Microsoft R&D – a user’s manual

The head of Microsoft’s Cambridge lab outlines the strategy: to scout the science world for new software markets.

There’s an old saying in Detroit: What’s good for GM is good for America. Now, Redmond, Washington, headquarters of the world’s biggest software company, has a new variant: What’s good for science is good for Microsoft.

Such is the underlying philosophy behind one of the world’s most ambitious, and international, corporate research programmes. The company, like many multinationals, has been pumping billions into R&D – and in fact, recently announced yet another budget increase. Yet the way it’s spending its pure-research money is noteworthy: seeding it across scientific domains, from biology to the environment, and across scientific institutions – so far, 219 announced lab partnerships in Europe alone, involving about 500 independent researchers.

The aim, according to Andrew Herbert, managing director of Microsoft’s European research centre in Cambridge, England: to push forward, and give the company a window into, a broad range of scientific fields that will shape our future – and also shape the future of the software industry. He cites a joint research project with France’s state computer institute, INRIA, in applying computers to mathematical theorem-proving. Another example: developing a better computer model, with Italy’s University of Trento, for systems biology.

Both are research topics critical to the future – in these cases, the future of mathematics and biology. But both could also give Microsoft commercially valuable insights into whether there’s a market for new software in those fields, and if so whether it should enter the fray.

Scouting the science market

“If either of these (projects) work, there would possibly be a software-based business that Microsoft would want to be in,” says Herbert. Based on the results, the company might decide to add new features to old products or launch entirely new products; or it might decide to stay out of the market and simply license its technology to others. When considering these future market opportunities, Herbert says, “I am looking at it through a research pair of spectacles. Others in the company are looking at it through marketing spectacles” or other business points of view.

Of course, there have been many broad multinational research programmes: From the 1950s through much of the 1980s, for instance, AT&T and IBM supported the world’s most eclectic – and Nobel-winning – pure-research programmes in corporate history. And most multinationals today pursue R&D in collaboration with university researchers, to get more research bang for the buck. But truly grand visions in corporate research are rare today, when Wall Street has little patience with long-term planning.

The Microsoft plan was laid out earlier this year, in a report, Towards 2020 Science, that it compiled on the world’s greatest scientific challenges. The 86-page study, which made the cover of leading scientific journal Nature, summarised scores of problems – in complexity theory, topology, synthetic biology, forecasting methodologies, and other fields. It also outlined a series of software, computing and communications problems that, if solved, could advance research in all of the other fields and produce what the lead author, Stephen Emmott of the Microsoft Cambridge lab, called “an era of science-based innovation that could completely eclipse the last half century of technology-based innovation.”

‘A finite cheque’

Wall Street investors have already been jittery about the vast scale of Microsoft’s R&D spending – and if any financial analysts had read the ambitious scientific report they might have found even more cause for concern.

But Microsoft recognises, Herbert says, that it can’t do everything on its own and is consequently picking its bets. When asked if he has a blank cheque for R&D spending, he replies: “I have a finite cheque, but I have the choice of how to carve it up” among the many potential projects.

He declined to disclose the size of the European research budget, but said the lab’s funding has been rising by between 5 per cent and 15 per cent a year for the past four years, and is likely to continue growing at roughly that speed.

The red-brick Microsoft Cambridge lab – with 80 PhDs, 20 computer-support technicians, and administrative staff – is on the outskirts of the city, near the University of Cambridge’s famous Cavendish laboratory for physics. It’s one of five Microsoft Research labs world-wide, and the only one in Europe.

The company also has a bigger network of what it calls Innovation Centres, which work on development projects. For instance, on 7 November Microsoft announced a new innovation centre project in Portugal, to develop speech-recognition and other tools for translating Microsoft products into Portuguese.

At the Cambridge lab, a group of 10 to 15 works on external relations – in part, forging research partnerships with universities across Europe, the Middle East and Africa. The lab also is involved in IP Ventures, a Microsoft initiative to spin out discoveries that it doesn’t want to develop by itself.

It also funds about 60 students across the region to work on research projects in partner labs. And finally, it maintains two jointly funded research centres – one at the University of Trento, in computational systems biology, and the other in Paris with INRIA, chiefly on software security.

A model of you

One of the Trento projects illustrates the Microsoft approach. There, a group of researchers is trying to adapt an engineering tool, called stochastic pi calculus, to systems biology. The aim, Herbert says, is to develop a computer model of the sequence of chemical reactions that happen when an antigen triggers the human immune system. The broader goal, Herbert says, is “a complete model of you” – a silicon simulator that would allow doctors to test medicines on a computer model of each patient, before actually writing a prescription.

Of course, thousands of researchers around the world are also working on aspects of this problem, but the Microsoft project – funded equally by the company, the university and the regional government – focuses on software tools to speed the research. Preliminary development tools are already public, on a Microsoft Web site.

The result, for Microsoft, could be a new software feature, a new product or simply IP to sell to others; it’s too early to say which, yet. Herbert likens his lab’s role to “scouting” the world of science for future opportunities for Microsoft.

Some of those could simply be new features – in data-analysis, presentation or forecasting - for Office, Windows and its other general-purpose products. Others could be software tools specifically intended for scientists; for instance, he notes that astronomers could use better software for visualising sectors of space in three dimensions.

Does that mean that Wolfram Research, MathWorks and other scientific software specialists should start getting alarmed about new competition from Microsoft? Herbert demurs: generally, he notes, Microsoft products are “horizontal” – standardised tools that can be used by all industries, rather than specialised tools for individual sectors.

And even if Microsoft does act, he says, “hopefully it will grow the market. In a rising tide, all boats float.”