From Big Blue to NYU

On the eve of his departure for academia, IBM Research's Paul Horn talks about Blue Gene, open-source technology, and the strategy behind sharing patents

With a roster of 3,400 scientists, IBM Research is one of the largest corporate research outfits in the U.S. Paul Horn, the man who has been running it since 1996, recently announced his retirement—ending a 28-year career at IBM (IBM). He will be replaced by John Kelly, another longtime IBMer. Horn, 60, is becoming a Distinguished Scientist in Residence at New York University. He was only the sixth lab director since IBM set up the organization in 1945.

During his decade at the helm, Horn made the research organization an essential part of IBM's transformation from a computer hardware company into an organization that increasingly sells innovation and business expertise. On his watch, for instance, the company undertook a shift in its approach to performance computing, placing new emphasis on energy efficiency.

BusinessWeek Senior Writer Steve Hamm interviewed Horn after his departure was announced. Edited excerpts of their conversation follow:

When you took over as the IBM Research director a decade ago, IBM had just come through a near-death experience [related to its over-reliance on the mainframe business]. Yet rather than slashing the research department the company kept investing. Why?
There was a feeling by our then-CEO, Lou Gerstner, who brought me in as head of research, that research could and should be part of the difference and help turn the company around.

IBM Research had long acted on the belief that researchers should actively engage with people in the company's business units, but you dramatically increased the frequency and depth of these interactions.
The interactions with the product divisions are the channel through which our innovations get to the marketplace. You can have great ideas, but you can only be innovative if you impact the marketplace. There's a limit to how tightly you can be connected. You want to do exploratory work, or else you just become a development lab. But it's very important to have the credibility that your ideas are not some wacko things generated by a bunch of researchers who don't understand what it takes to bring out a new product.

We worked it so the business units do some of the funding of research. If you're really providing value for them, the real measure is that they take money out of their development budget and give it to you. If you're not critical for them to succeed, you don't get any of that.

Since the late 1990s, IBM has exited one commodity technology business after another. How have you harnessed IBM Research to come up with innovations that IBM could build new businesses around?
One of the reasons why research has been so successful is IBM's business model. As a corporation we want to be in high-value segments of the information technology market. Those are the businesses where R&D makes the most difference. For example, we got out of PCs, which have margins on a good day of 3% or 4%, and we got into businesses, like our software businesses, where you can really drive differentiation with the world's best R&D. We have a business model that keeps research in the center of things.

In the late 1990s, you pushed through a radical rethinking of high-performance computing, which resulted in IBM's Blue Gene supercomputer. It uses thousands of inexpensive, low-power processors rather than expensive, high-performance chips.
Blue Gene started as a research project to fundamentally rethink computation. How can you increase computational power and, at the same time, lower the wall-plug power and the cost of energy? Blue Gene was aimed at changing the game substantially, and I think we did. You'll see a lot of the concepts in Blue Gene ripple through the whole IT industry, all the way down to your PC.

Is there a role for IBM research in virtual worlds and social networking? (See BusinessWeek.com, 11/20/06, "Palmisano Gets a Second Life.")
About a year and a half ago we had an online innovation jam—internally and partially externally—where a whole bunch of research technologies we had been working on were discussed. The question was: How could we use some of these technologies in the marketplace? Ten projects came out of the jam for internal incubation. One of them was in virtual worlds. That project was given to the research division. So we have a little business in research in virtual worlds. We're helping companies like Reuters (RTRSY), Sears (SHLD), and Circuit City (CC) establish a presence in virtual worlds.