You can't ever really be sure about these types of things, but the recent downgrade from Goldman Sachs (GS) of shares of Advanced Micro Devices (AMD) may be a trap. Anybody following my writings knows that I'm not an AMD cheerleader by any stretch of the imagination - I am fully cognizant of the problems that the company faces going forward and that Mr. Rory Read and team will need to pull some real rabbits out of their collective hats to keep this company viable. That being said, I believe that the "hype" is nowhere near over, and as we enter the second half of 2013, AMD's return to profitability could drive further upside in the shares before it becomes time to "sell."
The Game Console Perception
Have you ever noticed that the markets get giddy about things that they've already "known" about? I'm fairly confident that when Microsoft (MSFT) goes ahead and announces its next generation "Xbox," it will be powered by both an AMD CPU and GPU, very similarly to what Sony's (SNE) Playstation 4 will come packed with. My view is that when this deal is formalized, the sell-side as well as the buy-side will start to run the numbers on what kind of upside this could engender for AMD's revenue and earnings base going forward. The conclusion is likely to be that in the very worst case, it helps soak up wafer commitments at GlobalFoundries, and that in the best case, these parts have pretty decent gross margins. It is also important to note that the vast majority of the IP for these part(s) is already paid for by AMD's CPU and GPU lineups in the traditional PC space.
It's a good business for AMD, for sure, and represents the first step towards becoming less of an "Intel (INTC) competitor" and more of a unique semi-custom SoC company that has some unique IP to deploy into interesting segments. In short, the hype for this is nowhere close to over, in my view.
Kabini and Temash May Disappoint Long Term, But Short-Term Hype Plentiful
I believe that in the short term, there will be significant hype around AMD's place in the "tablet" world, which will further juice the stock. While many of the die-hard AMD fans have believed that AMD magically stuffed a quad core, super-fast-GPU SoC into a 5.9W power envelope with "Temash," the first reviews of the chip paint a different picture. Quad core "Temash" is an 8W part at 1GHz and 15W at 1.4GHz ("turbo-dock'd"), so this is not something that is in the same power league as Intel's "Bay Trail" or ARM's (ARMH) various Cortex A15 implementations for smartphones and tablets. According to the aforementioned review,
So, speculations that the A6-1450 has a TDP between 8 and 15 Watt depending on the Turbo could be true: If the CPU runs at its maximum clock under full stress, the power consumption would be nearly 15 Watt. However, an official TDP specification of AMD is still missing.
Yikes! Remember how in an article from a few months back, I made the following observation?
There's no doubt about it: the numbers are quite good. In "tablet" mode it's about twice as fast as an Intel 2 core/4 thread Atom Z2760 and in "hybrid" mode it edges out the Intel 2 core/4 thread Core i3 based on the 2011, 32nm "Sandy Bridge" architecture. Generally speaking, it's easier to get more theoretical performance by throwing on more cores than it is to make single threaded performance better, but it's still interesting nonetheless.
The caveat here is that Temash isn't playing in the 2W tablet chip space like the Atom Z2760 -- it's an 8W TDP chip in "tablet" mode and a 15W one in "hybrid" mode.
Our resident AMD permabulls tried to claim that I was "bashing" AMD and "pumping Intel" (really, guys, you think I am trying to move a $120B market cap stock?), I was simply applying my knowledge of this industry to come to reasonable conclusions. With no process advantage, a low R&D budget, and dramatically higher performance, did you really expect magic? There was a compromise, and it came at the cost of power consumption.
My guess is that AMD's dual core "Temash" has a very tough time competing with Intel's "Baytrail" in the Windows tablet market, but it does reasonably well in the budget Windows PC market, as did its predecessor "Brazos." One thing, though, is that battery life seems to be not all that great according to that review,
In battery life, the small 30 Wh battery is obstructive. The user can only expect a battery life of about 3 hours during web browsing with medium display brightness and active WLAN.
Ouch! 3 hours of battery life? While I believe that the small battery is part of the problem, AMD's "Jaguar" design lacks a lot of the really interesting power saving features that ARM, Intel, and Qualcomm integrate into their cores thanks to their phone/tablet principal focus. AMD has created a really good, low power, low-cost PC oriented chip, but this is not a design with a "tablet first" mentality - and it shows. Whether this is the right design target or not for AMD to maximize profits remains to be seen, but don't be surprised when Microsoft doesn't put "Temash" in the next Surface!
Microservers: Going ARM To Save Money?
Semiaccurate, run by the ever-entertaining Charlie Demerjian, ran an interesting piece regarding AMD's new micro-server/ARM server strategy. I have maintained for quite some time in my writings that this move is primarily to save on the costs of developing the right core IP for this space, and that thesis has been proven spot on from the following remarks given in the piece,
By Mr. Feldman's count it takes 400 million dollars and three years to build an x86 server chip. But only 30 Million dollars and 18 months are required to make an ARM chip. As far as AMD is concerned the ARM ecosystem offers massive cost and time-to-market advantages over building traditional x86 chips.
It's really cheap to build an SoC when you don't have to actually design, build, and validate a CPU core, and are just putting together building blocks, the most expensive of which is licensed. The disadvantage of course is that everybody else is using the same core IP (and you have to pay royalties on each chip), so you have to really differentiate on the SoC side of things and give up having control over the core. This isn't an "ARM versus X86" thing as Mr. Feldman seems to try to push - it's that AMD as well as the little band of startups trying to play in this space don't want to pony up the bucks build their own cores.
Anyway, this is actually a little bit puzzling in light of the fact that AMD has a low power "Jaguar" core, but as I showed above, it's not in the same power class as the Intel "Atom" and the ARM Cortex designs, so you have more powerful cores, but you can put on few of them - and for these workloads, the cores don't need to be too, too powerful so more is really better in many cases.
But at the end of the day, AMD can't really do worse than its current server lineup, which has <5% market share against Intel, so if it can sell more volume, increase total revenues, and maybe have a shot of having non-trivial market share in this new growth area, why not?
Graphics Profitability Continues To Vanish
I recently penned a piece in which I pointed out that the "Never Settle" bundle (in which AMD gives away games and hurts GPU division's operating profitability) was costing AMD a bundle. Nvidia (NVDA) reported a very strong quarter at the high end of their guidance range along will killer margins and ultimately great profitability. This was driven by GeForce strength (Titan, anybody?), coupled with strength in professional/HPC with Quadro/Tesla.
The pro-AMD camp tend to think that the game console deals will suddenly make everything rosy for AMD's GPU division and put an end to Nvidia, but I can almost guarantee this is not the case. The Xbox 360 is powered by an AMD GPU, but this hasn't dented Nvidia's market share. AMD's "switchable" graphics (Enduro) is still inferior to Nvidia's, and Nvidia's developer relations isn't exactly going to go by the wayside because the game consoles have AMD chips in them. I also don't buy for a second the argument that discrete GPUs will go away anytime soon; no gamer would give up his or her beefy 200W discrete GPU for AMD or Intel integrated graphics. Also keep in mind that the consoles "reset" the baseline for gaming graphics, which may actually drive demand for higher performance PC GPUs as game developers raise the graphical bar.
While I still don't exactly buy the "blue skies" scenario that many AMD bulls perpetuate, this is still a trade that probably has some room to run before it's time to sell. There is a lot of mass "hope" for the game consoles, the "Temash/Kabini" parts, and even the ARM micro-server parts. They may or may not actually bring AMD the fortunes that the shareholders expect, but AMD claimed profitability in Q3 and Q4, so if we actually hit that target, the shares could go much, much higher before coming back down, even if the rosy long-term picture turns out to be exaggerated.