AMD (AMD) bulls can rightfully pat themselves on the back. AMD is now earning more per share than the cost of its stock price several years ago. The turnaround from its near-bankruptcy has been successful. However, as such, I would argue the investment opportunity for growth in AMD is quickly fading as it is becoming a mature company, operating in mature segments like PC, data center and graphics.
Nevertheless, looking at the stock and the general commentary by AMD bulls, it seems investors are expecting this fairy tale to continue indefinitely, with the company now having overtaken Intel (INTC) in market cap.
Hence, I would argue AMD investors are now seemingly expecting growth that simply may not exist. Even worse, investors are only contemplating all the upside without considering any of the risks. This is indeed a risky proposition, and hence the case for AMD being a long-term investment stands on very shaky grounds.
This analysis is a follow-up to my previous coverage of when AMD was nearing Intel’s valuation: AMD Is Not Intel. As an aside, the title for this article was chosen by Twitter. Combined with a previous article (Advanced Marketing Devices), I have now changed the AMD acronym to Advanced Marketing Delusion.
First and foremost, this article is not about the following one or two years. Given AMD’s current growth, its valuation at 30x P/E, and which is expected to drop to 25x P/E, isn’t all that unreasonable. As mentioned, AMD has been very successful.
What bothers me, however, is that I haven’t seen a single AMD bull or analyst or investor, in general, contemplate about what could happen to the stock and the company in the long term, in light of the Intel turnaround (which I have discussed here: Upgrading Intel: Torrid Tech Turnaround).
Indeed, a quick glance at the Seeking Alpha AMD page will show that analysts are almost unambiguously bullish on the stock, with just a few being neutral at worst.
1. Waning product competitiveness
So to get right to what I am referring to, I will provide one example as a case in point. As most will undoubtedly be aware of, over the last few weeks, some comments by Intel CEO Pat Gelsinger have been going around, where he said AMD is now in the rear-view mirror. This, in turn, has even led some analysts to make parodies on this comment, for example, by suggesting that this must mean that Intel is driving backwards.
So clearly, all AMD bulls are having a great time by making jokes about Intel. What could possibly go wrong? Well, a lot. The very first mistake already is that AMD bulls are neglecting any possibilities of downside: what if Pat Gelsinger was actually correct? Indeed, the whole point of this article is to point out that investors may be starting to live from quarterly report to quarterly report like the prototypical Wall Street analyst, expecting a beat-and-raise at every stage without considering what is actually happening at a larger, longer-term scale in the market, and the risks that are becoming apparent – which could be ignored when AMD was still in its growth stage – when a company is becoming $200B in size.
So for starters, the mistake AMD bulls made when they decided not to take Pat Gelsinger’s comment seriously is that they forgot what made AMD successful in the first place: everything begins and ends with product leadership. So what AMD investors ignored or missed is that Pat Gelsinger was talking about feeds and speeds. Not revenue, earnings and P&L. To put it blunt: Alder Lake is a leadership product. It has beaten AMD on every single metric imaginable (single-threaded performance, multi-threaded performance, connectivity and I/O, pricing and even performance per watt), as well as in every single market segment from laptop to desktop.
Why does this matter? Because, as I have had to painstakingly become aware of in my analysis of Intel, products usually trailblaze revenue by a year or more. For example, Intel announced that it sold 100 million Tiger Lakes to date (Alder Lake’s predecessor). While this seems impressive, and certainly dwarfs AMD’s scale, it is still a minority compared to Intel’s total 2021 CPU shipments of 250-300 million CPUs.
So this is a significant distinction to make and one that AMD bulls clearly didn’t make since they unanimously made fun of Gelsinger’s comment when they saw AMD’s latest earnings report: as if AMD’s Q4 results were “proof” that Alder Lake isn’t a leadership product technologically speaking. However, the simple reality is that Alder Lake launched in mid-Q4 in just the first few desktop SKUs. The rest of the line-up of >60 SKUs and hundreds of design wins, with which Intel can perhaps win back market share, will only launch and then further ramp over the course of 2022.
I am not saying this is what will happen and that AMD faces an immediate revenue and market share decline, but I am just saying that AMD bulls are ignoring the threat of Intel’s improved product line-up (embodied by Alder Lake) simply because they didn’t see any evidence in AMD’s Q4 results. This is short-sighted.
Okay, so the next thing AMD investors will surely do is point to the data center, but the exact same rebuttal applies here: Intel is still just in the early innings of the ramp of the very first CPU in its multi-year roadmap to regain product leadership.
As a case in point, Intel reported it sold a record number of server CPUs in Q4 – although the number was not disclosed, its previous record was 8 million – but simultaneously also said it shipped 1 million Ice Lake CPUs. This means the majority (7/8ths) of what Intel sells in the data center is still its old and deprecated 14nm line-up. The point is that investors should not assume that AMD will always be at such an advantage, and when that time comes, AMD’s momentum and financials are at a real risk of not performing like they currently are.
Indeed, in the worst case, Granite Rapids (on Intel 3) in 2024 will have up to 112-144 cores, with each core being faster than AMD’s merely 96 Genoa (N5) cores (putting Intel at least at a 17-50% advantage). In addition, given that TSMC’s (TSM) 3nm capacity is being bought by Intel and Apple (AAPL), AMD may not transition to N3 until 2025, at which point Intel will have its Diamond Rapids CPU on the superior 20A process ready.
2. Intel regaining process leadership
As another case in point, Intel has been talking about “5 nodes in 4 years” for the last year, but Intel has yet to even launch the first of these 5 nodes into the data center (Sapphire Rapids). Admittedly, though, Intel will skip one of these five nodes, and data center also seems to be a year behind, with 20A only launching in 2025 in the data center, in all likelihood.
So again, you don’t have to be Lisa Su in order to take market share currently when the competition is so weak. Everyone could be CEO of AMD right now and claim their victory lap for growing sales. Hence, as stated previously: the real test for AMD is yet to come and investors are either assuming it won’t matter or aren’t even contemplating this reality in the first place. Indeed, AMD bulls are publicly banking on Intel missing its process timelines (for their AMD thesis to play out), despite there being no evidence that Intel is behind on any of its 5 nodes. Intel in fact has pulled in 18A from Q1’25 to H2’24.
In conclusion, it is misguided to assume that everything will stay exactly the way it was after Intel has ramped those 5 nodes into the market over the next 2.5 years. (As I have discussed in other articles, in reality, there are actually only two full nodes, those which were previously called 7nm and 5nm, but the argument stands.)
3. Running out of growth markets
Thirdly, I consciously called AMD a mature company above. Investors are yet again behind by assuming that a $20B revenue company can stay in growth mode forever. It can’t. The simple reason here is that the market simply isn’t big enough for this to be possible. For example, in the PC space, a lot of debates are going on about what the “new reality” for the PC is after the incredible surge in PC demand in the wake of COVID-19: the PC market in two years grew from 250M to 350M units.
When AMD was still tiny a few years ago, all the talk was about taking market share from Intel. However, with AMD now having become a substantial alternative in the market and with competition higher than it has been at any other point in the last few decades or more, the only substantial and reasonably effective way for AMD to continue to grow going forward would be due to the general growth of the market.
Hence, here AMD is subject to those same discussions that investors and analysts to date have (arguably erroneously) solely focused on regarding Intel: will the PC market keep growing, and what if it declines back to its previous 250M level? Although Intel’s view is that 350M is the new reality going forward, the risk of a decline (in the aftermath of the shortages) is nevertheless existent, and not a single person will argue there is a real opportunity for substantial further growth. Ergo, the PC isn’t a growth market, and so at some point, AMD is poised to run into a brick wall – with or without Intel.
A similar argument could be made for the console space, which is another market that has existed for decades. Obviously, the new console generation launched just a year ago and there are still severe shortages (Sony (SONY) is actually behind in its shipments compared to the PS4), but a more prudent investor would again contemplate what will happen in the longer term. As happened the last generation, at some point, sales will fizzle out, and worse, AMD’s margins will become thinner.
In the absolute worst case, Intel might actually try to win at least one next-gen console and in one fell swoop take out about half of AMD’s console market share. Note that with Intel’s foundry business, Intel might not even have to “try” since either Sony or Microsoft (MSFT) in principle could simply design their next-gen console chip based on Intel IP without any intervention from Intel. Although some may see this as just a theoretical possibility, note that both Microsoft and Sony are (in theory) heavily incentivized to go this route, since bypassing the AMD middleman will result in lower cost, and consoles are well-known for their ultra-slim (or even non-existent) margins.
So overall, I am not saying AMD doesn’t have any opportunity for growth – it does – but clearly, AMD today already is facing substantial long-term growth issues in two-thirds of its business. So this actually provides a very bleak view, and which I would argue is not something investors should see as a beauty mark: the observation that the data center has now become AMD’s last remaining bastion for growth.
As discussed at length by now, obviously cloud is a nice and growing market and Intel’s current product portfolio is downright embarrassing, but so again I repeat my argument that the more sophisticated long-term investor will ask what will happen when this isn’t the case anymore (starting in 2024), which AMD investors aren’t bothered to contemplate yet. Again, it depends on your time horizon.
4. Xilinx dilution
This brings me to my fourth and last argument, which is the Xilinx elephant in the room. There is a lot that could be said about this, and Xilinx’s competitive position, which I actually have spent a whole article about a few months ago: AMD Vs. Intel: Battle For FPGA Leadership. But in short, and relevant to the discussion here, is that I would argue the Xilinx acquisition provides a one-time boost to AMD’s revenue, but this has come at a very severe cost: share dilution.
There are several ways I could make this argument, but overall, I would argue that there is a substantial risk for AMD investors that this acquisition over time will produce subpar returns (compared to the sum of the parts prior to acquisition), which is a quite paradoxical observation I have noticed in the stock market.
So what could happen is that while over the next year AMD’s growth rate will be artificially propped up by the inorganic Xilinx contribution, AMD shareholders could be in for a rude awakening in 2023 when they have to face substantially lower growth rates due to the commensurately tougher comps. This in turn will only exacerbate all the arguments so far in this article about competition from the likes of Intel and Nvidia (NVDA), as well as the aftermath of the shortages causing AMD to run into a brick wall with regards to the seemingly infinite growth targets of its shareholders.
I have two examples as evidence for this thesis – and which suggests this thesis may already start to play out in 2022, well before the “tough comps” for AMD start in 2023. First, in telehealth, there was a blockbuster merger between Teladoc Health (TDOC) and Livongo Health. One might think the market (back in 2020 when the market was still favoring growth) would have been excited by a combined telehealth powerhouse reporting triple-digit growth, but a quick look at the stock will show that at no single point this was the case. The stock dropped on the day of the merger announcement, and except for a brief rally has only declined since.
Secondly, Intel is currently actually doing the opposite by spinning off Mobileye, reportedly for a $50B valuation (which is similar to Xilinx’s market cap). However, contrary to AMD adding $50B in market cap due to the Xilinx merger, not a single (informed) person would expect that the Mobileye spin-off will result in the opposite of the current AMD merger: Intel shedding $50B in market cap when it loses Mobileye.
Both examples “prove” (indicate) that the combined result in the stock market often ends up as being worth less than the sum of the parts.
In line with this, one could even provide a third example, which is the Intel Altera acquisition. Although Intel funded the acquisition with cash instead of stock, ultimately Altera has shown to be a low-growth piece that ultimately has not provided any substantial contribution to shareholders. Note that I explicitly say shareholders: for Intel (as a business), Altera has actually been a very strategic acquisition that has played a very key role in Intel’s position in segments such as SmartNICs/IPUs/DPUs and 5G networking. In that view, the acquisition has been successful strategically but not particularly so financially.
Of course, perhaps AMD shareholders may be expecting a different result in Xilinx’s case, but isn’t the very definition of delusion to do the same thing and expect a different result? AMD is doing exactly the same as what Intel did half a decade ago, so a more prudent assumption would be not to expect any different results.
In summary, AMD bought itself a very, very expensive piece of revenue growth, and this was largely funded by existing AMD shareholders who got substantially diluted. In fact, it is obviously this merger that has caused AMD to surpass Intel in market cap, but as discussed, this raises the question for how long this will be the case, since going forward, the market will no longer treat AMD and Xilinx separately.
Some Seeking Alpha contributors have tried to make a case that the acquisition would yield a net contribution to AMD’s earnings, but the author seemed to be posing question marks himself: “What isn’t clear is whether the deal remains accretive with AMD growing sales at a 50% clip and Xilinx slightly beating consensus estimates”.
While math may suggest that something should always equal the sum of its parts, the stock market doesn’t always work like that. As such, the argument provided in this article could be interpreted in two ways.
First, AMD has been successful and has capitalized on its (growth) opportunity. The turnaround investment thesis has played out with AMD reaching a $200B market cap and in fact becoming the size of Intel, which itself has announced plans to regain industry leadership and will actually build $200B worth of new fabs in the next decade alone. AMD investors aren’t contemplating yet this new reality in both growth and market competitiveness dynamics going forward over the next few years. As argued, at some point, AMD will run into a brick wall in all of its segments.
Secondly, one could consider that the sole reason for AMD’s latest surge in market cap hasn’t been due to any increased investor sentiment, but rather due to the inorganic Xilinx piece that is being added (absorbed) and will drag down the growth rate in 2023. While Xilinx has higher gross margins, its growth has been very mixed (and also arguably has already been overtaken in product leadership by Intel’s Altera).
As such, Xilinx (worst case) could end up just like the console business, being a cyclical piece with low to no long-term growth prospects. In fact, when also adding the PC business to these two markets, one simply has to conclude that three-quarters of AMD’s business is now in segments with no real growth prospects.
This leaves the data center (cloud) as AMD’s and AMD shareholders’ last bastion that they have to clamp onto. This clearly isn’t the sign of a strong business but rather leaves the impression of a maturing one that has to face the strongest competition that it has ever faced in the next few years, from both Nvidia and Intel. For example, both of these companies have dwarfed AMD with regards to their prepayments to TSMC. Hence, AMD is poised to get back to its former position of being the technologically inferior (“cheap”) competitor.
In conclusion, AMD shareholders are correct insofar that they observe that nothing is what it was. But likewise – not so long from now – nothing will be what it is now. Since the AMD thesis is played out, investors may be advised or contemplated to look for a new investment opportunity. For example, as one such proposal, when it comes to turnarounds, there is one such option nearby competing in AMD’s very same markets: Intel.