At the moment, Advanced Micro Devices' (AMD -1.06%) share of the server chip market basically rounds down to zero. Rival Intel (INTC 0.05%) is responsible for 99% of server chips shipped each year, a near-monopoly that has led that company to produce incredible profits in its data center business. AMD has so far completely missed out on the demand for server chips created by cloud-computing companies building out massive hyperscale data centers.
This situation could change in 2017, when AMD launches new server chips based on its upcoming Zen microarchitecture. The company is promising major performance improvements, and the move to a 14-nanometer process will help close the manufacturing gap with Intel. Things could still go wrong: Zen server chips could be delayed, or could fail to live up to the performance promises AMD has made. But it's hard to imagine AMD's position in the server chip market getting any worse.
Besting Intel won't be easy, but AMD has done it before.
The rise of Opteron
In 2002, AMD was in a position similar to where it finds itself today. The company was unprofitable, forced to slash costs, and its share of the x86 server chip market was essentially zero. AMD was less diversified at that time, as it didn't acquire graphics chip company ATI until 2006, and the company still manufactured its own chips. But its position in the server market was nearly identical.
Four years later, AMD held a roughly 26% unit share of the x86 server chip market. Intel, once viewed as an unstoppable juggernaut, was unable to fend off AMD. Opteron, a line of server chips launched by AMD in 2003, proved to be a massive success, propelling the company back to prosperity and creating a major headache for Intel.
Opteron was a 64-bit processor that could run 32-bit applications with no performance penalties, something that Intel's Xeon processors could not do at the time. Intel's high-end Itanium processors were 64-bit, but legacy 32-bit applications ran slow, and the product line was completely abandoned a few years ago after continually failing to gain much traction. Intel launched a 64-bit Xeon processor in 2004, but it would take until late 2006, when the first Xeon chip based on its Core architecture was released, before the company reclaimed the performance crown and began to eat away at AMD's market share.
Can AMD do it again?
Despite the similarities between 2002 and today, Zen is unlikely to repeat Opteron's massive, albeit short-lived, success. That's not to say that AMD won't win market share from Intel. But expecting the company to claim a quarter of the server chip market in a few short years is probably overly optimistic.
In 2002, it took a good product from AMD and a major misstep from Intel to fuel the former's market share gains. Intel's focus was split between Xeon and Itanium, and it allowed AMD to gain technical and performance advantages that took years for Intel to correct. In contrast, Zen is AMD's attempt to catch up to Intel in terms of performance, and it's unlikely that AMD's new chips will be clearly superior.
Server chips are also a far more important part of Intel's business today. During 2015, Intel's data center segment accounted for 29% of total revenue and 56% of operating profit. Intel didn't break out server chip sales back in 2002, but in 2009 the data center group accounted for just 18% of revenue and 40% of operating profit.
The first pillar of Intel's new post-PC strategy is the cloud, and with the data center group being its main growth engine going forward, the odds of Intel making another major misstep that opens the door for AMD is slim. If AMD's Zen server chips produce the kind of performance gains that AMD has promised, the company can certainly win back some market share from Intel. But a repeat of 2003, when Opteron upended the server chip market, is unlikely.