AMD is at a crossroads. Recent product design wins give it a chance to become highly profitable. Yet competition and changes in user patterns could further reduce AMD's sales of traditional products. This article will present an in-depth qualitative analysis of relatively long-term scenarios (1 to 3 years out).
First, a couple of short-term catalysts. While I was preparing to write this article, on May 16 Goldman Sachs downgraded AMD and the stock price plunged. But this had followed a run-up of the stock from a 52-week low of $1.81 on November 16, 2012 to $4.40 on May 15. The 52-week high had been $6.46 back on May 29, 2012.
The run up was on the presumption that AMD would be designed in to all three major global gaming consoles [Nintendo, Sony (SNE), and Microsoft (MSFT)]. Nintendo and Sony are known wins, but there has not been an official announcement on the next generation Microsoft Xbox. That will come on May 21. If everyone is wrong, and AMD does not have the slot, that would be a short-term negative catalyst and an upside for the company that did get the slot.
The gaming console silicon is a good place to start to examine the competitive landscape. Why would the perpetual runner up in CPUs and GPUs (graphics processing units) win spots in all three gaming consoles?
Intel (INTC) is currently crushing AMD in high-end CPUs and in the mass market. But Intel is relatively weak in graphics. It has thrown a lot of R&D dollars into catching up, and is closing the gap, but is still roughly a generation behind AMD and Nvidia (NVDA). Nvidia, on the other hand, is arguably neck-and-neck with AMD in graphics, and generating better margins at the same time. But Nvidia decided that rather than compete against Intel and AMD in the x86 CPU market, it would take its expertise into the ARM CPU market. The gaming consoles required more computing power than ARM designs could provide at present. Using Nvidia for graphics would have meant using a discrete GPU with a discrete CPU, which would have increased costs without increasing usable performance.
Outside the gaming console market AMD has been in trouble. It has been nearly pushed out of the server market. It does not really compete in the high-end CPU market for PCs, and it is weak in the midrange market. Even in the mainstream, low-cost market, where it is competitive, in order to maintain market share it accepts lower ASPs and margins than Intel. In the discrete GPU market, where gamers, video editors, and others will pay for faster graphics performance, AMD ASPs have tended to match Nvidia overall, but margins have been much worse.
For AMD to have a comeback it needs to sell products with better margins across its product line. Since it does not have the economies of scale of Intel, nor is it likely to gain those economies in the next few years, it needs to succeed based on its new competitive strategy.
The core of that strategy is to exploit parallel processing, derived from its use in graphics chips. This merging of GPU and CPU on a single piece of silicon has been in progress for some time. It appeared first in low end and midrange PC chips. In effect, at the stage of process technology available to AMD in 2011, it put a low-end GPU and low-end CPU on the same silicon. This had two advantages. An on-silicon interface is much faster than a chip-to-chip interface. It also lowers design costs, since one socket in the motherboard did the work of two sockets previously. These APUs (advanced processing units) in a variety of flavors sold reasonably well.
But the technology trend setters did not care for them much; they were for mere consumers and ordinary office workers. Their graphics were good enough for most purposes, but technology reviewers favor products that combine faster GPUs with faster CPUs. Another problem was software. A lot of software would run faster if it was directed to the GPU instead of the CPU, but that was not a priority for most software companies. Also much of the work that was done was on Nvidia's proprietary CUDA system, which could not be run on AMD chips. The increasing use of OpenGL and OpenCL is gradually alleviating that particular bottleneck.
The transition to APUs coincided with another trend: the rise of the smartphone and the abandonment of netbooks for tablet computers. APUs were perfect for increasing the quality of netbooks, but they consumed too much power to run tablets or smartphones. They have done reasonably well in the $300 to $600 notebook and desktop markets.
The PC market started shrinking in 2012, leading to predictions of its demise, but I have seen the end of the PC market predicted too many times to go along with that hypothesis without critical examination. There is another trend worth noting: decreasing costs of large screens. 24 and 27 inch screens are now very reasonably priced. They really help increase efficiency. It no longer makes sense to buy a screen smaller than 24" for a non-mobile worker, or for home use. People will want and need a good graphics processor to feed their larger than ever screens. If cost is an issue, AMD has the best solutions. Its APU processors have plenty of compute power and can handle large screens without difficulty. There will be another PC refresh cycle at some point as Windows XP and computers running it die. Smartphones and tablets don't even make good word processors. The cloud is fine and good, but having actual processing power and graphics capability at your finger tips is crucial both to core business functions and to entertainment.
It is conceivable that AMD could gain enough share in the PC market to compensate for any overall market declines. At the same time, Intel's own graphics+CPU solutions are improving, so it may come down to marketing and OEM relationships, which favor Intel.
AMD could also see a comeback in the highly-strategic server market. That market is now dominated by datacenters and cloud computing. Increasingly graphics and other parallel computations must be performed on the server before the result is sent out to the end user. So far Nvidia has dominated this market, AMD has products available and an edge with its APU technology. When we see AMD Opteron server chips incorporating its graphics cores and its fabric communications technology acquired from SeaMicro, it could be a game changer. Nvidia can't put graphics and an x86 CPU together on a single chip. It may compete in ARM-based servers, but these are just starting to hit the market. The real danger here is the slow speed of development by AMD has given Intel the chance, with its massive R&D budget, to catch up in this area. There are no product announcements by either company yet, but I would not be surprised to see Opteron chips with parallel processing features in 2014.
AMD has to be selective with their R&D dollars. They have to keep up the pace on CPU and GPU development while continuing their work integrating the two architectures. They can't go on generating losses as if they were a startup company. They have said they are trying to not go head-to-head with Intel, but it is hard to give up on a market segment.
One reason the gaming triple crown is important is that it can generate a guaranteed stream of profits while showing the potential of APUs. It can also increase AMD brand awareness, which is important when consumers and even business IT departments are making buying decisions.
It also means games will likely standardize on AMD, OpenCL, and OpenGL. If you look at reviews of gaming cards (at say, Anandtech.com), if they are in the same price range you will typically see that some games run faster on the AMD card and some faster on the Nvidia card. This is because of software choices made by the game designers. If those designers standardize on AMD optimization (to hit all three consoles plus AMD GPUs), those test numbers will begin to favor AMD. Then AMD can gain market share or increase margins. But this is not likely to start tracking until 2014.
Despite the optimistic tone of this analysis so far, the main factor in the space has not changed. Intel remains a profit machine. In the past, whenever AMD has made headway, Intel has thrown engineers, price competition, and marketing dollars into the space to prevent AMD from making significant profits or holding onto market share in the long run. In addition AMD needs support from software developers. If an application can't take advantage of the potential of an AMD APU, its parallel processing superiority won't give it an advantage in the marketplace. Adobe (ADBE) has started to support AMD parallel processing efforts, as has Microsoft , but much more development is needed.
Going forward, I believe the most important strategic decisions AMD management can make are when not to compete, either in new or traditional product lines. I am willing to see how the next generation server chips compete, but if market share remains under 5% for long, I'd like to see AMD concentrate on arenas where they can win. By winning, I mean showing substantial margins that can support both ongoing R&D and begin to show earnings that will be reflected in the stock price.
Today while I was writing AMD's stock price bounced about half-way back from its Goldman plunge. Leaving it (at $4.07 per share) with a market capitalization of $2.9 billion. Visibility is limited. We know AMD's game plan, but not how end markets will react, or if OEMs will even allow AMD to reach end markets. Revenue in Q1 was $1.09 billion, and is likely to be only slightly better this quarter. Normally a revenue run rate of $4 billion a year in the semiconductor industry would support a much higher market capitalization, but then normally people run companies to make profits.
I believe AMD will return to non-GAAP positive EPS in Q3, or at latest in Q4, based largely on sales into the gaming console market. But even if the new gaming consoles are a hit, they tend to be a seasonal product. So Q1 2014 would be my litmus test. Can they make a profit in the seasonally weak first quarter? I know that is a long ways off for most investors. I am likely to hold my shares until then and may even buy on irrational dips. But I would want to see what chip demand and margins are like in the various segments in the coming year before making any bigger investment in AMD.
Additional disclosure: I have received money from a Microsoft contractor for work done for Microsoft in the last 6 months & over the last 10 years.