I have been interested in the microprocessor business since I was in grade school, and have followed the "war" between Advanced Micro Devices (AMD) and Intel (INTC) for well over a decade. I remember the triumphant launch of AMD's K7 and K8 designs, soundly beating Intel's "Netburst" series microarchitectures that formed the basis of the Pentium 4, Pentium D, and various derivatives. I remember seeing AMD cleverly beat Intel to producing a nice, fairly elegant set of 64-bit extensions to the IA-32 instruction set, and I remember when the AMD Opteron was schooling Intel's Netburst-based Xeons on CPU performance and power efficiency as well as the actual system architecture.
I remember when AMD was great, and I will never forget the small fortune that I made buying shares of AMD at about $7 ahead of the run to the $40s (I got out at about $33, though, when I saw the evidence that Intel's "Conroe" would indeed put an end to AMD's "reign"). To those who suggest that I have some inherent dislike, you couldn't be more off the mark - AMD paid for many years' worth of new computer gear for me.
Ever since those halcyon days of the K8, AMD has stumbled year after year with products that time and again fail to hit the mark. AMD's follow-on to the K8 design, "Barcelona," was hyped up significantly as the "Intel killer" (since Intel had since taken the performance crown). One tech press outfit even claimed that AMD's engineers were "dancing in the aisles" when they saw the initial performance of the chips. Well, "Barcelona" fell flat on its face - marred by low clock speeds, an inferior micro-architecture, and a performance-crippling bug.
In its drunken stupor, AMD, under the leadership of then-CEO Hector "The Wrecker" Ruiz, levered up its balance sheet (forcing AMD to sell its fabs to stay solvent) and acquired ATI for a substantial premium to market price. While the "promise" of this acquisition was that AMD would quickly integrate ATI's graphics IP into AMD's microprocessors, it wasn't until 2011 that the first "Fusion" (that was the "buzzword" of the day - similar to what "HSA" and "Mantle" are today; take heed ye newbies to the AMD story) parts hit the street. By then, it was already too little, too late. The integrated graphics portion of the first Fusion chip, codenamed "Llano," was quite good, but its CPU performance was weak, and power consumption wasn't great - it failed to deliver the K.O. against Intel that the company so desperately needed to gain back market share and pricing power in PCs.
Of course, not all was lost. AMD's low power CPU effort - known as "Brazos" - was quite good. While AMD's high end share (that is in both PCs and servers) continued to wane, AMD's low cost, but high volume Brazos parts did quite well in cheap notebooks and "netbook" systems. These parts were superior to the Intel Atom parts that were being peddled, and as a result became one of AMD's best-selling products in its history. Everything seemed okay, even if it wasn't perfect.
Of course, with Apple's (AAPL) introduction of the iPad, the low end of the PC market (AMD's bread and butter) was the first to get hit - AMD's 2012 was marred by multiple guide downs. Intel wasn't immune either as its 2012 forecast was lowered three times, with the third driven by a big revenue warning for Q3. While Intel remained massively profitable, AMD began to lose some serious money during 2012 and in the first half of 2013. Newly appointed CEO Rory Read and his team implemented a restructuring plan - lay off a bunch of workers (freeing up talent to be gobbled up by the new wave of ARM chip vendors), cut SG&A, and stop trying to go head-to-head with Intel at the high end PC market and instead target "high growth" areas like tablets and micro-servers, as well as semi-custom.
It's a great plan on paper - but it's not really going to work like many AMD bulls think it will. Sure, semi-custom is going to work out pretty well - AMD's IP and current situation makes it ideal for this market. After all, AMD has a lot of great GPU IP, can license whatever ARM IP that it needs, and even has some decent X86 processors that it can use. It's not a bad gig.
But the problem is that AMD bulls aren't just betting on semi-custom, but are also betting on micro-servers, low cost notebooks, and tablets to save the day. Sorry, but these expectations are just completely unrealistic. While AMD joins the "ARM bandwagon" for micro-servers, the truth is that using ARM is a headwind for competing in any sort of server market, not a tailwind. It's absolutely puzzling to me that AMD has no plans to fight in this market with its own low power X86 cores - the vast majority of the server market is X86 today, and Intel's new products suggest that this will be the case for many years to come.
Further, while AMD saw a "spike" in its low cost notebook chip sales ("Kabini" and "Temash"), and while management (and AMD bulls) pumped up the imminent arrival of "Temash" for tablets, your friendly Seeking Alpha author pointed out that it wasn't time to get excited over these for the following reasons:
· The one-time spike in "Kabini" and quad core "Temash" sales for low cost notebooks would not be sustainable as AMD was simply exploiting a one-quarter window of opportunity before Intel rolled out low cost Haswell chips as well as its Bay Trail-M (Atom based, low cost) parts for this market and aggressively competed for share
· The "3.9 watt" variant of "Temash" was completely unsuitable for tablets. In addition to being woefully underpowered on the CPU side, as well as having nothing meaningful in the way of dynamic voltage/frequency scaling, the "Jaguar" architecture itself lacks the deep active-idle states needed to see really good battery life. It also lacks key IP blocks on the chip itself like an integrated Image Signal Processor ("ISP").
· AMD was on the verge of being hit by the 22 nanometer refresh from Intel in the server space, likely further pressuring AMD's limited server market share.
It gets worse, however. The effects of the Bay Trail-M ramp on AMD's P&L won't really be evident until Q1 2014 and throughout most of 2014. Further, in micro-servers, while AMD plans to ship its first 28 nanometer part by the end of the year, Intel will be on its third generation micro-server part that not only will be built on 14 nanometer FinFET technologies (that's two nodes ahead of AMD), but will also be an X86 design, which is a big plus from a software compatibility standpoint in the datacenter. AMD will also continue to feel the pressure from "Haswell" throughout the balance of 2013 and into early 2014, and then once mobile "Kaveri" (AMD's "big core" APU for notebooks) ships, it will have to compete with 14 nanometer "Broadwell-ULT." The competitive situation is just as bad as it gets for AMD - it's competing against better funded and larger design teams, as well as a two generation process generation lead. This, of course, explains why AMD's "big core" server roadmap is largely absent.
At the end of the day, it's important to be realistic. It's great to think you're holding on to the next multi-bagger story (it may be, who knows?), but at the same time there's a reason AMD is worth less than what Intel makes in profits in a good quarter - it's got a lot of problems and despite what management tries to do, the odds of "success" aren't exactly stellar. This is, at best, a speculative stock, and for most it's just a trade on whatever "story" happens to be in favor today. After all, something's got to be seriously wrong when after over 3 decades of being a publicly traded company, shareholders still haven't ever seen the following announcement:
AMD declares quarterly cash dividend.