Is the Wintel alliance falling apart?

FOR OVER A DECADE the alliance between Microsoft and Intel (Wintel) has been one of the great constants in an ever-changing computer world. The two companies have worked hand-in-glove for years and between the two of them have defined and divided the home/office computer industry. Consumers have benefited from this alliance in some ways -- even back in the Windows 95 era you never heard about a Windows compatibility problem with Pentium processors -- but competitors have long claimed that the alliance between the two gives them an unfair business advantage.

The Wintel alliance has continued even as both Microsoft and Intel have fought for increased market share in the mid-range to high-end server market, but recent events in the IT world may drive a serious wedge in the long-running alliance between the two companies. For the first time, the long-term interests of Redmond and Santa Clara may be diverging, as both pursue the profitable revenue streams that are increasingly hard to find in a depressed computer industry.

Intel: fighting the forces of commoditisation
The problem on Intel's side of the equation is the falling price of CPUs. Intel's highest-end CPU today, the Pentium 4 3.06 GHz, sells for around $600. That's not a bad price, but it pales when compared to the $1500+ a Pentium II 266 sold for when that CPU was introduced. One might suspect the same recession that's hit the entire industry has depressed the CPU market as well, but even looking back to 2000 and the launch of the 1 GHz P3, its $1000 price tag was roundly decried as over-expensive. The cost of desktop CPUs, in real terms, has fallen dramatically in the past six years, and Intel's margins on these parts have dipped with them. Two factors have combined to keep Intel's overall margins high even in the face of increasing desktop commoditisation, however: the company's highly profitable line of server-based and higher margin Xeon processors, and the computer boom from 1997 to 2001 that greatly boosted the total number of chips Intel sold per year.

Intel cannot depend on either of these factors for long-term profitability. Just as competition and market trends dramatically reduced the price of home computers over the last six years, server costs have been dropping steadily as well. While Xeons still command higher margins than their P4 brethren, the difference between the two has dropped significantly since Intel first introduced the line. With prices on its Xeon DP (dual processors) dropping, Intel has locked Xeon DPs for dual-process operation only and introduced a new line of (expensive) Xeon MP processors for systems requiring more than two CPUs. While this will keep Intel's server margins buoyed in the short-run, this type of (re)action can only be temporary. There's little evidence that higher-end 32-bit server prices are immune to the same downward pressure that has dropped computer prices for the last twenty years.

Intel can't depend on the Itanium's extremely high margins for any kind of immediate fix. While Itanium sales increased 700% in 2002, that marks an increase from 500 to 3500 units sold. At this rate, the number of Itanic CPU's sold should outpace the number of copies of Daikatana sold in approximately two years. Considering DK ranks as perhaps the worst first-person-shooter ever, Intel clearly has a real goal in front of it. AMD's imminent entry into the server market may also complicate matters for Chipzilla. While the server market is far less competitive on the basis of price in general, tight economic times coupled with inexorably aging hardware may have more businesses buying on price than would typically be the case. It's an established economic principle that increased competition results in lower prices, but Intel will be more reluctant to start a price war in its Xeon DP/MP range than they would in the P4 market. Keep in mind, all this takes as given Intel's repeated statement that the Opteron is only competitive in the Xeon market. Should AMD prove successful at targeting high-end Opteron against Itanic, Intel could see its next huge source of revenue throttled before it even gets off the ground. I don't know what flushing $4.5 billion down the toilet sounds like, but I suspect it doesn't sound very tuneful.

Intel's strategy is to delay the commoditisation of its CPU's for as long as possible and, when that begins to threaten profits, invent a new product and price it higher. Itanium is a perfect example: According to Pricewatch, a 2.8 GHz Xeon sells for only $492. The cheapest Itanium listed clocks in at a "mere" $2758.

So, if you're Chipzilla looking to protect your own profit margins in the 32-bit server industry, yet want to simultaneously help lower the total cost of systems based on your CPUs, what do you do? You target other parts of the system which are significantly expensive, which, in this case, just happens to be Microsoft and its server line of Windows products. The "Microsoft tax" may be a smaller percentage of TCO on server systems than on desktops, but the amount dollar-per-dollar is far higher. With Linux gaining such popularity in the server space, what makes more sense for Intel—cutting its own margins to lower cost, or encouraging vendors and resellers to jump on the Linux bandwagon (while providing development resources to do so)? As the number of Intel-equipped servers running Linux continues to grow, you can bet Intel's outward support for the operating system will grow as well, whether Microsoft likes it or not.

Microsoft: stretching for sales
It's an open secret these days that Microsoft's old, upgrade-oriented business model isn't going to work much longer. The company is increasingly perceived as an established, "old" company, rather than the young, high-growth operation it was during the 1990s. Changing business models, however, is a lengthy process and prone to frequent mis-steps, as Redmond found out with its aborted, and severely criticised, original .NET My Services initiative.

As MS continues to shift into a more subscription-oriented business model it hopes will serve (fleece) customers more effectively, it still retains a need for upgrade-driven income in order to keep their margins padded and investors happy. Every sales division of Microsoft save Windows and Office hemorrhages money. This may partly explain why Redmond is developing a 64-bit version of Windows to debut with AMD's Athlon 64 CPU. Interest and demand for the Athlon 64 are high, and the CPU's 64-bit capabilities give Microsoft an immediate, concrete, and tangible marketing idea to shove down Joe Consumer's throat. 64 bits equals better, right?

Although all MS operating systems have worked on AMD's processors, there's little evidence the company has ever bothered to optimise for Chimpzilla's platform. If Microsoft believes it can use Athlon 64 availability to boost its own revenue, however, there's a real incentive for the company to heavily optimise Windows for a 64-bit operating environment and provide a software package that, in fact, has real benefits on 64-bit over 32. AMD, meanwhile, is hardly going to complain.

AMD's x86-64 technology, in fact, may present Microsoft with a unique, if perhaps short-term, revenue opportunity. It's not unreasonable to think that consumers making the jump to a 64-bit CPU would be interested in buying a 64-bit operating system to run on it. And if you've got a 64-bit operating system, why not buy a 64-bit suite of Office software? Of course, the attractiveness of such an option is going to depend on what Microsoft can do to make the upgrade worthwhile, but there's likely more attractiveness in the idea than the Win2K to WinXP upgrade.

Intel, however, is unlikely to be thrilled with the idea that Microsoft would develop an AMD-centric product, but Santa Clara's on record saying the need for desktop 64-bit processing may be as much as ten years away. Of course, Yamhill might be able to take advantage of a new 64-bit version of Windows, but that CPU doesn't exist, right? CPUs that don't exist are CPUs whose features Redmond can't use to sell software.

Of course, Microsoft in turn isn't going to be pleased if Intel chooses to use Linux as a method of cutting costs on its own server systems. How much does this matter? Again, this depends on how Linux growth continues for the next few years. The crucial difference between Intel and Microsoft in the server space is this: Every server sold with an Intel CPU is a victory for Intel against AMD in the 32-bit market and against Sun, IBM, and AMD in 64-bit areas. For Microsoft, however, the presence of an x86 CPU is an irrelevant necessity. Necessary, of course, because Windows primarily runs on x86 technology, but irrelevant because the presence of a compatible CPU does nothing to guarantee the system in question is using Microsoft products.

Will any of these events shake the Wintel alliance? Hardly. Microsoft and Intel-based technology are joined at the hip and, until Linux evolves into an OS fully capable of replacing its commercial rival on the desktop and server, always will be. If the best interests of the two companies continue to diverge, however, it won't be long before serious cracks begin appearing in that once-impregnable façade. That might not be such a bad thing. Intel's interest in Linux development could provide that community with many useful tools, and Microsoft's work with AMD, and the marketing efforts that go along with such involvement, could give Sunnyvale a very useful kick. µ