Apple shocked the world when it went to Intel.
As I write this, the new Macs with Intel Core Duo processors inside are available. I was about as shocked as anyone when Steve Jobs announced that he would discontinue Macintosh computers with the IBM PowerPC Architecture and instead develop new machines with Intel chips. It just didn’t make sense to me on several levels, and, while I could develop theories that made sense of the move, everywhere I proposed these theories on Mac blogs and boards, people debunked the theories. And because I work for IBM, specifically for a division that promotes partnerships, I worried that my bias in favor of IBM’s PowerPC Architecture would color my opinions.
With all this on my mind, I decided to do something I haven’t often done in this column: I consulted with an impartial expert–Mark Margevicius, Research VP for the GartnerGroup, specializing in the personal computing industry. We had a wide ranging phone interview in which I proposed a bunch of theories and he either debunked them or endorsed them. After checking my theories against the leading expert in the field, they make a lot more sense than the Mac bloggers give me credit for.
Before I get to the theories, here’s what didn’t make sense to me. First, it didn’t make sense to deny 15 years worth of talk, by Apple, that Intel chips weren’t good enough for the Mac. A key component of the Mac’s success has always been about differentiating from the vanilla PC, both in terms of hardware and software. Mac’s differentiation on hardware was the PowerPC. Look at what Apple said about the PowerPC G5 chip prior to announcing the switch to Intel.
I won’t reproduce what Apple says, except to briefly give the PowerPC advantages. The primary advantage is 64-bit addressing, which is a big deal for graphics-intensive applications. All other PC processors are 32-bit, limiting memory space to 4 gigabytes. In 64-bit processors, memory addressing is virtually unlimited. The PowerPC is also the first PC processor to provide symmetric multiprocessing, which allows several applications to work in parallel rather than one at a time. Some versions of the processor are dual-core, which typically increases performance in parallel environments by around 25 percent.
Contrary to Apple’s stated reason for switching, the PowerPC is also known for producing the least amount of heat of any PC processor. Consider this fact: IBM produces a line of BladeCenter servers based on the PowerPC 720 chip–the same chip that runs earlier versions of G5 Macs. The IBM BladeCenter chassis houses up to 14 blade servers–28 dual-core processors–in a box the size of a ’90s-style PC. Because the PowerPC is relatively cool, the whole box can be cooled by two fans.
Finally, there is a huge disincentive to switch chip architectures: You force independent software vendors and your installed base to migrate their applications to the new platform. Fortunately for Apple, most of the available applications compiled for the PowerPC platform are also available for the Intel platform: Quark, InDesign, Photoshop, Illustrator, Office, etc. Smaller partners will need to migrate, but my guess is Apple doesn’t care too much about them in the scheme of things.
But the installed base is a concern. What it means for the largest customers–publishers and design shops–is that they will need to either upgrade all their applications to take advantage of the new systems or they will need to run the old applications in an emulator. Any performance benefits from the new architecture are erased by the emulator, so there’s little incentive for Apple’s largest customer base to upgrade hardware from PowerPC to Intel. Given the cost of new machines plus new software in these high-end graphics businesses, the cost/benefit ratio for doing that cannot be overcome over the life of the product.
Now to the theories. The first theory is cost. As Margevicius confirmed, Intel gave Apple a sweetheart deal to switch to the Core Duo chip. Intel can afford to do this while IBM cannot because the vast majority of the market–Dell, HP, Lenovo, etc.–builds systems for the same chip line. So Intel can have thinner margins on its chips than IBM and make up the difference on volume. IBM’s BladeCenter JS20 and JS21 machines that feature the PowerPC chip are not produced in PC volumes at this time.
The risk for Apple, as I’ve said above, is to reduce its hardware differentiation. Apple owners love the fact that they “Think Different.” Margevicius likens them to BMW owners, who pay more for cars that are different from the norm. To extend the analogy, Apple’s move would be like BMW developing a partnership with Toyota to put Lexus engines in Beamers.
This risk is deferred by two factors: First, Apple can choose to pass some of the chip savings onto its customers. This will act as a promotional tool for its key differentiator: OS X. When more users can afford PCs with OS X on them, they will see first hand how vastly superior the operating system is from Windows (more on this later).
“It’s not as though the hardware differentiation is helping Apple gain market share now,” Margevicius says. “As long as Apple’s losing market share, it might as well save money in the process.”
Second, as the benchmarks show, the Core Duo is a very fast chip in its own right. I’m still a little dubious as to whether these benchmarks test memory addressing in a way that shows the advantages of a 64-bit architecture. But especially in the laptop market, where the PowerPC does not have a dual-core offering, the Core Duo shows up to 20 percent performance advantages over the G5 chip.
My main theory is that Apple wants to ultimately get out of the PC hardware business and focus on OS X. How does this work? Well, right now, Intel Apple computers do not allow users to create partitions for dual-boot mode, that is, you can’t have both Windows XP and OS X on the same machine. But future versions will allow this dual boot mode with Windows Vista. This will allow users to compare the OSes side by side and see how superior OS X is. That’s the promotion. Margevicius says this is likely ruled out by Apple’s agreement with Microsoft for Windows XP, and at least and not yet ruled out for Windows Vista.
“We don’t see those partnership agreements, but Microsoft can probably choose to discontinue Office for the Mac if Apple violates certain terms, such as not creating Intel-based systems that allow users to run Windows and OS X side by side,” he says.
It also makes sense that Apple’s agreements with Microsoft prevent Apple from licensing OS X to other PC manufactures in direct competition to Windows. If Apple really wanted to get out of the PC business, this would seem the most likely avenue. It would boost the use of OS X overnight and that would help Apple’s other businesses. Ultimately, OS X would be a Windows killer.
Though Apple is openly trying to thwart hackers from illegally attaining OS X and working to get it installed on non-Apple PCs, I think it’s inevitable and Margevicius agrees. Apple even wrote special code in OS X including a poem for hackers in anticipation of this possibility. When hackers get their hands on OS X, it could spread like Linux. In this way, Apple could be conforming to the letter of its agreements with Microsoft while secretly hoping that OS X spreads like a virus that kills Windows. Once OS X is out of the bag, Apple’s agreements with Microsoft can be reworked to allow for limited licensing.
Margevicius and I agreed that Jobs has short- and long-term plans that led to switching to Intel chips. In the short run, he can get bigger margins on the Mac by holding his costs down. Long term, he can get out of the PC business by allowing OS X to slip out into general use. Because margins on operating systems are better than on PCs, he can drastically change Apple’s business prospects in the process.
“Steve Jobs is a shrewd businessman,” Margevicius says. “He knows there’s no money in hardware. This is his play to finally get out of the PC hardware business.”
James Mathewson is editor at large of ComputerUser and lead editor for the IBM ISV Business Strategy and Enablement organization.