IBM is adding "big bet" projects in exploratory research, Dean says. "For example, how does DNA interact with carbon nanotubes for self-assembly of circuits?" he says. "[There is] also spintronics and real-time analysis of large amounts of data from sensors all coming at you at one time. These are things nobody has done before."

IBM is also increasingly collaborating with customers, Dean says. For example, it is working with "a prominent candy company" to apply a prototype Web analysis tool, called Business Insights Workbench (BIW), to find hidden patterns and meanings in structured and -- most important -- unstructured data. "The company wants to do business in some emerging markets," Dean explains, "and [BIW] will look at trends and biases within a culture to predict whether a particular brand of chocolate will be bought." Another big plus: The client company will be able to analyze Web-based information in any language without translation, Dean explains.

Microsoft Research: The university model

While companies such as HP rely on partnerships with universities to gain access to many basic technologies, Microsoft is trying to create its own university, Chesbrough says.

Indeed, Microsoft Research can reasonably claim to have already done so. It has been led from its beginning in 1991 by Richard Rashid, a former professor of computer science at Carnegie Mellon University. Rashid makes no secret of his operating model for Microsoft Research: "The work we do is not that different from what you'd find at Stanford or Berkeley or [Carnegie Mellon University] in the sense that it is publishable basic research that is peer-reviewed."

Rashid adheres to an outlook that sets Microsoft Research apart from many IT companies where every research project has a product line in mind: "Our research may have a short-term impact on the product groups, but that's not why we do the work -- it's a consequence of the work."

For example, Rashid says, he started a small research group in computer vision -- machines that can see -- in the mid-1990s when no Microsoft product at the time seemed to need that kind of technology.

"But in a few years, digital imaging and photography became a huge value to other parts of the company in areas like photo processing, image analysis and signal processing, and things like Windows Media and audiovisual codecs came out of that earlier work," he says. "And now things like Microsoft Surface are all based on computer vision technology. In fact, you could have read about the work leading up to Surface five years ago."

Because Rashid's philosophy is to first do good computer science, and then see where it might fit, he focuses first on people. "My biggest lever is who I hire and who I fire; it's not telling people what to do," he says.

Asked why he decided to launch a new lab in Massachusetts, that will work at the intersection of traditional computer science and the social sciences, he says, "You don't establish a lab without the right person to do it. We had a great researcher, Jennifer Chayes, and she was really excited about a lab in that area [of the country]. She's fabulous and has done incredible work. If it wasn't for her energy and initiative, it probably wouldn't have happened."