Pages: 1 2
Over the last year, AMD as a company has been in flux. An almost entirely new management team is in place, which started with Rory Read’s appointment as CEO in late 2011. Over the last several months, Read brought in a number of new executives to the company including Mark Papermaster as CTO, Lisa Su as VP of Global Business Units and Rajan Naik as Chief Strategy Officer. Typically, the company holds a financial analyst meeting at the end of every calendar year. However, AMD delayed the analyst day for two months, to assemble a complete management team. Last week, AMD held the long-awaited financial analyst day, which was an opportunity to introduce the new management team to the financial community, articulate a new strategy and update the public product roadmaps. The main presentations were from Rory Read, Mark Papermaster and Lisa Su. There was also a financial updates from CFO Thomas Seifert. Seifert has been with the company since 2009 and served as interim CEO during 2011, so he is a known quantity and the first three presentations were generally more significant.
The overall impression of the new management team was fairly positive. While some of the presentations were a little opaque and confusing, the executives seem to clearly understand many of the critical industry trends and they collectively painted a picture of where AMD is heading. Perhaps the most likely explanation for the confusion is that all the new executives spent many years at IBM, which has a very specific culture, mindset and vocabulary that is different from the rest of the industry. The other good news is that Papermaster and Dr. Su have strong backgrounds in both semiconductors and markets beyond just the PC, which hopefully will complement Read’s experience at one of the more innovative PC OEMs (Lenovo). However, these are merely first impressions and the real metrics are the results that the new executive team can deliver over the coming months and years.
Few Strategic Changes
Overall, AMD’s strategy does not seem to have materially shifted from when Dirk Meyer was CEO. The company is also continuing to adapt to a fabless business model. The new management indicated that they would target leading, but not bleeding edge process technology. This seems to be a signal that they will be content with the roadmap set by TSMC and Global Foundries, which realistically lags Intel’s relentless pace by roughly 18 months. In context, this means that AMD will be at the head of the pack for fabless companies, taking advantage of the internal semiconductor expertise. The company is continuing to shift away from custom chip design towards a System-on-Chip (SoC) philosophy, where key IP blocks such as AMD’s CPUs, GPUs and memory controllers are combined together to create products targeted at different markets. The crux of the strategy is to focus on products where AMD’s IP is differentiated and can provide compelling value to customers, despite the manufacturing disadvantage.
One of the most consistent themes was an acknowledgement that AMD had simply failed to execute and deliver products to the market, based on the timelines and performance expectations set by roadmaps. This has cost AMD credibility, and had a negative impact, particularly on the server side of the business. It was clear that consistent execution is a major goal for the new management, and some of the roadmap changes (for servers) seem to reflect a more conservative and deliverable roadmap. One area that was discussed was using more of synthesizable logic, which is simpler and easier than semi-custom or custom design, but comes at the cost of area and power overhead. This complements the overall SoC design philosophy and should help improve time-to-market. However, it is unclear how much performance is being left on the table, especially since AMD’s main competitor can customize both the manufacturing and design flows.
One change in strategy is that the management signaled that they are more open to collaborating with third party IP vendors. Of course, the most controversial third party IP is ARM’s microprocessor cores. It is still exceptionally unlikely that AMD would supplant x86 and adopt ARM, for reasons which we discussed in our analysis of AMD’s mobile strategy. However, it seems like AMD is at least keeping the door open for future limited uses of ARM’s IP, perhaps in a hybrid system. AMD’s new executives also clearly indicated that they would be happy to take a page from IBM, and work on bespoke projects for customers, using a variety of internal and external IP blocks. This is probably most relevant for consoles, where AMD is fairly strong, but perhaps this is viable in other markets.
The openness to third party IP was particularly emphasized in a presentation on the Heterogeneous (formerly Fusion) System Architecture from Phil Rogers. HSA is an ISA-agnostic interface to seamlessly use CPUs and GPUs, which is software compatible across different hardware platforms. AMD plans to eventually offer their work as a free standard and would clearly like ARM’s support for HSA. Longer term, HSA might enable a hybrid SoC for AMD, perhaps using ARM cores for extremely low power. Historically, AMD has not been very successful with creating standards, because the company lacks a pervasive presence and influence in the industry, but perhaps support from the ARM ecosystem will help them overcome this hurdle.
The most puzzling aspect of AMD’s strategy is what it says about the board of directors. Reportedly, Dirk Meyer left AMD based on disagreements over the company’s mobile strategy and outlook. However, the new management presented a relatively minor overall shift that seems to generally continue down the existing path. This implies several possibilities. One alternative is that the existing strategy was evaluated by the new management and determined to be a good plan. This would represent a major mistake by the board of directors, as a one year leadership void has not helped AMD. Since there was little discussion beyond 2013, another possibility is that the new management team is making longer term adjustments that they are simply not comfortable revealing (or have yet to be finalized). Only time will exactly what transpired.
Discuss (38 comments)