The 2004 R&D Scoreboard, issued by the UK Department of Trade and Industry, contains a table of companies in the FT Global 500 with high R&D intensity (defined as R&D spend as a proportion of sales). [see earlier discussion: Investor Perspective]
Among large software companies, Computer Associates (CA) comes out top. It spends nearly $400m on R&D, which represents more than 20% of sales.
Perhaps more than any other software company, CA has largely grown by acquisition. It has accumulated a remarkably diverse portfolio of software products - typically mature products with an established and loyal customer base - which it largely manages as cash cows. Most products are upgraded at least annually, to reflect changing platforms (e.g. new versions of Windows), emerging standards (e.g. web services), security threats, and other external requirements. This clearly represents a massive amount of development effort, but it is hard to see much of this expenditure as directly contributing to innovation. CA does have some genuine and potentially valuable innovation, but this only accounts for a fraction of its R&D expenditure. [see earlier discussion: Death of Software?]
CA is spending a large amount of development money on maintenance - in other words, simply keeping its products up-to-date. CA's customers are spending large amounts of money too - not just in buying CA's products (thus helping to fund CA's maintainance expenditure) but in additional hardware and other resources.
This maintenance cost is necessitated by innovation elsewhere. Whenever Microsoft releases a new version of Windows, and whenever some malware producer discovers a new vulnerability in Windows, this forces large numbers of software producers to upgrade their products and services, and even larger numbers of users to patch or upgrade their hardware and software. There may indeed be some innovative new features in the latest version of Windows, but the total cost of this innovation is spread right across the IT industry, and is thus mostly incurred by people who have no use for these innovative new features.
Does this mean that people who use these innovative new features (the early adopters) are being subsidized by people who don't? Well in the short-term, perhaps yes. But the early adopters incur high risk and may provide valuable feedback - thus ultimately benefiting everyone. So it's all fair in the long run. Isn't it?