Inside the labs at Microsoft, HP and IBM
- 30 November, 2010 08:48
At Microsoft, Hewlett-Packard and IBM, investment in research and development is a reflection of corporate culture. This three-part piece examines the different approaches taken by each of these influential tech companies. Hewlett-Packard prides itself on its pragmatism, while Microsoft holds the flag of basic research aloft -- and IBM continues to file more patent applications, year after year, than any other tech company.
HP Labs: Inventing ways to do more with less
by James Niccolai
HP Labs has seen some big changes in the past few years. In 2007 it hired Prith Banerjee, the dean of engineering at the University of Illinois-Chicago, as its new director. A year later the labs started to narrow its focus from the 150 or so projects its scientists had been working on to 20 "big bets" -- projects that, if they paid off, HP hoped would contribute directly to its bottom line.
The labs had lost its focus, with researchers squirreling away in small groups on projects that, while interesting, didn't always serve HP's wider goals. Banerjee calls the new approach "innovation with purpose." He has arranged the big bets around eight broad areas that HP sees as core to its future, including analytics, data management, intelligent infrastructure, sustainability and, of course, the cloud.
Banerjee brushes off the recent criticism that HP Labs may be underfunded. Last fiscal year the company invested 2.5 percent of its total revenue in R&D, compared to 6.1 percent of revenue at IBM and 14 percent at Microsoft. That's still $2.8 billion, however, of which the labs budget is only a fraction -- in 2008 it was $150 million. The vast majority is spent on later-stage development in HP's product groups.
"People get all hung up on this percentage of revenue figure, but I could spend a ton of money doing completely useless research and development," Banerjee said in a recent interview. "It's the research output that matters, not the amount of dollars you put in."
Plus, he says, the labs devotes more of its resources today -- about one third -- to basic exploratory research that may not pay off for five or 10 years, in areas like nanotechnology and quantum computing. That's up from 10 percent a few years ago. Another third is for applied research, which covers fundamental technologies like compression that can be applied to multiple product areas, and the remainder is for research tied to specific products that could pay off in as little as a year.
Displays are one big focus for the company. HP doesn't manufacture displays, but it sells around 70 million of them each year in PCs, printers and other products. If it could come up with a radically cheaper way to make displays, HP could license the technology to others and also lower its own product costs, says Carl Taussig, director of HP's Information Surfaces Lab.
To do that it is trying to come up with a completely new manufacturing technique. Today, displays are produced essentially one at a time, using a photolithography technique similar to that used for semiconductors. But Taussig and his team have been developing a "roll-to-roll" process that allows manufacturers to almost literally print circuitry onto screens one after the other.
The process could go into production for certain types of screens in as little as two to three years, he said. HP has custom-built machinery running in its labs that can "print" displays at 5 meters per minute from long sheets of a thin polymer material that resembles aluminum foil. HP's breakthrough was to invent an imprint lithography process that allows the circuits to be layered on top of each other on the flexible screens but still stay in perfect alignment, Taussig said.
LCDs cost about $100 per square foot to produce today, and the roll-to-roll method could reduce that to $10, he said. Of course, the big display makers are working on similar technologies, and one challenge for HP will be getting the technology out of its lab and into at least small-scale production in order to show that it works.
Printing, not surprisingly, is another focus, but not the type found in homes and offices. Most of the optimization for those products has already been done, so the labs are focused instead on digital commercial printing presses, a fairly new area for HP where it hopes to challenge the incumbents, said Eric Hanson, director of HP's Commercial Print Engine Lab.
The labs have already developed a new type of ink for HP's Indigo digital press. Hanson wouldn't reveal anything about the newer presses HP is developing except to say they are making "good progress." But he was keen to show off a technology that could help smooth the industry's transition from analogue to digital printing, and thus eventually help HP to sell more equipment.
One obstacle for the industry is that paper mills don't have an efficient way to remove digital ink from all types of paper -- something they need to do when they recycle magazines and brochures to make new paper. So one of Hanson's team, Hou Ng, has developed a "surfactant" that allows digital and other inks to be skimmed off in a foam after the pulping process.
HP plans to give away the formula so that other companies can produce the chemical, Ng said. The idea is remove any obstacles that could prevent the digital print market from expanding.
The Labs' narrower focus doesn't mean it works only on projects tied to HP products. HP doesn't design its own server chips any more, for example, but it continues to invest in microelectronics. In August it announced a partnership with Hynix to commercialize a new memory technology called memresistor that derives from work in HP Labs.
The labs also contributed a video compression technology for HP's Halo and Skyline videoconferencing systems, and its on-demand magazine printing service, MagCloud began as an idea in HP labs. The labs also helped develop HP's ePrint service, for printing from any e-mail device to a printer anywhere in the world. It's also finding ways to bring research to market through its services arm, including a MEMs sensor technology being used by Shell for oil exploration
As part of the effort to make HP Labs more efficient, its researchers must now submit longer pitches that include a business proposal when they are seeking funding for new projects. The pitches are reviewed by business people and technologists within HP product groups as well as by other scientists.
It has also been working to get technologies to market more easily. For instance, HP created a spin-off company to help prove the effectiveness of its roll-to-roll display manufacturing process. It's designed to solve what Taussig called "the jumping the chasm problem" -- getting a new technology from the labs into widespread use.
"That's a missing piece that we haven't always done very well here," he said. "We're trying to move ourselves into something that's more like a manufacturing environment than an R&D environment. It's not easy for researchers to get that type of discipline, but we're getting better at it."
Microsoft: Keeping the focus on pure research
by Nancy Gohring
Leaders of Microsoft's research group sometimes speak with nostalgia about the glory days of computer science research, when labs like the revered Xerox Palo Alto Research Center and Bell Labs worked on ambitious projects.
Most of those famed institutions are gone or are shadows of their former selves, victims of budget cuts and investors who demand that every dollar be clearly earmarked toward development of profitable products.
While its research group never quite ranked among those famous labs, Microsoft likes to boast that it's one of the few remaining public companies that still do pure research, including kinds that might not turn into products for many years.
"Many of the other companies that used to have research labs, in previous downturns they told their labs to focus on doing something incremental, to fill gaps in their product lines," said Andrew Herbert, managing director of Microsoft Research Cambridge. "The track record is that people who are good at research aren't good at product development. If researchers get under the feet of people who are good at doing product engineering, it doesn't work very well. The labs get demoralized if they aren't doing interesting research and it all goes away."
Despite the economic downturn, Microsoft views research as critical to its future. In fact, it says that pure research is especially important during downturns.
"I think of research as one of the things that we have to do and elect to do in order to ensure we survive over the long term," Craig Mundie, Microsoft's chief research and strategy officer, said early last year at an annual research event. Companies that cut research in the face of short-term pressure or never start pure research tend not to last very long, he said. "My belief is the company would struggle to survive and prosper if we didn't have research investment," he said.
Microsoft's research group, which includes 1,000 people spread across six labs around the world, has a mission of advancing the state-of-the-art in computer science, Herbert said. It is also tasked with looking at ways that those advancements might support Microsoft products.
Microsoft researchers come from a wide range of disciplines and have the freedom to pursue subjects that interest them, Herbert said. They include psychologists, ethnographers, sociologists, mathematicians, astronomers and physicists.
The variety of backgrounds of the researchers reflects the way that computer science has progressed over the years, he said. "Historically, when computer science started as a subject... it was about building better computers, designing better programming languages, inventing algorithms to take on computing tasks we had at the time," he said. "As the subject developed, we as computer scientists needed to build conceptual tools and models to tackle these problems which it turns out are useful in other areas."
For instance, Microsoft researchers built tools that are helpful in testing very large and complex software, essential to try to guarantee that the code does what it's supposed to, he said. Those same tools happen to be useful to biologists at the University of Southampton in the U.K. who are working on modeling the human immune system.
When asked how he decides what the researchers should focus on, Herbert said: "The glib answer is, we don't." Instead, the goal is to hire smart people and give them the chance to pursue topics they want to work on, he said. "When you give them that freedom, they naturally want to work on things that are intellectually challenging in their own right. And the people who commit to work in research labs not only have that academic passion but they want to have an impact on the world and change something."
He encourages the researchers to work together and be aware of what each other are doing so that they can collaborate and then develop projects, he said. "That defines the agenda. It's very bottoms up," he said.
Microsoft's research strategy has its skeptics. The company has been criticized for not doing a particularly great job turning research projects into products, said Matt Rosoff, an analyst with Directions on Microsoft. "The transfer between research and product groups has been slow at times," he said. "They might come up with good ideas but it's not clear how to turn those ideas into products."
Still, he also said that even when Microsoft does find a way to use technologies developed in the research group in commercial products, it doesn't always do a good job talking about it. For instance, technologies developed in the research group ended up in Microsoft's digital rights management products, its SQL Server, and development languages like C#, he said.
It's happened often enough that products from the research group turn into technologies driving important Microsoft products, Herbert said. "Sometimes we find ourselves looking at things in research and people in the product groups probably say, 'we'll never do it that way,' but then there's a change in the market or the business model that suddenly makes those things important," he said.
That happened a few years ago when Microsoft decided to more aggressively pursue the search market. The research group had already invested in technologies like information retrieval which allowed Microsoft to get moving very quickly on a search product, he said. "When we were doing information retrieval, I'm sure we would have found people in the product groups saying, 'why are we doing this?'" he said.
Microsoft Research is most currently boasting about its role in developing technologies behind Kinect, the new Xbox product that will allow users to play games without a controller. The Xbox group had done work developing the system that tracks user movements. "But what they needed was technology that would help their applications confirm that what the tracking system claimed happened, had indeed," Herbert said.
The research group had done work on object recognition technology, which is most often directed toward image search in a search engine or image classification. But in this case, the technology developed in the research group went toward helping Kinect ensure that it is tracking user movements accurately.
That same technology might also turn up in other products. Microsoft has had some meetings with Tesco, a large supermarket chain in Europe, which has expressed interest in using the technology as an alternative to barcode scanners at checkout lines.
Herbert described some other research projects that may become commercial projects. One is known as SenseCam. "The original concept was a small wearable digital camera that decided when to take pictures for itself," he said. Researchers imagined that kids might wear one on a school field trip or workers in safety-critical environments might wear one to be able to later recall exactly what they'd done.
In addition to the camera, the SenseCam includes sensors and processors that determine when to take a photo based on changes in the environment, like movement or changes in lighting and sound.
A hospital heard about the project and began using SenseCam's for patients with various memory loss problems. Those patients were often previously instructed to keep written diaries that they could review. "That's very inefficient. Over time you build up so much diary you can't really review it," Herbert noted.
With SenseCam, the patients can flip through photos from an event. "It turns out that just watching that a few times actually seemed to burn a bit of memory into their brains," Herbert said. A company called Vicon has since licensed the SenseCam technology and is developing products around it.
It's hard to say exactly how much money Microsoft spends on research each year. The company typically spends between 14 percent and 15 percent of revenue on research and development, said Herbert. But it does not break out research on its own.
In a company like Microsoft, that makes it hard to know how much goes toward pure research, Rosoff noted. "With a software company, a lot of your cost is development. That's how the business works. You don't have a high cost of goods. It's about the cost of development and marketing," he said. That means the bulk of the research and development budget is likely development.
The last time that Rosoff's firm investigated Microsoft's research budget was 2002, when it concluded that about 5 percent of the company's research and development budget was pure research. At the time, that amounted to about $250 million, he said. He estimates that budget has grown to around $500 million a year now.
Microsoft is "somewhat unique" among its peers for having a pure research group, Rosoff said. The group is sometimes "judged more like a university research group," based on the papers that Microsoft researchers publish and their participation in setting the agenda for computer science research, he said.
The research group has grown since it was founded around 20 years ago, but otherwise not a lot has changed, Herbert said. When the group was growing rapidly, it was adding the equivalent of a university computer science department each year, he said. As the labs got bigger, the researchers had to adapt to working together as a global organization. Still, "we've had the same mission statement since day one," Herbert said.
IBM: Research is survival
by Joab Jackson
When the economic conditions are difficult, the temptation of any large company is to slash its research and development. After all, for most corporations, research does not directly contribute the bottom line, and given its speculative nature, may never do so.
Still, when the global economy hit the doldrums in 2008 of 2009, IBM shielded its R&D work from budget reductions.
"I've never been pressured to look for places to cut," said Robert Morris, the IBM vice president who heads up the company's research in the field of services. For IBM, research is not a luxury or a public relations play; it is essential to the company's survival.
"Many companies are reacting to the current global downturn by drastically curtailing spending and investment, even in areas that are important to their future [while] we're continuing to invest in R&D," IBM president and chairman Sam Palmisano instructed company shareholders in a 2009 letter. "In other words, we will not simply ride out the storm. Rather, we will take a long-term view, and go on offense."
Since 2002, IBM has increased R&D spending 21 percent. In 2009, it had spent $5.8 billion on R&D. The company now employs 3,000 researchers across eight labs worldwide, and is building a ninth lab in Brazil.
Rarely a week goes by without word of some new IBM innovation: In the middle of September, the company announced that it was shipping the world's fastest microprocessor. The week before, Big Blue announced that it invented an optical bus that added another 50 percent in throughput speed. And the week before that, it announced a virtualization technology that solved the problem of moving live virtual workloads across different data centers.
And some of IBM inventions have changed the world, even if they sometimes show up in fields far from IT. IBM can lay claim to not only inventing the personal computer, the disk drive and relational database, but also the SABRE travel reservation system, the technique that led to LASIK eye surgery and a blood separator technology used to treat leukemia.
IBM has held the record for most U.S. Patents issued in a year for the past 17 years -- and the recession hasn't slowed the company. In 2008 it received over 4,186 patents and in 2009 the number had jumped to 4,914.
But perhaps the best measure of success is not the number of patents it is awarded, but how well IBM profits from this work.
In fact, IBM has made headway on one of the most difficult problems facing any company investing in R&D: How to turn the research into business.
Venture capital investors sometimes refer this challenge as crossing the Valley of Death, meaning it is a slow, difficult trek to transform some solid research into a profitable product or service. And many good technologies have died on the way.
"We cracked that problem. We never have any problem getting the research close to the development," Morris said.
The trick, Morris explained, is to get the business units and even clients involved in choosing which projects to pursue. "We do not build things in the lab and figure out how to transfer them into the field. We're not an ivory tower. We're not a sandbox," Morris said.
Each year, the labs presents an outlook report to the IBM chairman suggesting future trends IBM should follow. In many cases, the company follows the lead of the report. Analytics, for instance, was highlighted in a report nearly a decade ago. Earlier this year, IBM announced that it anticipates that analytics will generate up to $16 billion in annual revenue for the company by 2014.
And work that the labs have done in the field of analytics has already been reapplied in IBM systems. For instance, IBM Research teamed with the IBM Global Business Services unit to develop analytics software to catch tax cheaters.
"We built in a real time analytics process, so as a tax return hits the computer system, we score it for the probability that it's a fraud," said Shaun Barry, an IBM global solutions executive who oversaw the implementation. "For those high-probability things we immediately stop the refund" so it can be examined manually.
The New York State Department of Taxation and Finance, which suspects is losing up to $1 billion annually due to suspicious tax returns, has been using the system. Use is expected to generate an additional $100 million revenue over a three year period.
Not surprisingly, the company is investing heavily into areas where it feels the application of information technology may revolutionize some field, such as analytics, city planning, health care, biology or energy.
One such project is an urban traffic prediction system. This system, being tested in a number of different cities such as Singapore, can take input from various road sensors and, using a traffic flow model for that city, not only show where the traffic jams are now, but even can predict where traffic jams may occur. With this knowledge, traffic management departments can make adjustments of the traffic, through road signs that suggest alternate routes.
"We calibrate a set of models on the most recent data, and in real time these models are applied to the real-time data feed," said IBM researcher Laura Wynter who works on the system.
Health care is another area of interest to IBM. In July, the company announced that it would invest $100 million in health care technology. The company's Zurich research center is working on what it calls a lab-on-a-chip that could radically cut the costs of lab testing. This device is actually a small strip that can soak up a sample of blood and detect the proteins are tell-tale signs of viruses and diseases.
"Our intent was to leverage our expertise in microfabrication. IBM makes a lot of processors and we can use these fabrication techniques to make chips that interface with biology," said Luc Gervais, the researcher working on the technology.
For IBM, research is a way of anticipating, or even creating, new markets. If you look over its 100 year history, you can see IBM moving from one field to the next. While it started out selling tabulating machines, IBM is now the largest services company in the world.
"When I first started doing services research, it was all about IT services," Morris said. "Now we're working on our clients' services: Health care services, city services, government services."
Research Labs At a Glance:
History: IBM opened its first formal research lab in 1945, using a renovated fraternity house near Columbia University in New York. In 1961, the lab moved to Yorktown Heights, New York, where it was renamed the T.J. Watson Research Center.
Number of researchers: 3,000
Lab locations: In the U.S.: Almaden and San Jose, California; Austin, Texas; Yorktown Heights and Hawthorne, New York; Cambridge, Massachusetts; Brazil; China (Beijing, Shanghai); Haifa, Israel, India (Bangalore, New Delhi); Tokyo, Japan; Zurich, Switzerland
Key research areas: Computer science, science and technology, service science, storage systems, analytics, distributed computing, networking, future systems, health care IT, "Smarter Planet" initiatives (IT-led research in urban planning, agriculture, energy management, government, infrastructure, education and others).
History: Microsoft Research was founded in 1991
Number of researchers: 850
Lab locations: Bejing, China; Cambridge, England; Bangalore, India; Cambridge, Massachusetts; Redmond, Washington; Mountain View, California
Key research areas: Computational linguistics, economics, health and well-being, human-computer interaction, machine learning, social science and theory
History: Research group founded in 1966
Number of researchers: 500
Lab locations: Palo Alto, CA; Bangalore, India; Bristol, UK; Singapore; Haifa, Israel; St Petersburg, Russia
Key research areas: Analytics, It Infrastructure, Digital commercial printing; Displays, Interaction and communications, Sustainable computing, Cloud platforms, information management