Microsoft's advance team

REDMOND, Wash. -- Tacked to a corkboard in the office of Rick Rashid, senior vice president for research at Microsoft Corp., is an uncashed check for 25 cents. It's booty from a 1992 bet he made with his friend Dave Gifford, an MIT computer science professor.

Rashid had left a secure job on the Carnegie Mellon University faculty to start Microsoft Research, gambling that a world-class research organization could be built and sustained within a software company. Gifford was one of the skeptics, and he wagered that Microsoft would not be around as long as Digital Equipment Corp. in Massachusetts.

Microsoft Research by the numbers

Microsoft Corp. has been increasing its research outlays as others scale back. Here are some measures of its global research effort:

Total research and development spending: $6.8 billion in 2004
Microsoft Research employees: about 700 worldwide
Patents: 538 issued in 2003
US research hub: Redmond, Wash.
Other research centers: Mountain View, Calif.; San Francisco; Cambridge, England; Beijing
Technology disciplines: more than 55 domain areas, from social computing to machine learning to human-computer interaction.

SOURCE: Microsoft Research

A dozen years later, what remains of Digital has been absorbed into Hewlett-Packard Development Co. And Microsoft, which romped and strutted through the 1990s, has overtaken IBM Corp. as the largest corporate spender on research and development. In its 2004 fiscal year, Microsoft's research outlays totaled $6.8 billion, spread among dozens of technology and product areas, ranging from search and machine learning to large-screen displays and wireless multimedia applications, artificial intelligence, and natural language processing.

Because of its collaboration with Microsoft product groups, Microsoft Research consumes a relatively small share of the overall R&D budget. The company has disclosed only that it spends more than $250 million a year on its research arm, though some estimates put that figure at closer to $1 billion. Whatever the number, Microsoft Research is clearly the forward line of the company's technology thrust, though some question whether Microsoft has gotten a bang from its research bucks.

"Our main mission is to move the state of the art forward in the all of the areas in which we conduct research," Rashid said. "And when ideas pan out, our second goal is to move the technology into products as soon as we can. . . . We really are the future of the company. We're here to make sure that Microsoft is here 10 years from now."

Such a statement might sound alarming for a software behemoth that earned $1.3 billion on revenue of $9.2 billion for the three months ended March 31. But the famously hard-charging company has entered the mature phase of life that all rapidly growing businesses eventually must face. (Remember Digital?) Over the past three years, Microsoft's expenses have grown faster than its income. And this month, chief executive Steve Ballmer sent an open letter to employees announcing the technology world's former enfant terrible would be launching a $1 billion cost-cutting program.

Against that background, Microsoft Research, the unit Rashid built at the urging of Nathan Myhrvold, then the company's high priest of advanced technology, has assumed a role as a kind of guarantor of Microsoft's survival. Rashid's organization, modeled after top academic research labs in schools like Carnegie Mellon or MIT, has assembled a 700-person brainpower trust that mixes rising young research stars with graying technology luminaries, such as Digital veteran C. Gordon Bell, the minicomputer pioneer, and Xerox PARC alum Gary Starkweather, who invented the laser printer.

Microsoft Research has set up satellite labs in San Francisco; Mountain View, Calif.; Cambridge, England; and Beijing. And it has strengthened its ties with the company's product groups, sitting in on their quarterly meetings with Microsoft's chief software architect, Bill Gates, known as "billg reviews," and hosting an internal fair, called TechFest, to loop product developers into research projects. Rashid likens such outreach to running a dating service within Microsoft.

"Their research is more strategic now," said Alva Taylor, professor of strategy and technology at Dartmouth College's Tuck School of Business. "Microsoft executives know they have to keep growing, and the growth is not going to happen just through another operating system on another laptop.

"So they have to figure out what other things they can do to give themselves the kind of growth that Wall Street wants."

But critics question whether Microsoft's research investments have been paying off in compelling new products for consumers -- or merely capitalizing on the company's dominant Windows platform to help cram more marginal features into new versions and upgrades.

The company's $6 billion-plus annual commitment to R&D "is literally enough to fund at least 2,000 start-up companies," suggested free software advocate Andy Hertzfeld, one of the designers of the original Macintosh computer. "If only 1 percent succeeded, you'd still get 20 hit products a year. . . . The only reason they [Microsoft] spend so much is because their monopoly brings in so much. Even given that level of spending, the cash is piling up so fast that it's embarrassing to them."

Microsoft Research officials defend their intellectual output, noting that almost every Microsoft product released in the past decade has contained programming code or engineering designs that originated in their labs.

They also point out that many of their research projects -- one example: a "virtual observatory" that incorporates astronomy databases from around the globe -- have nothing to do with Windows. The company was issued 538 patents last year, and now holds 5,100 in the United States and 1,300 abroad, with thousands more pending. And its researchers are among the most prolific authors of papers for scientific journals and conferences, in disciplines ranging from graphics to computer science.

(Reflecting his academic roots, Rashid does not require researchers to seek his approval before publishing papers.)

Some of Microsoft Research's initiatives are short-term and quick to market. Research leaders cite Microsoft's new Smart Watch, the first in a planned series of "smart devices," and the handwriting technologies behind the Microsoft-powered Tablet PC as examples of new products that arose from research innovations.

Researchers are working closely with the Windows group to incorporate new search and social computing features into the next version of the computer operating system, code-named Longhorn, which is due in 2006. And while many of its competitors have redirected their efforts almost entirely toward applied research, Microsoft still does significant basic research, though even that is intended to eventually generate products.

"We're not doing research in a vacuum," said senior researcher Susan T. Dumais, a Maine native working on search technologies. "One of the exciting things about being at Microsoft is all the avenues we have of getting our research into products."

A little over a decade after its founding, Microsoft Research has made its presence felt in a corporate research landscape still dominated by IBM, which runs eight research labs in six countries. Though it spends about $5 billion a year on R&D, second to Microsoft, IBM does research in a wider range of areas but vies with Microsoft for talent and breakthroughs in fields such as software and databases.

"In some parts of our business, they are trying to catch up with the research we have," said Armando Garcia, vice president of technical strategy and worldwide operations for IBM Research in Yorktown Heights, N.Y. "But we do work in much broader areas. The mode that they're in is the mode we were in 30 years ago: Go out and hire the best people and have them do exploratory work."

Such work is underway throughout Microsoft Research, much of it led by PhDs with backgrounds as diverse as psychology and mathematics.

Lili Cheng, group manager for social computing, is readying "social software" that creates ever-changing personal maps of people with whom a computer user interacts on various projects.

Natural language researcher Lucy Vanderwende uses text-mining algorithms to prepare computer-generated "contextualized summaries" of news and academic articles, and machine learning to translate them into other languages.

Research psychologist Marc A. Smith is exploring "social accounting" technologies that will enable Web surfers to "get the cream out of the huge jug of slime" in online user groups.

"When you get good online community, it's like spinning gold from straw," Smith said.

And Kevin Schofield, general manager for strategy and communications, is making sure Microsoft avoids the missteps of historic corporate research labs such as Xerox PARC (Palo Alto Research Center), which often failed to incorporate their technologies in their sponsoring firms' businesses.

"To me, the most important metric is, 'Do our product groups like the fact that we're here?' " Schofield said.

"They want to see our research in their products. We need to listen to them and understand the business problems they want to solve."

Robert Weisman can be reached at weisman@globe.com.