So without that change in accounting, would Microsoft still have ranked above No. 2, Ford Motor Co., on our list? And, more generally, would companies not currently on the list have appeared on it if they had counted stock options granted to researchers? There is simply no way to know.

We might know a lot more next year, however, if the new accounting rule takes effect. The rule's proponents claim that it will make corporate finances more transparent. Opponents, such as the American Electronics Association, in Washington, D.C.—whose members include Advanced Micro Devices, Broadcom, Intel, and Texas Instruments—say the requirements are difficult to implement and will make it hard to recruit talented engineers; they support legislation pending in Congress that would derail the new standard. But unless the new rule is adopted, comparing one company's R and D spending with another's is a difficult, if not spurious, exercise.

Even With All The Accounting Uncertainty, it is possible to divine some macro trends from the list [see table, IEEE Spectrum Top 100 R and D Spenders"]. Overall, the Top 100 upped their R and D by 2.2 percent to $236 billion despite sluggish sales growth of 0.8 percent. Bucking the trend, technology hardware companies, which account for about 20 percent of the overall R and D spending, reduced expenditures by 4.9 percent in 2003, even though their sales increased by 2.3 percent. The auto industry had relatively flat spending for R and D, with an increase of 3.2 percent. Semiconductor firms increased spending by 5.3 percent, but that was less than half their increased sales, 12.6 percent, which rebounded from a 10 percent slump the year before. The software and services sector increased the most, with an 18.6 percent gain in R and D spending, mostly because of Microsoft's spending.

Software dominates R and D, and not just in the sense that, for the first time ever, a software company is the world's top R and D spender. At most companies, software is taking up a bigger chunk of the development budget year after year, and it is at the heart of everything from drug discovery to automobile telematics. In Stockholm, Sweden, for instance, Hakan Djuphammar, vice president of systems management for LM Ericsson, says that over the last 40 years the telecommunications equipment sector has shifted from making hardware for phone networks to developing software. His company today spends about 85 percent of its R and D money on software.

"With the continuing standardization of powerful processors and hardware components, there won't be any differentiation in hardware 10 years from now," says Djuphammar, adding that this trend extends through industries like semiconductors, information technology, and consumer electronics. "You'll buy extremely powerful processor boards off the shelf and the only differentiation will be in software."

So now corporate labs are filled with programmers and software engineers. Indeed, the romantic image of monomaniacal scientists toiling for years in a company's central lab to produce the transistor, polyester, and the scanning tunneling microscope no longer holds. The dominant trend in industrial R and D over the last two decades has been to decentralize research and look outside the company for innovations. The result: company labs spread across the globe, populated by researchers who must fill several roles—investigator, inventor, technology scout, and, increasingly, consultant to customers.

The change in how companies and their researchers conduct R and D mirrors shifts in corporate structures and missions. Henry Chesbrough points out in his 2003 book Open Innovation: The New Imperative for Creating and Profiting From Technology that the vertically integrated enterprise exemplified by Xerox Corp.—which made everything from the toner to the lasers used in the copiers it leased and serviced—continues to give way to horizontally integrated companies that combine components produced elsewhere. Instead of using R and D resources to make better, faster, cheaper widgets, companies like Hewlett-Packard, IBM, and Lucent are employing researchers to service customers that need help putting those widgets together into novel systems.