November 20, 2004

Don't buy Microsoft's $100 PC pitch

Author: David Sugar

I read with some amusement Microsoft CEO Steve Ballmer's recent call for $100 computers for the world's less affluent consumers. While many developing nations struggle with overwhelming poverty, and at least some struggle with just trying to feed their population, the cost of computing and technology are often not among the immediate concerns of such governments.

Some governments, such as Thailand's, have looked at making computing more affordable for their citizens. However, even with government support, the cost of computing hardware, even before any operating system is licensed or used, is often beyond the means of most of their citizens.

While standardization of PC hardware led to competition and commoditization that drove down hardware cost over the past decade, today this process has come to a halt, and it appears that the cost of new computers has reached a price floor. Ironically, this floor is a result of the same standardization process. How so?

Today, for a new PC to be compatible with the Microsoft Windows operating system, it must retain an accumulated set of odd architectural choices, starting with the Intel or Intel instruction set-compatible microprocessor, and ending with standardized BIOS that must be licensed. Part of their cost is in supporting these specifications. What if PC equipment were designed without such constraints, and were able to use components and architectures that best met the needs of business and home computing in the developing world?

Such systems would not need to use Intel processors, but instead could use alternate chips (perhaps ARM or PowerPC or something else) that run with far less power and cooling needs and still prove fast enough for ordinary business tasks. They would not need the same bulky and expensive power supplies. They could use less RAM. Furthermore, by eliminating any vestige of PC architecture compatibility, system vendors could produce machines with fewer specialized components that have licensing fees attached, and put them into smaller cases that take up less physical room and material costs.

The cost to provide such a machine may indeed be close to $100 when manufactured in quantity, compared to the minimum deliverable cost of $400 or so for an Intel- or AMD-based box capable of running the latest versions of Microsoft Windows and entertainment software.

Such PCs would of course not be compatible with Microsoft Windows, thus there would be no issue with their enabling copyright infringement of Microsoft products. However, GNU/Linux systems are very portable, and already have been ported from microcontrollers to mainframes. A complete and fully usable GNU/Linux system could run on such a machine. Indeed GNU/Linux systems have already been ported to and demonstrated on many of these non-Intel architectures. Over time, newer iterations of GNU/Linux systems tend to improve in computing efficiency and system requirements to run, rather than the trend of requiring ever more computing hardware just to perform the same tasks that one finds in proprietary solutions such as Microsoft Windows.

Indeed, the cost of computing can be made more affordable. This does not mean that computing hardware will ever be free. The simple reason is that, unlike software, whose marginal per-unit cost for the nth copy to manufacturer does approach zero, there remains a substantial and real fixed cost in parts, labor, and materials for every PC produced. Economy of scale may reduce these costs somewhat, but they never approach zero as they can for software.

When looking to provide affordable computing for the needs of developing nations, the best option to consider is free software. The economics of free software generally recognize and approach the ideals of a free market, where costs for software can approach zero.

However, some claim that free software delivered on standardized PCs for the developing world somehow promotes copyright infringement. Part of this is because of the economics of the pushcart vendor, who can and does sell proprietary licensed software illegally and unconstrained at its natural free market value. Hence, the cost difference between software that offers freedom and software that does not is not very great.

When Ballmer speaks of making PC hardware for $100 for developing nations, he is also speaking of selling users Microsoft Windows XP and Microsoft Office for an additional $900. True, Microsoft does offer "discounted" versions of its software, but it is provided as incompletely functional, such as Windows XP Starter Edition. This does nothing to reduce copyright infringement, but the company sells it on the principle that it is better to force hardware vendors to include a "broken" version of Microsoft Windows than to permit other operating systems to enter the market.

Contrary to Microsoft's apparent expectations, consumers in developing nations are no dumber than those in the rest of the world, and know a bad deal when they see it. Surely the most immediate result of making PCs a few hundred dollars cheaper will not be a sudden surge in the purchase of $900 worth of software, but rather more business for the pushcart vendor.

Even with a $100 PC, consumers in the developing world today must either do without computing altogether, or accept the necessity of being copyright infringers. However, there is another choice and a better way forward for universal computer access. We can actually achieve the goal of meeting the world's computing needs at $100 per machine without additional or hidden costs by removing Windows from consideration on such hardware, and using GNU/Linux instead. This will have the happy side effect of eliminating copyright infringement against Microsoft products.

Isn't it nice when free software can make everybody happy?


  • Free Software
Click Here!