Some of you may wonder why x86 smartphones aren't a thing anymore, but many more probably didn't know they even existed in the first place. Starting in 2012, companies began launching smartphones using Intel's x86 Atom CPUs, a product line that the company touted as one of its most important. The smartphone market was lucrative enough on its own, but there were other use cases Atom would be great for. Intel only had to break Arm's stranglehold on phones and Intel CPUs would be anywhere.
As of 2018, x86 smartphones have gone the way of the dodo, and Atom managed to make its way onto our list of Intel's worst CPUs ever. In hindsight, it's not unbelievable that Intel could fumble something this badly. After all, it blundered in literally every segment of its business from 2017 until recently. But Intel's failure to get into smartphones was more complicated than bad technology or business mistakes.
The brief history of Atom and smartphones
From the mid-2000s, both Intel and AMD were focused on developing smaller, more efficient versions of their traditional silicon. AMD was content to just make smaller PCs and laptops with its Bobcat APUs, but Intel had big ideas with its competing Atom chips, first announced in 2008. It wasn't just going to be in-home theater PCs and tiny laptops; it was going to conquer the world. We'd see Atom in music players, televisions, GPS devices, handheld gaming consoles, and yes, smartphones. Intel was going to march right into Arm's most important stronghold and just take it.
Of course, Atom didn't immediately get into smartphones because Intel had to lay the groundwork. So 2008 and 2009 came and went with no x86 phones. Intel finally unveiled the Atom CPU it would use for phones in 2010, called Moorestown. Sure, it still had to contend with how phone manufacturers were used to making ARM chips, but Moorestown was so advanced and powerful that Intel was confident it would get three of the five top smartphone companies to make devices using Atom chips.
Lots of ink has been spilled over why Intel's phone strategy never got anywhere.
2010 came and went without any x86 smartphone announcements, but nobody expected any that soon. Then 2011 came and went with no smartphones or even announcements for a future one. The first Atom-based phone did come out in 2012, but it was only a reference design Intel and Google had made, not a high-performance device like everyone wanted. Around the same time though, Motorola, ZTE, and Lava became Intel's first partners in smartphones. Finally, we were seeing some momentum.
But for the next four years, nothing really happened — no big design wins, no stunningly fast Atom CPUs coming out. But in 2016, Intel made a big announcement: It was canceling its upcoming Atom SoCs for phones. And that was it. No SoCs meant no more x86 smartphones, even though Atom was still getting updates. Intel did create one last Atom SoC for a company it had made an agreement with, but that was it. The last Atom-powered smartphone came out in 2018, and it was bad.
That's where the very brief story of x86 smartphones ends. Lots of ink has been spilled over why Intel's phone strategy never got anywhere, but there were a few big reasons why Intel had to call it quits in 2016. Here's the autopsy report.
Atom had a hard time breaking into the software ecosystem of phones
The biggest and most obvious hurdle for Intel was software. Many people knew it would be a struggle the moment it launched in 2008 because Arm ruled the smartphone market. Now, it wasn't just about companies being used to working with Arm the company or using ARM chips in their phones. The bigger issue was that software made for ARM CPUs couldn't run on x86 chips.
Basically, every CPU makes use of an instruction-set architecture (or ISA), which defines what the CPU can fundamentally do and how it reads code (and I mean actual ones and zeroes and not a coding language like Python or C++). Arm had (and still has) a big incumbency advantage in phones because all the software was made for ARM chips, from operating systems like iOS and Android to the apps that ran on the operating systems.
Intel knew about the challenges in introducing a new ISA to a market that was accustomed to using a different one. Itanium, the company's first 64-bit CPUs, used the new IA-64 ISA rather than an upgraded version of x86 that was capable of 64-bit, which was ultimately a fatal mistake for Itanium. AMD's competing Opteron chips used the x86-64 ISA and took nearly 25% of the server market. Eventually, Intel had to throw in the towel and make its own x86-64 server chips, Xeon, and it also used x86-64 for all its other CPUs and has ever since.
Still, this was a thing Intel could see from a mile away, and with enough dedication to smartphones, it was something that could be overcome. Indeed, there were lots of smartphones that used Atom CPUs, such as Asus's Zenfone series, which was one of Intel's bigger wins. However, there were other complicating factors.
Intel didn't give Atom the resources it needed
Atom is remembered for being pretty slow, and it's not entirely unjustified. Although Atom chips weren't universally bad (one of the first x86 smartphones was actually pretty decent performance-wise), they couldn't measure up to ARM-based chips from companies like Qualcomm and Apple. This wasn't just a consequence of bad engineering on Intel's part, but also a lack of prioritization that disadvantaged Atom.
Process nodes are really, really important for smartphone chips. Upgrading from one process to the next improves not only density (meaning you can make smaller chips or cram more parts into the same space) but also improves performance and efficiency, which is especially important. Higher efficiency means better battery life and also better performance at the same power consumption. But Intel always let its desktop, laptop, and server CPUs get first dibs on its latest processes, with Atom getting upgraded about a year or two later every single generation. No wonder Atom wasn't all that fast.
Extremetech also put forth a theory that Intel didn't want to change its business model for Atom either. Writing in 2016 soon after Intel canceled its Atom smartphone chips, the publication said Intel "wasn't willing to risk upsetting the economic model that had transformed it into a titan of computing." Intel didn't want to get into making low-end, cheap processors for phones when it could make bigger margins in other markets. After it had lost billions upon billions of dollars half-trying, it gave up as soon as times looked tough for the company.
Ultimately, Intel got too big for its britches
Between the extreme difficulty in breaking into an established hardware-software ecosystem (especially considering Intel had firsthand experience with that already) and general negligence towards Atom, it's clear Intel overestimated itself when it came to smartphones. It thought that just because it was an industry titan that it could walk into the phone market and own it as it had with desktops, laptops, and servers.
That very same hubris is what led to Intel thinking it could just buy company after company for billions of dollars, aim for an absurdly high generation-on-generation gain with its 10nm node, and capture 30% of the entire silicon market, including CPUs, GPUs, and FPGAs. All of this blew up in Intel's face, just like it did with x86 phones, and although it was always going to be an uphill battle for x86 to survive in the smartphone market, Intel's recklessness was perhaps what doomed it to failure.