The Memo That Spawned Microsoft Research—Analysis and Download

Microsoft Research, the central research arm of that tech giant of the northwest, was founded in September 1991. This is the prehistory.

With the recent stepping down of Rick Rashid—the head of the organization since its inception, who took a new job at Microsoft in July—I have decided to post the memo that gave birth to the lab and led to his hiring. I consider it not just an historic document in the annals of corporate R&D, I believe it holds some important lessons for any organization thinking about how to invest in its long-term future.

The 21-page document, entitled Microsoft Research Plan, was written for Bill Gates in early 1991 by Nathan Myhrvold. At the time, Myhrvold headed up advanced technology work for Microsoft but believed the company needed to look even farther out, and so argued for the creation of a central research organization. (Myhrvold later became Microsoft’s CTO, before eventually founding Intellectual Ventures in Bellevue, WA.)

Gates approved the creation of Microsoft Research shortly thereafter. MSR has since grown to employ more than 1,000 scientists and engineers working at seven major research labs and six other technology facilities all over the planet.

Embedded as a PDF in this post you will find the entire original document. What follows below is my analysis of the memo, where I try to provide some context for its proposals and also draw attention to what I believe are important lessons for today.

Warning: when it comes to the history of research and development, I’m a kind of junkie. Or I’ve been bit by the vampire of R&D. Or something like that. If you’re largely of the same vein, so to speak, please read on. If your condition is milder, I have also posted a much shorter version of my analysis, along with the 11 PowerPoint slides Myhrvold used to brief senior Microsoft executives after Gates approved the creation of Microsoft Research. They make for much quicker reading. You can find them here.

The Background

First, here’s a bit of context. The lab’s creation was a seminal, against-the-grain achievement in the modern corporate landscape—coming at a time when places like IBM Research, GE Research, Bell Labs, Xerox PARC (the Palo Alto Research Center, now simply PARC), and basically any centralized corporate research lab were being labeled dinosaurs. Far-out research was too chancy, better left to universities. Closer-in research should be done inside business divisions—or so the basic thinking went. Creating a central lab independent of business division budgets partly to explore advances potentially years down the road, work that might never pay off for its mother corporation, was anathema.

But that is what Microsoft Research was created to do. That was only part of the job, though. You will see in the memo that Myhrvold, an Xconomist, focused from the beginning on aligning the lab with Microsoft’s business aims—drawing lessons from the failings of other famous research labs (especially Xerox PARC).

At the time he wrote the memo, Myhrvold headed Advanced Technology and Business Development for the company. His operation was zeroed in on relatively near-term products and advancements for Microsoft (as advanced technology development should be). But Myhrvold foresaw a great need for expanding the company’s technology horizons beyond what his group or any product group could, or should, take on. So he set out to convince Bill Gates and other powers that be that Microsoft needed to set its sights farther along the road ahead, so to speak.

Microsoft Research Plan

A Lab Is Conceived

“Once upon a time,” Myhrvold wrote for Gates and others, “it was very clear what our product agenda was—simply take the ideas which were successful on ‘big computers’ and move them to ‘little computers.’”

But, Myhrvold asserted, “the days of taking something off the shelf and adapting it to our ‘little computers’ are over.” Not only has most of that shift already taken place, he argued, “microprocessor based systems aren’t necessarily ‘little’ anymore—they rank among the most powerful general purpose computers on earth.

“The onward march of hardware technology is taking personal computers to new heights of processor speed, memory and general functionality per user…We are also faced with the staggering potential offered by entirely new developments in other parts of the system, such as optical storage systems, new graphics capabilities, digital data broadcasting, digital video and a host of other hardware and software innovations which will change the landscape of our industry.”

Myhrvold concluded his Introduction by noting, “The challenge, of course, is to create the software which realizes the potential that this hardware offers. We have to explore and research the new software technology necessary for the PC environment of the future, evaluate the impact on our upcoming products, and then follow up with the leadership and investment of resources necessary to turn ideas into commercial reality….

“In very general terms, we have to invest in our future by doing more work in research and technology creation.”

Commentary: OK, Myhrvold didn’t get the optical storage part right, but the rest was pretty spot on. More importantly, he knew that great corporations need to find ways to create new products and services rather than just copying or acquiring the advancements of others. Startups and entrepreneurship are vital to future economic growth. But we can’t forget the role large corporations play in our innovation landscape—and this innovation needs to be pursued, nurtured, and developed if that corporation wants to be around in 50 years, 100 years, or longer. This was the key point of Myhrvold’s memo.

The Argument For Research

Myhrvold identified six primary areas of high-technology activity that Microsoft as a company “could and should be involved in.” (Memo, p. 3) These were:

Tracking the state of the art

Advising product groups

Technology acquisition

Advanced development

New technology & business projects


Myhrvold defined Research as “the task of working on unsolved problems in computer science which are critical to our strategic needs in the future.” He stressed that he was not speaking of basic research 10 or 20 years down the road. “This is really applied research, because we would expect to incorporate this work into products within a 2 to 5 year time horizon.” (p. 4)

The Advanced Technology Business Division that Myhrvold headed concentrated mostly on the first five areas on the above list. However, he noted, “Our plan for some time has been to expand the group, both to be able to handle more work in the existing areas and also to expand into new areas such as research.” But Myhrvold no longer thought that the ATBD was the proper home for research. “After looking into this in more detail…” he continued, “the need for a [separate] research group has become a much higher priority.” (p. 4)

The memo laid out a string of arguments for why he reached that conclusion, but here is the gist. “In the course of the next couple of years,” it said, “our development organization will need access (and in some cases exclusive access) to technology which does not exist today. Although we are quite capable in creating products, our developers are not equipped for this task by themselves. This isn’t just for extra credit, in many cases it is needed just to fulfill the vision and commitment that we have already embarked upon. There is a lot of technology that we can and should count on getting from outside—either [from] university research or outside companies, but there are also some areas where we need to, or have an opportunity to, create new technology ourselves.” (p. 5)

Commentary: This multi-dimensional approach to R&D is a cornerstone of Myhrvold’s argument. Great corporations must watch the outside world and incorporate advances from it, they must do product iteration and small-scale, “safe” development, and they must take some risks on longer-range bets, some of which they do in house. A key quote (p. 5):

One of the key distinctions which has to be made is the difference between doing advanced development, product development and research. Many companies confuse these issues to their detriment. We believe that Microsoft needs to have a full spectrum of development.”

Great Failures Don’t Mean Research Is a Bad Idea

Myhrvold acknowledged some high-profile failures at other companies to capitalize on path-breaking research. “There are many examples of good research which does not benefit the company involved, often to extreme proportions,” he wrote. “Xerox is an example of a company that did the right stuff at the right time with the right vision, and still lost. They invented the GUI [graphical user interface], and yet it never did Xerox any good.” (p. 7)

But, he cautioned, “research is no different than any other corporate activity—there will always be some spectacular failures. Every aspect of business is mismanaged by somebody, and it is not at all surprising that research is among them. When people focus on the question of ‘why doesn’t corporate research work?’, and use examples like those mentioned above, they are almost always overlooking the fact that you could equally ‘prove’ that finance, marketing, advertising etc don’t work either.” (p. 7)

“The fact is that research does work at a lot of companies. When research fails it almost never is because of an intrinsic problem in research itself (i.e. the inability to think of something new). Instead, the research usually falls prey to problems that can be traced to general management issues—having the right goals, transitioning technology to company benefit etc.” (p. 8)

“Xerox’s failure to capitalize on PARC is certainly a shame, but from any financial or strategic standpoint it pales in comparison to the $1 billion (in 1970 dollars) they lost in the mainframe business. Even worse was Xerox’s strategic failures which cost them many billions of dollars worth of market share. The hard bottom line is clear—to kill your company, you need a bad product strategy, and to waste a lot of money you need a bad product group. The actual cost of research never amounts to much in that context. This is not an apology for doing a poor job at technology transfer, but one should keep things in perspective.” (p. 8)


Myhrvold identified three key advantages that research can provide a company: a time advantage in introducing new products and technologies; better access to strategic technology; and knowledge and intellectual property for the company. “You shouldn’t start research in an area unless there is a strong chance of getting a unique edge in one of these three ways.” (p. 9)

The Technology Transfer Problem

“Once you have created some great technology, there remains the problem of effectively transferring it to the development organization,” Myhrvold wrote. “Failure to do this effectively is a primary reason that research work is ineffective at many companies.” (p. 12)

Myhrvold laid out four imperatives or guidelines for overcoming this challenge:

High level strategic support is vital

Selecting the right research agenda is more than half the battle

“The largest single technology transfer problem is that the technology is off target and nobody wants it or needs it in their product. This is a very vital point – no amount of technology transition ‘process’ can help the wrong technology at the wrong time.”

Proper program management keeps the agenda relevant

“The process of tracking the rest of the world, and measuring research goals against our strategic needs is not just an up front thing, but has to be maintained throughout the process.” (p. 12)

Communication with product groups is essential

Commentary: Reading this, I can’t help but look at Microsoft today, and think of the broader product strategy problems around search, online advertising, smartphones, cloud computing, and more that now plague the company. As someone who has visited various MSR labs over the past two decades, I also can’t help but think of all the good ideas I have seen at MSR (and other parts of the company) around mobile phones, tablets, speech and other interfaces, sensors, and more, that were on the cutting edge or well ahead of the curve—but through some combination of bad market timing, poor product execution, or just failure to get buy-in from the right groups fell behind Apple, Google, and others. A lot of it can probably be chalked up to failing to avoid the very pitfalls Myhrvold recognized from the start, something not unique to Microsoft by a long shot.

Why Research Should Be Centralized and How It Should Be Structured and Funded

Myhrvold noted that others had argued that research expertise should be distributed throughout a company, but he rejected this idea for four reasons: 

The best researchers won’t come to work in a product group

You can’t create the right atmosphere for research in such business groups

Synergy between researchers would be impossible to attain

Product groups are not equipped to incorporate researchers working on different time horizons and with different goals than the rest of their staff

Much of the rest of the document is about the structure a research group should take. Myhrvold argued for a central director “similar to the chairman of a computer science department in a university,” (p. 14) who has both respected technical capabilities and management skills and who can recruit new researchers to the staff. (This would turn out to be Rick Rashid, recruited from Carnegie Mellon.)

He proposed several research groups, with 5-10 people in each group, “depending on the area, the scope of the projects, etc.” Overall, the lab would total 30 people in first year, 50 in the second, and 60 by the third. He felt that it had to be a minimum of 50 to have critical mass. And, he said, great people were knocking on the door: “We have already been contacted by research groups at HP Labs, IBM, and the IRIS [Institute for Research in Information and Scholarship, which was working on a “scholar’s” workstation] group at Brown who want to leave en masse.” (pp. 15-18)

Each of the research teams should be lead [sic] by world class people—there is really no excuse for settling for second best,” he wrote.

Commentary: Myhrvold laid out the recruiting plan, and his first bullet point was to hire Gordon Bell, the famed “father of the minicomputer” from his days at Digital Equipment Corp. as a key advisor to help him find the director. That did indeed happen, and Bell, now 79 and also an Xconomist, is researcher emeritus at MSR. After Rashid, the lab’s first hires were three key members of the linguistic and natural language group from IBM Research—Karen Jensen, Stephen Richardson, and George Heidorn.

Draft Research Agenda 

Myhrvold laid out a draft research agenda which was pretty spot on for the next five years—and longer. This agenda, he stressed, was not set in stone, but “to illustrate the kinds of projects that are envisioned.” (p. 19)

Information at your fingertips

The digital world

Creating the digital office and home of the future

“If these sound redundant, it is by design. They are just different ways of looking at the same thing—how the personal computer will evolve between now and the mid 1990s.” (p 19)

Closing Thought (for those who have read this far): After each item, Myhrvold wrote a paragraph laying out more of what he meant. You might have some fun reading them, and thinking about what you might say about a draft agenda for the next five years in computing. If you care to share your ideas, drop me a line at [email protected]








Bob is Xconomy's founder and chairman. You can email him at [email protected] Follow @bbuderi

Trending on Xconomy

By posting a comment, you agree to our terms and conditions.

7 responses to “The Memo That Spawned Microsoft Research—Analysis and Download”

  1. Blake says:

    Whaddya mean “he didn’t get the optical storage part right”?

    CD ROM drives were just starting to roll out and were still pretty expensive. A burner cost well over a kilobuck. The software distribution medium of the day was floppy disks (yes, I didn’t have a CD ROM drive and thus ordered Windows 95 developer preview that came as a shoebox with dozens of 3.5″ floppies instead).

    Sure we don’t use optical storage as our primary medium but if you look at DVD-RAM, blu-ray, etc etc. his comments about optical storage are just as relevant as the rest.

    • Robert Buderibbuderi says:

      Good point. I confess, I was thinking along the primary medium line.

    • Stuart Warren says:

      I suggested to a client that Optical disks were the way to go and they invested in Panasonic magneto optical disks and drives. Now no one uses that technology, and even Panasonic no longer supports drivers for their technology, leaving lots of data lost on this orphan technology.

      • Blake says:

        I’m not sure what your point is… There were certainly plenty of dud optical technologies, just like everything else. But basically every machine except ultrabooks now has a DVD/blu-ray drive in it, and in the early 1990s they certainly did not.

  2. guest1 says:

    Hmz and what did MS Research produce ? :) Apart from Kinect and comfortable shape of the mouse (intellimouse) i can’t think of any successful innovations done by MS in 30 years.
    …oh and there is aslo a grass that grows on the walls in their offices…

    • Milo Crotoni says:

      The idea, I think, was to have software that would be ready to utilize new hardware technologies:

      “The challenge, of course, is to create the software which realizes the potential that this hardware offers. We have to explore and research the new software technology necessary for the PC environment of the future….”

      This mean mean things like caching design, security, generic driver design … etc. I’d hardly compare that with the hardware research other companies did, though.

  3. Jörg Debus says:

    The focus was wrong (1991). And it is still wrong. They focused to exploit new technology for new functionality and new marketing stories. They totally overlooked the cost of ownership of their products. Deploying and using Windows and MS Office is until today more than two times more costly than the license fee customers pay. They still do not take care of the users/customers investment in data and knowhow and configuration when migrating to new versions. And with Windows 8 they devalued billions of dollars invested in home made applications of their business customers. Would be an great deal for both MS and the customers if they had helped saving this money and they would have had a chance to get some part of that as license revenue.