With driverless cars on the verge of going mainstream, Google now has its sights set on another goal: building a rival to Intel’s processors. In April, Google, the principal division of Mountain View, Calif.-based Alphabet Inc., announced it has software that would enable it to use non-x86, and consequently non-Intel, processors in data centers, and that it was going to give away data center designs so that others could do the same.
Google is taking this course of action for several reasons, says Paul Teich, principal analyst with Tirias Research. “Google is looking to both spread its supplier risk and move the industry state of the art forward for compute.” To explain this, Teich cites history: For 20 years since the early 1990s, rivalry between Intel and AMD has driven processor improvements faster and further than at any time before or after, he says.
Mark Hung, research vice president at Gartner Inc., notes that Google’s x86 alternative program may lead to chips only for internal use, following the strategy the search giant used in the past when it designed its own network switches. If Google becomes a credible threat with processors that get high praise for performance, features, power consumption, and cost—even if used internally—it could prod Intel to reduce costs. “How much better can you be, by doing internal development, in squeezing a little more out of your existing vendor?” Hung asks.
No matter what Google does or how the project plays out, neither analyst sees this processor push as having much impact on system builders for several years. One reason is that Google has the resources to not only create the chips, but to do so without a supporting ecosystem. Having that ecosystem in place is an important consideration for everyone else. “It's not just a processor price and performance issue,” says Teich. “The rest of the infrastructure has to compete with Intel infrastructure that is already shipping in high volume—motherboards, memory buffer chips, commodity DIMM server types, and so on.”
Teich also notes that a lack of cost-competitive infrastructure has hurt AMD’s Opteron server processor program for much of the last decade, resulting in shipments below a critical threshold. The situation may change with the upcoming Zen architecture, expected to hit production volume in 2017. For that reason, system builders should keep a close eye on the chipset’s progress, Teich says.
From the Trenches
System builder Joshua Liberman, president of Albuquerque, N.M.-based Net Sciences Inc., says that short term Google’s actions will have little or no impact on his business. Mid- to long-term, though, could be a different—and beneficial—matter.
Liberman says the Windows-Intel duopoly dominates the market, with AMD only in the bargain bin. There are Linux systems available, but these are not viable mass-market alternatives. Google’s entry could change that equation over the next two to four years. “If anybody really breaks that [duopoly] open and it becomes something that looks vaguely like competition, that would be a good thing,” Liberman says.
Other system builders don’t see this push into non-Intel processors as ever likely to have an impact on their businesses. Some, in fact, see this move as a possible step backward that does not take advantage of what Google does best. “The future is not hardware, but software. And Google rocks at software,” says John Kistler, owner of J&B Technologies Ltd., of Maryland Heights, Mo.
Kistler thinks the play may eventually involve not only servers but tablets or video processors. However, he doesn’t see any of these possible directions as having a significant effect on his company’s business.
Given the uncertainty surrounding Google’s intentions and plans, and the current processor landscape, what should system builders do? Gartner’s Hung advises, “For at least the next three to five years, Intel is still the best bet.”