At Advanced Micro Devices' (AMD +13.85%) most recent Financial Analyst Day, executives from the company outlined their goal to become a significant player in the server processor market. The company is banking on new processors using AMD's redesigned processor core, code-named "Zen," to try and gain share in the market for x86 server processors.
These share gains, according to a presentation from AMD CEO Lisa Su, are expected to come in the 2017/2018 time frame.
While AMD is telling a good story, I am not convinced the company will have such an easy time gaining share.
AMD's track record is pretty poor
Back at AMD's 2010 Financial Analyst Day, the company showed the following server CPU roadmap:
Source: AMD Financial Analyst Day presentation
In 2011, the company did, in fact, launch the "Interlagos" and "Valencia" CPUs mentioned above, but "Terramar" and "Sepang" never showed up. Instead, AMD launched "Abu Dhabi" and "Seoul" in place of "Terramar" and "Sepang," respectively, featuring the same core counts as their predecessors, but using the company's follow-on to the "Bulldozer" CPU core known as "Piledriver."
Since then, though, AMD hasn't released new server processors based on the company's newer CPU cores (Steamroller and Excavator), even though at the company's 2012 Financial Analyst Day, the company suggested that such products were incoming:
Source: Advanced Micro Devices Financial Analyst Day 2012
What the company is promising today
Earlier this year, AMD showed the following server product "roadmap":
Source: Advanced Micro Devices Financial Analyst Day 2015
According to this, AMD is going to put out high-core count X86 server processors for more "traditional" server segments, high-performance ARM (ARMH +0.00%) based processors aimed at storage, networking, and embedded segments, and will also put out a "high-performance server APU," which will presumably combine AMD's new Zen processor cores with its next generation graphics.
At first glance, it seems like AMD is going to have a whole suite of competitive products that should allow it to capture meaningful market segment share across a large number of data-center applications. But I think AMD may, in this case, be overpromising and, if I'm right, will wind up substantially underdelivering.
What is the basis for these claims?
Take a look at the following chart illustrating AMD's research-and-development spend over the last 10 years:
Source: YCharts
The server roadmap that AMD outlined in 2010 was published when AMD's total corporate research-and-development spending was in the neighborhood of $1.4 billion. Looking back, it would appear AMD didn't have the resources to execute upon its prior roadmap.
Where were the Steamroller-based server chips? How about the Excavator-based ones?
A more recent example of AMD's failure to actually launch products in a timely fashion would be its ARM-based microserver part code-named "Seattle." This part integrates eight off-the-shelf ARM Cortex A57 cores in a proven 28-nanometer manufacturing process, and AMD says that it won't start seeing revenue from shipping systems until the second half of 2015.
Note that back in 2013, AMD was telling people Seattle would sample in early 2014 and that "volume shipments" would begin in the second half of 2014.
If AMD was a year late on what I believe should have been a fairly straightforward microserver processor, should investors really believe that AMD will be able to actually hit its 2016/2017 launch plans with a whole suite of these arguably (much) more complex server processors built on a relatively new manufacturing technology on what is a substantially reduced research-and-development budget?
Keep the optimism in check
I think that investors who buy AMD on the expectation the company will execute perfectly on the roadmaps outlined at the 2015 Financial Analyst Day will wind up being disappointed. Whether that disappointment ultimately comes as a result of products that aren't as competitive as implied, or from significant product delays, is hard to tell at this point.
