Intel’s Challenge: WinARM

by Jean-Louis Gassée

The Wintel alliance, Windows and Office running on Intel x86 processors, enjoyed its dominant position for more than three decades. Today, ARM processors have shuffled the deck, presenting serious challenges to Intel’s new CEO.

Two weeks ago, we looked at the history and culture Intel CEO Pat Gelsinger has inherited. After ascending to become Intel’s first CTO, Gelsinger left the company in 2009, and now returns to undo the damage that was wreaked by his three predecessors.

Last week, we outlined the depth, breadth, and ambition of the new CEO’s goals for his second Intel life. At first we were skeptical, but after a closer examination of Gelsinger’s unusual combination of intellectual, psychic, physical, and spiritual energy, we were drawn to an optimistic conclusion: the Intel 2.0 reboot is in capable hands

This week, we temper our optimism by considering the influence that Microsoft and other companies could have on Intel’s future.

Let’s begin by taking a look at Microsoft’s Surface laptop/tablet hybrids. The first Surface, which ran on an ARM version of Windows called Windows RT, came out in October 2012, a good eight years before Apple’s ARM-based, pardon, Apple Silicon-based Macs. This initial attempt didn’t work out: Windows RT was “deprecated”, left by the roadside. But the road continued.

In December 2016, Microsoft announced a new Surface that ran on a Qualcomm Snapdragon ARM processor, followed by updated ARM-based Surface Pro devices in 2019 and 2020. Why such persistence? Because Microsoft engineers, who had accumulated intimate knowledge of the x86 architecture, understood the DNA defects of Intel CPUs. Microsoft decided long ago that ARM had a structural advantage, a decision later buttressed by the increasingly “desktop-like” benchmark results of Apple’s own Axx processors.

So far, consumer adoption of the Snapdragon Surface devices has been lukewarm…but that could change. Qualcomm just bought NUVIA, a firm started by ex-Apple engineers who are focused on the design of ARM-based CPU chips. In a not-too-distant future, Microsoft could wield competitive ARM-based Surface machines, where competitive has two meanings: Not too far from the Mac’s performance and, as a result, a threat to x86 laptops from Dell, Lenovo, and so on. Not to be outdone, PC clone makers will move to ARM processors made by Qualcomm and others.

Gelsinger knows this very well, even if he hasn’t always been an ARM true believer. At a Microprocessor Forum in 2008, he sang the rote “Just You Wait” refrain to argue against ARM’s structural advantage:

“Over the last five years, the performance gap has been steadily diminishing. It is an unfounded myth that the gap between RISC and CISC, or between x86 and everyone else, is large. It’s not large today. Furthermore, it is getting smaller.”

This is no place to relitigate the RISC (Reduced Instruction Set, e.g. ARM) vs CISC (Complex Instruction Set, eg. X86) debate. We’ll just note that Gelsinger soon will have no choice but to announce Intel’s commitment to the ARM architecture. However, unlike in the good old x86 days, Intel won’t have a monopoly on chips that run ARM-ed Windows and Office. By supporting ARM CPUs, Microsoft has removed the x86’s advantage, thus forcing Intel to join the ARM ecosystem (one bad pun avoided).

Now that Wintel — and the x86’s inflated margins and sales volume — is on its way to obsolescence, Gelsinger is left with an obvious move: Make it up in volume.

The highest quantities are to be found in smartphones, a market that former Intel CEO Paul Otellini declined to bet on. (According to industry lore, Intel felt that Steve Jobs’ bid was too low, and Otellini, by his own admission, couldn’t see the volume. He was wrong by at least two orders of magnitude.)

However, for Intel to win any significant volume in the smartphone market, it will have to successfully update its manufacturing process to be competitive with TSMC and others. It will also have to make price concessions to win accounts away from existing suppliers.

Next, we have the PC and tablet markets with lower quantities but not necessarily higher prices. Consider the chips in the latest iPhones and the newest Mac laptops: They’re closely related in performance and, most likely, cost.

Gelsinger’s final challenge is in what Intel calls the Data Center space. The high-end x86 Xeon chips used to command both good volumes and very high margins, but a look at the last quarter of 2020 shows Data Center product sales are down 16% and operating income down 40%. The downward trend is hard to explain given that Work From Home helped create the hottest Cloud Data Centers market we’ve ever seen.

While its recent Data Center decisions haven’t been as embarrassing as the iPhone blunder, Intel has missed out on a few opportunities. The company passed when they had the chance to acquire NUVIA while that company still concentrated on making chips for cloud servers. Before that, Intel watched as then-President and cloud computing expert Renée James leave the company in 2017 to found Ampere, self-proclaimed creators of “the world’s first cloud native processor”, the ARM-based Altra. Even more embarrassing, a look at Ampere’s leadership team shows it’s almost exclusively composed of Intel alums totaling close to two hundred years’ experience at the Santa Clara company. One imagines that the group lost the ARM battle inside Intel, and then took their skills and energy into the Ampere startup. I’d love to hear Gelsinger discuss Ampere now…

The way I describe the future might sound like everything, such as ARM’s rise everywhere, happened suddenly. As we know, it was long in the making. At the January 2011 CES, Microsoft CEO Steve Ballmer made it clear that his company would no longer exclusively rely on Intel for the processors in its devices. A week later, I penned a Monday Note titled Wintel: Le Divorce in which I blithely wrote:

For Intel, getting back into the ARM business (they sold the previous one to Marvell in 2006) seems like a straight shot: A bit of paperwork, some money, a team of engineers and they’re in business. Intel could very well decide to follow Microsoft’s lead — once again — and make ARM processors for the new Windows + Office combo.

I was very wrong on timing, definitely misunderestimating three factors:

• Moving Windows and Office to ARM is easier said (by a columnist hiding behind his iMac) than done in the real world, especially while also keeping the x86 moving forward.
• I well underestimated the time it would take for ARM CPU makers (besides Apple) to come up with competitive desktop-class silicon.
• Above all, I didn’t realize Intel culture would stay in a state of denial for another decade.

In 2011, the year when the divorce began, Microsoft’s market capitalization was $218B; at $123B, Intel’s cap was about half that number.

Ten years later Microsoft’s market cap is very close to $2T, grown by a factor of 9. In the same amount of time, Intel barely doubled to $265B, revealing what investors think of its performance and future.

Restoring Intel to its former primacy will take all of Gelsinger’s multi-faceted skills.

— JLG@mondaynote.com