AMD’s Mobile Strategy

Pages: 1 2

In the last year, AMD has gone through a number of wrenching changes. CEO Dirk Meyer was pushed out in early 2011 over a disagreement on mobile strategy. The company was in a holding pattern until Rory Read was appointed as the new CEO earlier this year. Rick Bergman, who was responsible for the product groups at AMD, almost immediately left to take the top job at Synaptics. Shortly thereafter, Read cut roughly 1,400 jobs to reposition the company. However, the new strategy for AMD has not been clearly articulated and will remain under wraps until their financial analyst day. Incidentally, the financial analyst day was rescheduled from November, 2011 to February, 2012. This gives the company and its new management a bit more time to formulate and begin executing on their strategy.

In the mean time, speculation concerning AMD’s plans has been rampant. Clearly mobile devices such as tablets and smart phones will be a key element as that part of the semiconductor market is growing rapidly. However, the company has also made comments indicating that it may focus less on competing with Intel in the future. In light of these comments, there has been speculation that AMD might shift its mobile strategy to work with ARM for low-power devices. However, any such speculation is baseless and irrational. AMD is nearly certain to continue using x86 and is unlikely to adopt ARM for both technical and business reasons.

AMD Today

Taking a step back, it is important to understand AMD’s current position, before speculating about a future strategy. Fundamentally, AMD is a fabless company that designs x86 microprocessors, GPUs and SoCs for the PC and server market.

AMD’s ability to design and more importantly validate and support x86 microprocessors is extremely unique. There are only 3 such companies: Intel, AMD and VIA. The latter of which is a marginal player in terms of revenue and shipments. For all intents and purposes, AMD is the only company outside of Intel that can do x86. The problem is that Intel has a huge manufacturing advantage and more resources. While AMD’s CPUs are good, Intel is simply better.

AMD graphics solutions are extremely competitive with Nvidia for the consumer market. Their actual hardware design is slightly better than Nvidia, and their software and support slightly weaker; on the whole the two are very evenly matched for consumers. For workstations, Nvidia’s software ecosystem is a significant advantage, but also an opportunity for growth into a high margin market. AMD’s strength in graphics is a saving grace for their SoCs, and helps to mitigate Intel’s advantages for consumer PCs.

The ARM World

ARM is a rather different animal from x86. Nearly every company has some sort of license or another. For some, this means minimal effort and just getting a reference design from ARM. Others have custom ARM-compatible designs that are tuned for specific markets.

In terms of mobile devices, the largest players are Apple, Nvidia, Mediatek, Qualcomm, Samsung and TI. Apple is the vertically integrated darling of the industry. Nvidia is a new entrant, hoping for their graphics to differentiate their SoCs. Mediatek is the low-cost and high volume threat from below. Qualcomm is the leader in cellular networks and baseband, particularly 4G. Samsung is partially vertically integrated, with their own fabs and system designs. Rounding out the pack is TI, which has deep analog and DSP expertise and is a very solid vendor.

From a technical perspective, adopting ARM for SoCs makes little sense. It utterly eliminates one of AMD’s core competencies, namely their x86 expertise. Without that unique capability, the company is left with good CPU design and superb graphics, albeit both are targeted at the PC. In some sense, there would be little distinction between AMD and a new-comer like Nvidia. Moreover, a great deal of AMD’s technical expertise is closely tied to dealing with the idiosyncrasies of x86. A deep understanding of optimizing variable length, memory-register instructions and x86’s ordering model is not all that useful in the context of ARM.

More importantly, trying to compete with the ARM ecosystem is just bad economics. As a general rule of thumb, less competition means higher profits and AMD is just barely profitable today. Even if AMD functions as a second source for Intel, that is a vastly more lucrative position than trying to compete to be the fifth source for Qualcomm.


Pages:   1 2   Next »

Discuss (338 comments)