wintel

Wintel: Le Divorce Part II

 

At CES 2011, Ballmer told the world Windows would “fork”, that it would also run on lower power ARM chips for mobile devices. This was seen as a momentous breach in the long-standing Wintel duopoly. Two years later, the ARM tooth of the fork looks short and dull.

This is what I wrote almost two years ago:

After years of monogamy with the x86 architecture, Windows will soon run on ARM processors.

As in any divorce, Microsoft and Intel point fingers at one another. Intel complains about Microsoft’s failure to make a real tablet OS. They say MS has tried to shoehorn “Windows Everywhere” onto a device that has an incompatible user interface, power management, and connectivity requirements while the competition has created device-focused software platforms.

Microsoft rebuts: It’s Intel’s fault. Windows CE works perfectly well on ARM-based devices, as do Windows Mobile and now Windows Phone 7. Intel keeps telling us they’re “on track”, that they’ll eventually shrink x86 processors to the point where the power dissipation will be compatible with smartphones and tablets. But…when?

Today, a version of Windows (RT) does indeed run on an ARM processor, on Microsoft’s Surface tablet-PC hybrid. Has Microsoft finally served Intel with divorce papers?

Not so fast. The market’s reaction to Redmond’s ambitious Surface design has fallen far short of the heights envisioned in the company’s enthusiastic launch: Surface machines aren’t flying off Microsoft Store shelves. Ballmer himself admits sales are “modest” (and then quickly backpedals); Digitimes, admittedly not always reliable, quotes suppliers who say that Surface orders have been cut by half; anecdotally, but amusingly, field research by Piper Jaffray’s Gene Munster (who can be a bit excitable) shows zero Surfaces sold during a two hour period at the Mall of America on Black Friday, while iPads were selling at a rate of 11-an-hour.

Traditional PC OEMs aren’t enthusiastic either. Todd Bradley, head of HP’s Personal Systems Group, is unimpressed:

“It tends to be slow and a little kludgey as you use it .…”

Acer exec Linxian Lang warns:

“Redmond will have to eat ‘hard rice’ with Surface…it should stick to its more readily-chewed software diet.”

To be sure, there are happy Surface users, such as Steve Sinofsky, the former Windows Division President, as captured in lukew’s Instagram picture:

(An aside: I went back to Sinofsky’s 8,000 words blog post that lovingly describes the process of developing “WOA” — Windows on ARM. At the time, WOA was presented as part of the Windows 8 universe. Later, Microsoft swapped the “8″ designation and chose to use “RT” instead. These naming decisions aren’t made lightly. Is there any wonder why WOA was moved out of the Windows 8 camp?)

It’s possible that the jury is still out… Surface sales could take off, Windows RT could be embraced by leading PC OEMs… but what are the odds? In addition to the tepid reception from customers and vendors alike, Microsoft must surmount the relentless market conquest of Android and iOS tablets whose numbers (210 million units) are expected to exceed laptop sales next year.

So, no… the Wintel Divorce isn’t happening. Intel’s x86 chips will remain the processors of choice to run Windows. Next month, we’ll have CES and its usual burst of announcements, both believable and dubious (remember when 2010 was declared the Year Of The Tablet PC?). We’ll have to sort the announcements that are merely that from those that will yield an actual device, but in the end I doubt we’ll see many new and really momentous Windows RT products out there.

Microsoft’s lackluster attempt at Post-PC infidelity doesn’t help Intel in its efforts to gain a foothold in the mobile world. Intel’s perennial efforts to break into the mobile market with lower power, lower cost x86 chips have, also perennially, failed. As a result, there is renewed speculation about a rapprochement between Intel and Apple, that the Santa Clara microprocessor giant could become an ardent (and high-volume) ARM SoC foundry.

As discussed here, some of this makes sense: Samsung is Apple’s biggest and most successful competitor in the smartphone/tablet space, spending billions more than anyone else in global marketing programs. At the same time, the South Korean company is Apple’s only supplier of ARM chips. Intel has the technology and manufacturing capacity to become an effective replacement for Samsung.

This wouldn’t be an easy decision for Intel: the volumes are high — as high as 415M ARM chips for 2013 according to one analyst — but the margins are low. And Intel doesn’t do low margins. Because of the Wintel duopoly, Intel’s x86 chips have always commanded a premium markup. Take Windows out of the picture and the margin disappears.

(As another aside, the 415,000 ARM chips number seems excessive. Assuming about 50 million iPhone 5s and 15 million iPads in the current quarter, and using the 4X rule of thumb for the following calendar year, we land somewhere between 250M and 300M ARM chips for Apple in 2013.)

Also, Intel would almost certainly not be Apple’s sole supplier of ARM chips. Yes, Apple needs to get out of its current and dangerous single source situation. But Tim Cook’s Supply Chain Management expertise will come into play to ensure that Apple doesn’t fall into a similar situation with Intel, that the company will secure at least a second source, such as the rumored TSMC.

The speculation by an RBC analyst that Intel will offer its services to build ARM chips for the iPhone on the condition Apple picks an x86 device for the iPad is nonsensical: Apple won’t fork iOS. Life is complicated enough with OS X on Intel and iOS on ARM.

Historically, a sizable fraction of Intel’s profits came from the following comparison. Take two microprocessor chips of equal “merit”: manufacturing cost, computing output, power dissipation… And add one difference: one runs Windows, the other doesn’t. Which one will get the highest profit margin?

In the ARM world and its flurry of customized chips and software platforms, the “runs Windows” advantage is no longer. ARM chips generate significantly lower margins than in the Intel-dominated world (its competitor AMD is ailing).

This leaves the chip giant facing a choice: It can have a meager meal at the tablet/smartphone fest, or not dine at all at the mobile table…while it watches its PC business decline.

In other news… Paul Otellini, Intel’s CEO, unexpectedly announced he’ll leave next May, a couple years ahead of the company’s mandatory 65-year retirement age. No undignified exit here. Intel’s Board pointedly stated they’ll be looking outside as well as inside for a successor, another unusual move in a company that so far stuck to successions orchestrated around carefully groomed execs. This could be seen as a sanction for Otellini missing the mobile wave and, much more important, a desire to bring new blood willing and able to look past the old x86 orthodoxy.

JLG@mondaynote.com

 

Wintel: Le Divorce

The eponymous flick is mildly interesting, but we’re gathered here today to examine the Wintel breakup. After years of monogamy with the x86 architecture, Windows will soon run on ARM processors.

As in any divorce, Microsoft and Intel point fingers at one another.

Intel complains about Microsoft’s failure to make a real tablet OS. They say MS has tried to shoehorn “Windows Everywhere” onto a device that has an incompatible user interface, power management, and connectivity requirements while the competition has created device-focused software platforms.

Microsoft rebuts: It’s Intel’s fault. Windows CE works perfectly well on ARM-based devices, as do Windows Mobile and now Windows Phone 7. Intel keeps telling us they’re “on track”, that they’ll eventually shrink x86 processors to the point where the power dissipation will be compatible with smartphones and tablets. But…when?

Intel: We constantly improve our geometry. Year after year we shrink the size of the basic semiconductor building block. Our newest fabs run 22nm processes!

MS: Yes, but why does ARM continue to win the battery game? Not to mention ARM’s flexible licensing which has created a thriving ecosystem of third-party processor extensions. Graphics, radios, networking functions…they all end up on the same low-power hardware, an entire system-on-a-chip (SOC) customized for each application, from navigation systems to tablets.

Intel: We’ve had this talk before. You’re a software company, why can’t you create a “mobilized” OS? At least we tried with Moblin.

MS: …and then you mated it with Nokia’s Maemo and spawned Meego. Ask any credentialed engineer what they think of corpocrats who condone such unnatural acts.

Let’s mediate.

Throughout the PC era, the Redmond company has cleaved to Intel. Intel insiders may dislike their second billing in the Wintel monicker and insist that the company is more than a junior partner, but the numbers tell a different story: Microsoft makes $62.5B in yearly revenue with a market cap of $242B; Intel? $44B/ $118B.

A thought experiment to illustrate Intel’s dependence on Windows: Take two PC processors, same size, computing output, power dissipation, and manufacturing cost. They differ in only one regard: Processor A doesn’t run Windows, processor I does.

Which chip will fetch the better market price?

The I chip, of course, the x86 processor.

Intel executives have chafed under the Microsoft yoke, they want to monetize their semiconductor design and manufacturing might in ways that don’t rely on Gates & Co. They even started their own venture portfolio, Intel Capital, to find and fund young companies that have the potential to open new sea lanes for the mother ship. They’ve gotten into all sorts of businesses, from toys to server farms, from modems to memory, and, lately, Wind River, Infineon and McAfee, none of which has done much to change Intel’s subordinate role in the PC market.

(This isn’t the case in the server sector where the dominant life form is x86 running Linux variants. Indeed, Intel’s strong Q4 2010 results show a 15% growth in the “data center group” while PC-related sales were flat.)

While the PC reigned, Wintel put on a happy face. But really personal computers–smartphones and tablets–broke up the marriage.

All smartphones run on ARM processors. (A handful of tablets use x86 hardware, but without much success.) The modern generation of mobile computing employs an array of operating systems, from Android to Bada, QNX, Meego, WebOS, iOS, but when you scratch down to the metal, you’ll find an ARM-based SOC. All of the subsystems that were cradled on a PC motherboard are now integrated on a single piece of silicon, ARM processor included.

Microsoft got tired of waiting for Intel to step up to the plate and, at the January 2011 CES in Las Vegas, Steve Ballmer announced that the next version of Windows would also run on ARM (transcript here).

But what is Microsoft’s CEO really saying?

I think Ballmer intends to bring the full market power of the franchise to tablets. Microsoft will step back and take the time–target 2012?–to make a “WinTablet” OS (WinTab? Wablet? Winblet? Register the domain names now) that includes a tablet version of Office. Third-party developers will follow.

The Wintel breakup causes profound and welcome changes in the industry, best summarized by Horace Dediu, in a recent Asymco blog post:

“At this year’s CES two unthinkable things happened:
• The abandonment of Windows exclusivity by practically all of Microsoft’s OEM customers.
• The abandonment of Intel exclusivity by Microsoft for the next generation of Windows.”

Intel professes to be unconcerned by the ARM flirtation, but below the calm surface the company’s executive must have their doubts. The low-end Atom business for netbooks hasn’t been doing well lately and will do even worse if tablets continue to eat into that category. Yes, there isn’t an enormous amount of money at stake in the low-end, but tablets and, more generally, ARM-based devices could seriously upset the x86 PC cart.

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.

In reality, however, it won’t be that easy.

PC OEMs have little choice: x86 processors are available from Intel and AMD, and, in a limited form, from TSMC. They can’t design an application-specific x86 device and send it to be manufactured in Taiwan or Korea.

Tablets and smartphone manufacturers, on the other hand, have the benefit of design flexibility, the choice of sources that come with the ARM ecosystem. Intel can’t take advantage of the quasi-monopoly it enjoys today in the PC world. Plus, the new ARM world means lower profits, so the company may decide against getting into the fray and, instead, focus on the servers. And even there ARM could become a threat: x86-based server farms run huge electricity bills and cause operators to look anew at ARM’s power-saving advantages for data centers.

None of these changes will happen overnight, but they won’t take long. A year ago, tablets were nonexistent. The PC market has been irreversibly changed. The George and Martha Wintel bickering makes for an interesting story, but…there are businesses to be run.

JLG@mondaynote.com