Over the past few years, as ARM CPUs have become more powerful, we’ve seen an increasing number of companies express interest in an ARM server solution. Early providers,like Calxeda, planned to launch relatively modest microserver solutions with larger chips and more capable cores coming at a later date. Other companies, like AMD and Applied Micro, plan to enter the market with fairly aggressive ramps in the microserver space. Now, Qualcomm has flung its own hat into the ring — and it’s got the potential to dwarf all competitors.
According to Steven Mollenkopf, Qualcomm’s CEO, the company is already engaged with multiple customers, including Facebook. Unlike the other early competitors, Qualcomm doesn’t necessarily need to lead with a microserver design or take a page from AMD’s book and introduce a standard ARM core while it works to bring its own custom architecture to market. Qualcomm doesn’t just have experience building ARM cores, its previous Krait design was a wide core that outperformed the Cortex-A9 against which it competed, clock for clock.
Qualcomm has been working on its own 64-bit chip for several years, and has relied on ARM’s standard Cortex-A15 and Cortex-A53 for its interim products, but the company didn’t fire its chip developers or retrench around the idea of just using ARM IP.
Qualcomm could push ARM deep into the datacenter
With the collapse of Calxeda, companies like AMD and Applied Micro appear to be the best near-term challengers to Intel’s x86 datacenter business, and both should have hardware in the field before Qualcomm has shipping products to launch.
What does Qualcomm have that these other companies don’t? To be blunt: money. AMD and Applied Micro are taking similar strategies, with an emphasis on particular markets andunique features where the ARM architecture is best positioned. There’s nothing wrong with that — it is, in fact, the smart strategy to take, particularly for companies with limited cash flow.
Qualcomm, however, has the ability to challenge Intel on its own turf. That’s not to say that this is an automatic win — if large companies could be assured victory simply by spending money, Windows would already own the Android ecosystem and Intel’s Atom would powerall but the cheapest tablets.
Intel’s datacenter business is in no near-term trouble. The company’s Xeon processors still utterly dominate the market and Intel has taken steps to address the microserver business by bringing its lower-power Atom cores to market. Broadwell, which already debuted as Core M, undoubtedly has a future role to play in low-power, high-density product divisions. Anyone who thinks that an ARM vendor, even an ARM vendor the size of Qualcomm, will simply walk in and pick market share off Intel’s shelf like grocery shopping is kidding themselves.
That said, there’s still a genuine threat here. Intel hasn’t faced serious competition in the enterprise server business for six or seven years. It hasn’t faced serious competition from a company of comparable size and strength to itself for well over a decade.
Qualcomm’s server business could kickstart the long-awaited ARM-x86 wars
One of the striking trends in software and hardware development over the past few years is something that didn’t happen. Despite the confident predictions of a great many people, the ARM and Windows ecosystems did not merge on either a software or a hardware level. If you own a Windows device, it’s virtually guaranteed to run x86. If you have an Android tablet, it may use an Intel Atom processor thanks to Intel’s contra-revenue programs, but it still uses a suite of tools that’s distinct from traditional Windows software. Applications likeOffice for iPad are the exception that proves the rule.
Qualcomm’s entry into the server business could finally bring the ARM and x86 ecosystems into direct conflict rather than simply skirmishing around the edges. Intel and AMD both historically built server cores that were essentially repurposed into cheaper consumer products. Qualcomm has no plans to launch into this segment in the near term, but if it can demonstrate a successful server business, building consumer products around the same chip families is a logical next step (and a way to improve yields and margins by using chips that don’t make the server cut).
For now, the company has announced just one partner, but Facebook surely isn’t the only company eyeing future Qualcomm servers. This announcement could fundamentally reshape the market..