2 years to see if AMD can gain. But you are right when you say hope is the motivation for all investments, but the numbers are the grease to that hope. We will see. Thanks again.

Andy

1 Like
10

I’m not sure where you’ve derived these numbers but aren’t they a bit apples and oranges? In November NVDA projected next quarter (i.e., the one that just ended and will be reported in 3 weeks) growth of 230%. But that’s on a base of Q4 23, which predates the huge ramp up in Q2 24. Sequentially, they forecast 10% growth, or 48.5% annualized, which is obviously great but not 200%.

Maybe the best comparison is with analyst estimates compared with the most recent quarters.
AMD: $6.17B in Q4 23 (Oct-Dec 2023); $26.0B est for FY24 – 5% above 4x latest quarter
NVDA: $20.00B est in Q4 24 (Nov 23 - Jan 24); $93.4B est for FY25 – 17% above 4x latest quarter

So NVDA will likely grow revenue faster at something like 15-20% annually vs ~5% for AMD, unless something surprising happens. I have no crystal ball, but I think the big “surprise” was the emergence of ChatGPT, not something in the future that accelerates their production or makes the world want GPUs even more than it already does.

Just as a disclaimer, I am long AMD and don’t own NVDA. I compared them in summer 2022 and settled on AMD as I thought it had a more diverse line of products to grow and greater headroom, despite NVDA’s dominance in GPUs. Overall, I found NVDA to be a bit stronger of a company, but was pricing in too much growth. Then GPT happened :man_facepalming:. But I won’t complain about a triple from my lowest purchase and a >100% return overall. Now they both seem to me to be valued beyond a reasonable future and I may look to trim when I can get back to redoing my bull/bear DCF models.

Softie

2 Likes
11

When I first joined the fool late last century the environment was very different. Companies were far more willing to share real information. The AMD board was very technical about the making of chips. Today that information is harder to find.

My impression (and I could be wrong, I’m aging, post-covid, and making more mistakes, especially posting on dialysis days, which is not today) is that all of AMD’s critical production is purchased from TSMC. TSMC is wafer limited, so TSMC’s customers are wafer limited. I don’t know the percentage of good chips per wafer (yield), but assume that since NVidia also uses TSMC, and Intel started using TSMC a few years ago, that almost everybody is dependent on getting more wafers by TSMC growing capacity. Intel still has fabs, but for about the first time ever has clearly lost the process lead to somebody else. For a while I considered that Intel was buying wafers to limit AMD’s access to more wafers while AMD had the better products, but Intel’s TSMC chips are going into high end Intel products. Intel was passed by TSMC, and now has near endless supply of lower end products. That is where Intel has gained market share from AMD.

Side note: I suspect Intel will replace AMD in the next generation of gaming consoles. AMD has been capacity restrained for years, and repurposing everything they make into higher margin products will grow the revenue without costing much in expenses. But that’s a suspicion; I have seen no articles on that.

So NVidia and AMD (and many others) will get more chips from two processes. TSMC will make finer lasers or other ways to make chips (which is hard and expensive as all existing fab labs must be upgraded) and they must build new foundries, which is only expensive. TSMC is indeed working hard to build outside of Taiwan in the hopes that they can survive as an international company in case China takes over Taiwan, so they have political problems affecting their choices as well. Finer lasers mean smaller die sizes, smaller die sizes mean less power loss and more efficient chips as well as more good chips per wafer, which will increase the chips everybody can get from TSMC wafers. More fabs means more wafers. Both benefit all TSMC’s customers.

NVidia IMO did a great job of forecasting how things would go, and focused on extreme high end products whose sky high prices would be justified by savings in Tflops per dollar of operations. As electricity costs are rising (I expect a tipping point there as renewables replace fossil fuel power plants) that makes $20,000 chips reasonable from savings on power and cooling the server farms. AMD did a good job, maybe even a great job, in deciding they couldn’t compete with NVidia at the highest margin and focusing on beating their oldest competitor, Intel. Intel is trying to return to a culture and profits that they lost. Nobody is too big to fail, but demand is so high for chips in all kinds of products that their fabs will guarantee Intel a revenue stream. Those fabs are why Intel (of those three) is not capacity constrained. Intel could probably double the price of a lot of low end products, which would do great in the short term, but that will offend customers and at some point it will present a market opportunity for AMD. AMD’s architecture is much more flexible than Intel’s, so they can far more easily repurpose their limited supply of chips into whatever niche is showing the greatest return.

I think all three companies will do okay until something changes. I already mentioned a Chinese invasion of Taiwan, which would be awful for a lot of folks worldwide but good for Intel. Intel could also catch or pass TSMC in production; they certainly are spending enough to get back on top. Other bad things could happen which are beyond my perception right now. But barring those two events, NV will make the highest margins, AMD will continue slowly taking high end margin products from Intel, and Intel may wind up struggling to survive or find a pathway to turn low cost high volume into enough profit to survive and maybe grow.

Fool On!
Richard

3 Likes
12

Hi Softie, I was quoting numbers YoY. But you are correct. Last quarter NVDA projected YoY Revenue growth of over 200 percent. This coming quarter they project YoY Revenue numbers of over 400 percent and 37 percent sequentially. Q424 revenue 20 billion Q423 Revenue 3.8 billion

If you think that NVDA is only going to grow at a 15-20 percent annually then you are projecting a large decrease in Revenue growth. If you take the 20 billion run rate that they will have this coming quarter and project it out for fiscal year 25, 20 bill * 4 it should be 80 billion for the year which would give them a 63 percent annual growth rate. 49.1 billion run rate for fiscal 24 80 billion run rate for fiscal 25 = 63.26 percent .

I have no problem with AMD and would start a position in it if it drops much lower but at this price I will just be patient. I would start taking a position around a P/S of 5 or so.

Andy

13

I am trying to estimate what their current growth rate is, now that the huge step-change in demand from GPT is in the past. In Q2 and Q3 of FY24 their revenue exploded from $7.19B > $13.51B > $18.12B. They guided for Q4 to be $20.00B, which would represent 231% YoY growth over Q4 FY23’s $6.05B, but “only” 10.4% QoQ growth. Maybe you are referencing the data center segment only? I didn’t find projections for that.

Apologies, though, I did do my math a bit wrong. The right growth rates for NVDA and AMD for the coming 12 months are 28% and 9%.

A constant growth rate through FY25 that takes them from $80B run rate today and finishes the year at the $93.4B consensus estimate is 6.3% QoQ or 27.6% annualized:
$21.26B + $22.60B + $24.02B + $25.53B = $93.4B

For AMD it’s 2.1% QoQ and 8.7% annualized:
$6.30B + $6.43B + $6.57B + $6.70B = $26.0B

Unless the consensus estimates are way off, the growth rates are going to moderate substantially.

1 Like
14

I apologize Softie I was just looking at the Data Center and projecting it out and then putting out the yearly growth.

I don’t calculate it that way Softie. I take YoY and QoQ growth estimates. The way you are calculating it doesn’t show the explosive growth that is happening. It levels it out for one year and makes the company look like it is growing much slower than it actually is. I think by looking at a YoY growth shows the strength of the growth and then the QoQ will show if the growth is slowing down.

Here is a look at my spreadsheet.

image
image1600×860 109 KB

I am not saying i am right but it’s the way I choose to look at it. I can understand going by the Analysts viewpoints but if you do that with NVDA I would think you would always be behind. I will take it quarter by quarter because they didn’t get this spike in growth correct so why would they get the growth going forward correct? But let’s see if your numbers are close at the ned of the year. I will book mark it , thanks.

Andy

2 Likes