Why Isn't Software Getting Any Cheaper?Why Isn't Software Getting Any Cheaper?

For many years now, two traits have defined the PC hardware industry: falling prices and skyrocketing performance. So why is software playing the heavy in this high-tech lovefest?In my last post, I quantified a trend most of us instinctively recognize: Every year, PC buyers pay less -- often a lot less -- for increasingly powerful and reliable hardware.

Has the software industry kept up in the price-performance race? If not, what are the implications for hardware makers and for buyers, especially businesses that regularly upgrade both their desktop hardware and software?

Windows Pricing: A Historical Context First, let's look at some numbers that illustrate long-term software price trends. I think it's fair to do this by looking at an obvious business software staple: Microsoft Windows. As I'll discuss in a moment, this is a perilous, but still necessary, exercise.

It takes some digging to find historical price data on Windows, especially for the earliest releases. According to this source, the first four Windows releases each retailed for around $100. Windows 3.0 -- the first version to offer any real usability improvements over DOS -- retailed for around $150 for a full version or $80 for an upgrade.

The next generation of Windows started off a bit cheaper: According to another source, a full version of Windows 95 sold for around $110. Windows 98, however, reversed the trend, retailing for over $200. Finally, Windows 2000 Professional upped the ante considerably, costing well over $300.

Price comparisons for subsequent Windows releases are more difficult to make; starting with Windows XP, Microsoft offered multiple versions of the same product at widely varying prices. It's safe to say, however, that the price for a Windows license certainly has not dropped very much in recent years. A retail copy of Windows 7, for example, will apparently cost just a bit less (perhaps by 10 percent) than a comparable copy of Windows Vista.

Fortunately, Microsoft itself offers some perspective on Windows pricing. In 1998, the company rebutted an Economist editorial which claimed that "the cost of a retail upgrade of Windows has more than doubled since 1990."

Instead, according to Microsoft, it had actually offered "relatively stable pricing during the 1990s" on its Windows upgrades. It's safe to assume this is a best-case scenario -- a fact that actually reinforces my point about software pricing.

General Software Price Trends Of course, any attempt to study software price trends -- even for the same product -- is a risky business. Microsoft now segments each Windows release into multiple versions (and multiple prices). OEM pricing is an entirely different, and particularly inscrutable, topic. And of course, the process of deciding what represents a software performance "enhancement" is far more subjective than it is for PC hardware.

Economists who study IT price trends often employ what is known as a "hedonic" approach, in which they attempt to quantify specific improvements in software quality over time. It's analogous to the "cost per MB" standard used to track PC memory and storage price trends (although less objective, for obvious reasons).

What is the bottom line here? According to a summary of several studies published in the mid 1990s, PC software generally dropped in price no more than a few percent each year. While the numbers backing these studies are now somewhat dated, I don't think there is any good reason to assume that the situation is much different today.

In other words, there is, indeed, a colossal disparity between software and hardware price trends. Since the 1980s, key hardware component prices dropped by three or four orders of magnitude. During the same period, software prices have remained nearly flat.

It was much easier to overlook this difference when new PCs cost hundreds or even thousands of dollars more than the software running on them. Today, however, the cost of a desktop system has dropped perilously close to the cost of a retail Windows license. That doesn't include the cost of other desktop applications, including business productivity tools (such as Microsoft Office) and non-optional "options" like anti-virus software.

What Accounts For The Gap? What accounts for the price differences separating hardware and software today? Certainly it isn't a question of production costs: Moore's Second Law, for example, states that while CPU transistor counts increase exponentially, so does the cost of the semiconductor labs required to fabricate them. Software "fabrication" costs, by comparison, are negligible.

Could R&D costs account for the disparity? It's at least a more tenable claim: Software development remains a stubbornly labor-intensive process (some economists still define it as a "craft industry"). Machine-generated code and automated QA tools play an important part in the development process, but neither has really supplanted the role of individual developers.

That's bad news for software buyers. It suggests that software developers may never realize the gigantic productivity gains that have revolutionized so many other segments of the IT industry. Also, while economies of scale play a clear role in driving down PC hardware prices (consider the premium some users pay for bleeding-edge hardware), the same process isn't nearly as apparent in the desktop software market.

Does Value Still Trump Price? Where does that leave IT consumers? For many years, they have tolerated a situation where desktop PC hardware was commodity-priced -- and desktop PC software most definitely was not. Buyers either did not notice or did not care, and vendors certainly were not going to raise the issue.

Now, however, buyers must decide: Does commercial software deliver enough value to justify this disparity? Will buyers tolerate a market in which commercial software vendors hold the line on pricing, even as hardware costs continue to plummet?

Open-source software, and most notably Linux, certainly offers one alternative approach to managing software costs. Over the past few years, Linux distributors and open-source software providers have demonstrated that it is possible to build a business model around no-cost software, based upon service and support revenue rather than up-front licensing fees.

Moving software to the cloud, and thus paying for access on an as-needed basis, offers another alternative method for pricing software. Yet it remains unclear, at least to me, whether cloud-based software will replace desktop software either as quickly or as completely as some folks seem to think.

It is also possible, of course, that IT users will decide that commercial software is worth the premium they pay for it. In some cases, for example, businesses will decide that it's a fair deal to buy a desktop system when Windows 7 alone represents 50 or 60 percent (or more) of the total initial investment.

In other cases, however, I suspect that they will not. It may not be easy or even very fair to compare hardware and software price trends, but I think it's inevitable. As a result, many buyers are sure to wonder why software costs are consuming ever-larger slices of the IT budget pie.