Advanced Micro Devices (NASDAQ:AMD) has, for some time now, been chasing 64-bit ARM chips for use in servers and data centers, and as a possible replacement for Intel's (NASDAQ:INTC) x86 server chips. AMD plans to launch ARM chips for servers in the fourth quarter of the current fiscal year. Intel controls 97% of the x86 server chips, while AMD controls just 3% of this market. How hard will it be for AMD to challenge Intel's dominance in server chips?
A history of AMD vs. Intel server chip battles
AMD has been trying to compete with Intel's x86 standard for many years. The company enjoyed considerable success about a decade ago, when it introduced the Opteron family in 2003, initially based on the ''Hammer'' architecture. The company's server chip market share peaked at 25% in 2006.Opteron chips enabled 64-bit processing a few years before Intel finally launched its own 64-bit chips.
Intel's 64-bit chips have since become the standard for server processors. As it matured, the server market gradually moved to multi-core chips, and AMD again had the upper hand when it bet the farm on multi-core chips that used the ''Bulldozer'' architecture. AMD began shipping these chips in 2011. The chips paired two integer cores with a single floating-point core, as well as various other shared components.
But, again, Intel responded quickly by building chips with more full cores that had both integer and floating point capabilities. As a result, Intel's multi-core chips, most notably its Xeon E5 line, have dominated the market. Intel sold 92% of all server processors worldwide in 2013, and accounted for 75% of server chip revenue. High-end server chips such as IBM's Power and Oracle and Fujitsu's Sparc, accounted for just 1% of server chips shipped, but 20% of overall revenue due to their much higher average selling price, or ASP.
Challenges of shifting to a different architecture
Any movement toward a new architecture would require companies to modify their software stacks to be compatible with the new architecture. The new chips would also have to first gain critical mass to make the switch worthwhile for enterprise customers. Companies that write their own software, such as Google and Facebook, are more likely to become early adopters of new technology. But for other companies, this might be a major hurdle, which would considerably slow adoption of ARM server chips.
There are also major questions regarding the actual potential of low-power servers and micro-servers, which are the main target of ARM server chips. AMD estimates that the current market for microserver chips is 4%-6%, with the potential to eventually grow to 10%-20%. Gartner is less optimistic and sees the market growing to just 5% of all server chip revenue by 2017.
A recent Tech Pro Research report revealed that only 5% of organizations are currently deploying microservers.
The research also found that organizations are not deploying microservers to replace traditional servers, but are using them to handle static elements in high-traffic networks. In other words, microservers are not likely to replace their bigger traditional counterparts.
But, Intel is not taking any chances and has diversified its server line to cover microservers as well. The mainstream category is now called Xeon E5 and is aimed at both single and dual socket servers. The company also has Xeon E7, aimed at multi-thread servers. Then there's the 64-bit Atom C2000 line, commonly known as Avoton, which is available in eight cores.
AMD can still win
But, this does not mean AMD won't benefit from its 64-bit ARM server chips. Although the company's goal to take 25% of the overall server chip market by 2017 looks overambitious, investors should remember that the company is starting from near the bottom with just 3% of the x86 server chip market. Moreover, according to the Tech Pro study, microservers are a niche market, which will allow AMD to grow its 64-bit ARM server chip business without competing directly with Intel.
The server chip market is currently worth about $12 billion. Assuming the microserver chip market grows according to Gartner's predictions, it will be worth nearly $600 million by 2017 (5% of $12 billion). If AMD takes half of this market, that would translate to $300 million in new revenue, or 5.7% of the company's 2013 revenue.
Considering a more bullish outlook, AMD achieved a peak market share of 25% in server chips in 2006. If the company manages to hit that level again by 2017, it could be looking at an extra $3 billion-$4 billion in revenue, or 56%-75% revenue growth.
AMD might see considerable revenue growth from its 64-bit ARM server chips, without causing a dent in Intel's market share, as the new chips are likely to serve a niche market. Many current analyst estimates might be off by a big margin, and investors are advised to keep a close eye on events in the space to be ready to capitalize on any pleasant surprises that might arise in the near future.