When Microsoft founder and chairman Bill Gates stepped down from daily involvement with the company this past summer, two Microsoft employees had been selected to fulfill Gates’s role as the company’s technological visionary. Ray Ozzie, who was interviewed by Knowledge at Wharton last year, inherited Gates’s former title of chief software architect. The other heir to the Gates legacy is Microsoft chief research and strategy officer Craig Mundie.
Microsoft grew out of one of the great transformations in the history of computing — the rise of the personal computer — and has been driven by Bill Gates’s vision of “a computer on every desk and in every home.” Although Mundie is only six years older than Gates and Ozzie, he is a veteran of several earlier shifts in computing technology. That experience may prove valuable as Microsoft faces the next disruption in computing: the shift away from individual PCs to networks of Internet-connected devices.
Mundie holds a bachelor’s degree in electrical engineering and master’s degree in information theory and computer science from Georgia Tech. While studying at Georgia Tech, he went to work for Systems Equipment Corp., which was later acquired by Data General, a company that arose from an earlier industry shift — the move from large, mainframe computers to smaller “minicomputers.” Tracy Kidder’s Soul of a New Machine immortalized the work of Data General during that period, focusing on the team in Massachusetts who raced to beat the North Carolina branch of the company where Mundie worked.
In 1982, Mundie, along with fellow Data General alumni Ron Gruner and Rich McAndrew, co-founded what would become Alliant Computer Systems to develop high-performance parallel computers for engineering and scientific applications. After several years of success, Alliant ran into problems. As former Alliant president Ron Gruner writes on his web site: “Just as minicomputers in the 1970s had undercut mainframes and our mini-supercomputers had undercut supercomputers, smaller computers — workstations from Sun and Apollo and even the desktop PC — were beginning to undercut us.” Alliant went out of business and filed for bankruptcy protection in 1992.
In 1992, Nathan Myhrvold, then Microsoft’s chief technology officer, recruited Mundie to spearhead Microsoft’s efforts to address another emerging shift in technology: the convergence of PCs and consumer entertainment products. As head of the Advanced Consumer Technology Group, Mundie focused on non-PC PC platforms such as Windows CE, the Pocket PC and early console-gaming products. Mundie also oversaw Microsoft’s acquisition of WebTV and championed the company’s early efforts in interactive television, including its video-on-demand experiments in the mid-1990s.
Mundie brings this knowledge of the vicissitudes of technological change to his work charting Microsoft’s path through the next technological transformation.
Knowledge at Wharton recently met with the 59-year-old Mundie in Microsoft’s corporate headquarters in Redmond, Washington, to discuss his vision of the future of computing and how he intends to keep Microsoft relevant through the next paradigm shift in computing. An edited version of that conversation follows.
Knowledge at Wharton: Now that Bill Gates has stepped down from daily involvement in the company, how will his responsibilities be split between you and Ray Ozzie?
Mundie: We actually made the changes two years ago, [but] now that Bill is actually gone, we have to take the last step — which is more an acculturation thing for the company to deal with.
How did we divide his job? In three dimensions: in/out, up/down, long/short. I’ll explain each of those.
In/out: For many years now I have spent a lot of time dealing with the geopolitics of Microsoft. Even when Bill was here, he and I spent a lot of our time talking about the company’s vision of the future. So I will be more outbound and Ray will be more inwardly focused from a communications point of view. Ray will talk whenever he wants to, obviously, but I’ll spend more time than he will outside.
The long/short is the time horizon. I took over running the global research operations. The things I do are in the 3- to 20-year horizon: starting new businesses, doing fundamental research. Their gestation cycle is longer. Research is centralized in the company, where the development activities are distributed within the business groups.
The third dimension, up/down, is within the technology stack. I start at physics and work up. Ray starts at the cloud [distributed Internet-based services] and works down.
Each of those dimensions is a very serendipitous division of labor, because our own styles and interests map to those three dimensions in a very natural way.
Knowledge at Wharton: You said your personalities fit the way the responsibilities are split. Can you explain that a bit more?
Mundie: For example, on the inside/outside dimension, I’ve now spent so much time dealing with government leaders, business leaders and other people outside the company, and giving many, many speeches and talks and demonstrations. It’s something that I enjoy doing and it’s natural. It’s something that Ray is good at but doesn’t do as much of and I don’t think [he] enjoys it to the same degree.
Within the technology stack, my interest tends to be closer to computer architecture and the lower-level issues. A lot of Ray’s experience was at the application layer, writing [Lotus] Notes or the other things. Today, as the Internet has evolved, he’s very involved in the service components of our business and integrating them across all the existing business units. So these are nice, natural partitions of the work.
Knowledge at Wharton: You oversee Microsoft’s research group. What are two or three Microsoft research projects that are particularly interesting?
Mundie: One that we’ve been working on for quite a long time relates to the underlying shift in the architecture of the microprocessor. The physics have forced the microprocessor folks to shift into building machines with multiple cores in order to get more performance. They can’t continue to increase the clock rate [the speed of the computer’s microprocessor]. That represents one of the more fundamental shifts for our industry in 40 years. We have a lot of research dealing with making parallel programming easier, more reliable and ultimately something that many, many more people will have to confront.
We have a lot of work going on in what I call this “natural user interface” concept, which is beyond GUI [“graphical user interface” — pronounced “gooey”] to NUI [pronounced “newie”]. People’s ability to interact with computing will be less and less about interacting with a computer. There will be certain [tasks] that we do with a traditional [device] that looks like a computer. But your cell phone, the walls and surfaces of rooms, the ability to talk and speak with the computer — that all is coming fairly quickly now. The advent in the next five years of chips that are wildly more powerful than the ones we have today is going to accelerate this shift into a more humanistic relationship between the computer and people.
[In] other areas… we’ve got a group in Bangalore focused on developing technology on a global basis for people at the base of the pyramid in a demographic sense. That represents an interesting opportunity…to bring health, education, more jobs and productivity to five billion more people. I’m quite enthused about that work, too.
Knowledge at Wharton: Microsoft is a big company; you’re undertaking all these initiatives. But one of the critiques one hears about the company is that focus is an issue. With Bill stepping down, to what extent is keeping focused a challenge for the company?
Mundie: First, I don’t think the issue of whether Bill’s here or not changes the basic nature of the company relative to that issue. A lot of this boils down to: How does the company stay vital in the long term and what is the business we’re in?
In the last decade, we have had to get Microsoft to realize it’s not a PC software company per se. At the size we are and with an aspiration to continue to grow and be economically vibrant in the future, diversity is going to matter more and more.
In the days gone by, people said, “I know what Microsoft did.” It was the objective Bill set out: to put a PC on every desk and in every home. Well, we did a pretty good job of that, but only for a billion and a half people on the planet. There are another five billion to go. If you look across the planet, most people’s first computer is unlikely to be a PC in the future. It’s more likely to be a phone or their television than a traditional personal computing device.
As we sought to say — “Where is software going to matter?” and “How does it actually have to get integrated to make a difference?” — we have been moving to diversify the software into these other environments.
Some would call it a lack of focus. But, my view has been — and I guess I helped set the strategy more than a decade ago — that eventually the cacophony that would ensue when all the devices were smart and connected — but had no relationship to one another — would create a demand for a company that would be able … to integrate these things in some useful way. We’re really just at the tip of that iceberg.
A lot of what we’ve done in terms of cloud services and what we’ve done to put our software into phones, cars, game consoles, televisions, watches and everything else, make us the only company today that has any chance to bring these things together in some unifying way. Some people will take it, some won’t. But I think that that’s important.
The other thing to think about is that we operate in a high flux environment. There are very few businesses as big and as global as this one, whose technology base and products shift as fast as this one. It’s impossible to accurately predict the future and it’s certainly not possible to comfortably rest on the laurels of the product lines that you have and believe that 10 or 20 years from now you’re going to be relevant.
Our nature is to explore many things and be prepared to disrupt in some new areas, to continue to enhance the ones that we’ve got and then to respond to opportunities that emerge simply because of other people’s creativity.
Knowledge at Wharton: Looking at two of those disruptive areas: When Steve Ballmer spoke at the Wharton School, he stated that he saw Microsoft as having two major competitors: ad-funded software and open source software. Would you agree?
Mundie: If you want to focus narrowly on saying, as perhaps we should, “We’re just a software company and we intend to make a business out of software” — then I think his statements are true. It’s obviously a little bit more complicated than that. But, at a conceptual level, the idea that software is not an intellectual property asset is something that we do not agree with. We spend billions of dollars a year to make [software]. Most of the established countries in the world accord it intellectual property status and we work hard to develop patents and uniqueness.
There is a movement in the open source world which has tried to move away from saying “having intellectual commons is good” to saying “having intellectual property on [software] is bad.” That is where we have tried to make a strong distinction. [We] support the idea that people need to have an intellectual commons and to share things for learning — but we tend to fight back on neutering intellectual property in software as a concept.
Ad-funded software is really ad-funded products. We don’t believe that most businesses are going to end up showing ads to their employees as a way to pay for the productivity software that’s integral to their business. In the consumer space, there are many ad-funded things that are important. It’s one of the reasons the company has said we are committed to being a major player in the advertising business.
There are only three ways to get paid for this stuff: transactions, subscriptions and advertising. Microsoft started life in the transactions model: You paid us once, you got a right [to] quiet enjoyment of that version of the software until we could offer you something better that would induce you to buy the next version.
Over time, particularly in enterprises, [our customers] said, “We really just want all of the latest stuff, so let’s have a subscription model where we’ll pay you a fixed amount of money.” The vast majority of our big customers around the world operate under that model.
The thing we never really had is this third leg of the payment stool, which is the ability for people to use the software based on advertiser supported compensation. This doesn’t make it free to us, it just makes it free to the consumer. Somebody is paying the bill on their behalf. We think that’s a good thing, and as consumer-based computing has become more a part of the business, it’s natural for us to take that on, too. That’s a challenge, just being able to have a prominent role in that ecosystem.
Knowledge at Wharton: One of the big drivers of ad-funded software is search. Despite Microsoft’s enormous resources, the company seems to have a hard time taking on Google in search and ad-funded search products. Why is this so difficult?
Mundie: There are not very many businesses. We happen to have a few of them ourselves. Google stumbled into one in the combination of the technically challenging problem of very large scale search coupled to a scale economic business — the advertising model.
Once you get to scale, the economics are very favorable. Even if you had perfectly equivalent technology, if you start late — which we admittedly did on the advertising side — you have to work hard to get back to scale. You’re always at a disadvantage. No matter how much money you have, you have to build up to scale. It just takes time.
Think about this like Pepsi versus Coke. Despite all of the [resources] of PepsiCo, why did it take them so long to get to a significant competitive position in the market relative to Coke? It’s one of these same effects. People develop habits, they order Coke; Coke is ever present.
You have to realize that you can’t keep playing the game on the terms of the person who established it. But it doesn’t do any good to say you want to play the game on a different basis until you’ve reached at least parity on the fundamental feature of [the competitive] product. In Pepsi versus Coke, remember the infamous taste test? It doesn’t do any good to hold the taste test if you don’t think you taste as good. We’ve been spending the last few years getting our search to taste as good. So if you tried it, you might like it. Then the question becomes not “Would you like it?” It’s “Would you try it?”
You’ll see in the next year or two that the company will continue to innovate in elements of the business model, the technical approach and user experiences around search and advertising. From those we’ll begin to offer people not only a product that tastes as good but has some incentive to try it.
We recognize the role of scale in this. One of the reasons we pursued Yahoo was as an accelerant for scale. We don’t view it as essential, but it certainly would have been a convenient way to spend money in order to get to scale more quickly. But, absent that, we will continue to go down the path that we’ve outlined.
You can see the first example of this with the [Live Search] “Cashback” offer. It changes the dynamics between the consumer, the advertiser and the person running the advertising system from an economic point of view. Early returns are encouraging. But we’ll see what it takes to shift this market.
Knowledge at Wharton: Back during the portal wars, iWon.com used a similar ploy — offering a daily cash award to a randomly selected person [who used its portal]. That didn’t seem to work. How is Live Search Cashback any different?
Mundie: I don’t think a lottery has the same appeal as something that affects every single person who makes the decision to use search for commercial purposes. I would have [characterized what] you described as a marketing ploy — as opposed to something to change the fundamental business model of portals. I wouldn’t equate the two approaches at all.
Knowledge at Wharton: The reception of Vista has been rather mixed. Microsoft is now talking about some of the features that they will be introducing in Windows 7. Are we at the end of the line for the large monolithic operating system?
Mundie: If you think about the Internet as just a big distributed computing machine — then you could say the trend is toward more distributed asynchronous concurrent types of programming.
[Whether] you’re doing a web app that’s pulling stuff from multiple servers in the cloud or writing an application using multiple computing and storage facilities within a single machine, more and more — latency aside — those are similar problems.
The heritage of operating systems was managing the resources of a single machine and providing facilities for the serial execution of a program. Programs are moving into this world where they get performance through parallelism and they get functionality through distributed concurrent services.
The power of the machines will continue to encourage more and more complicated software systems. But complexity is bred historically through the scale of these software systems. The thing that’s fragile — and to some extent Vista has suffered from this — is as you increase the scale of software built using traditional methods, you increase the complexity.
Many of the failings — not just of our software but of all large software — is that the security problems, the lack of reliability, the difficulty in maintenance, the difficulty in testing, all of these things are symptomatic of software still being too much of an art form and too little of an engineering discipline. I believe that over the next 10 to 20 years, you’re going to see a dramatic shift in the way people write software.
I don’t think describing it as “monolithic” or “not monolithic” will be the question anymore. Software will be built through the composition at every scale of a lot of distributed, asynchronous services. The question is: How can you specify, compose and operate those services? The operating system will itself adopt some of those models — although its role in providing an abstraction of the hardware capabilities means that it will be slightly different in that aspect. But I think that will happen.
The challenge we have is how to move the world — that has a lot of investment in the technology we’ve got — gracefully into an environment where we can introduce these new capabilities and shift the underlying platform and tools in a way that promotes the construction of these new types of applications.
One of the areas I’m personally a big supporter of, because I think it’s a precursor to this, is the new business we started about three years ago in Robotics. We have a robotics toolkit that we’ve been shipping for about a year and a half now. When people look at it, they realize this isn’t really just about robots. It’s about distributed concurrent programming models. And, in effect, it has a visual programming language.
Many of the things that the programming community has suffered with — as software has become larger and more complex — will be remediated by moving to these new techniques. But there’s a huge amount of inertia in the system, a huge amount of investment.
We’re in a very long-term shift to the next computing platform. This happens in our industry about every 15 to 20 years, like a clock. The question is: Will we or will somebody else be the ones to navigate to this new platform model? I think that we are well prepared for that change.
Knowledge at Wharton: One of those big evolutionary changes that happened in the mid-1970s is one that helped to build this company: the change from mainframe and mini-computers to personal computers.
Mundie: Since 1980, we’ve been through about two and a half of these cycles. My view is that each new platform cycle has two phases. The first I call the diffusion phase, where something has to force the diffusion of the new platform into the environment. Once it’s there, it becomes a target for the world’s programmers to move up to some new use of the platform and to create a lot of applications for it.
The PC platform was established by two killer apps: spreadsheets and word processors. They had enough appeal that individuals went out and bought a personal computer just to get access to those tools. That was the birth of the personal computing industry.
The next platform shift came along with the Internet as another place you could target programs. It didn’t obviate what was happening at the desktop. There were two killer apps for that [phase]: email clients and web browsers.
Once you establish a platform, the second phase is the exploitation phase. We’ve had literally hundreds of millions of people exploiting the PC platform. There’s a huge community now exploiting what you can do on the cloud side. We’re in the exploitation half of the Internet, which shows because it’s getting more diverse in terms of people programming for it and using it.
Where that ends up is what I call the “cloud plus client” model. If you look back a little bit, people said, “Hey, forget that software stuff, it’s just software as a service.” That’s pretty well gone now.
Knowledge at Wharton: SalesForce.com’s phone number is still 1-800-NO-SOFTWARE.
Mundie: Yeah, and people aren’t calling it that often either.
[Despite] the rhetoric, any of the really compelling Internet-based applications are now being built by software that runs in the client — whether its browser-based or real local execution. Take Virtual Earth, Google Earth, the Robot Telescope or any of these truly compelling experiences — more and more you’re running some sophisticated client and it’s being fed by a data or orchestration service in the cloud. This “cloud plus client,” which we generally call “software plus services,” is going to be shown in every dimension — economics, the experience of the user and other things — to be a much more durable use of this composite platform.
What you have to ask is: What is the next really big shift in the platform? I’ve named it “fully productive computing” — where there’s a lot more use of the computing cycles that are in the system. Take your laptop, your desktop, your cell phone — if you measured the utilization of the computer [inside these devices], it’s very low because it’s only been designed to respond to an input from you.
If you realize there are a lot of latent cycles that are unused in the devices and then you say, “I’m going to make those things go up by a factor of 100 in the next decade” — there’s no way you’re going to leave that lying fallow. The programmers of the world will do something with it.
The question is: What do you do when you make it a hundred times more powerful? We have a lot of interesting ideas now about how to change the way people deal with computing and the kinds of things that it can do. That’s where I’m aiming in the future: the realization that the way in which you build and design the computers that underpin both the client and the cloud are going to go through a big transformation.
So it doesn’t matter whether you’re a cloud guy or a client guy. Your future is going to require the absorption of a new model of computation and a new model of writing programs.
Knowledge at Wharton: You identified the two killer apps for the first two phases. What are the two killer apps for this next phase?
Mundie: Nobody knows yet.
I contend that no one knows what they are in any of those phases until they emerge. Platform evolution is a bit like cooking a primordial soup. You keep adding things into the pot and cooking it, and eventually something pops out due to ingenuity, and you reach a critical mass. If there’s not enough change and if it isn’t compelling enough, then the hysteresis [inertia] is so high that the world just stays on the platform it already knows and loves.
Only through investment in basic research and the development of these new technologies by the whole industry, not just one company, and then pouring smart people into that equation, does this thing pop out. I don’t think anybody can predict exactly when it happens or exactly what the killer apps are going to be. I can tell you that if you look back at the pattern, it’s very, very consistent.
So you could say that we’re doing all the foundational work in our company to make the contributions to the research and development line. We’re trying to hire as many smart people as we can and we’re on the quest for the next killer apps. Hopefully we’ll find them.
Knowledge at Wharton: You spoke about your “software plus services” model and the richness you get with a client app connected to the cloud. In Microsoft’s paradigm, those are traditional clients written to the operating system. Adobe Systems has a product, AIR, that’s attempting to build a new abstraction layer above the operating system to bring that richness to cross-platform applications. What’s your take on that approach?
Mundie: This is not a comment about Adobe AIR per se, but in general: When you approach a platform inflection point, everybody thinks they know enough about the underlying platform to just abstract it into something that commoditizes it. That’s a natural tendency, but that doesn’t work when the new platform emerges underneath it, because it always emerges as a singularity. It doesn’t emerge simultaneously from many quarters all at once.
Our view is: We have to keep evolving the platform. If we can’t add value with a correspondent set of applications that take advantage of it — if it, in fact, is just homogeneous — then yeah, you’ll get AIR and [Google’s] Gears and any other thing that you want.
When Mark Andreessen was at Netscape, [he made] his infamous quote that “Windows is just a poorly debugged set of device drivers. What people really want is just this browser thing.” That hasn’t really worked. It’s very important to look at what it is that people want to do. Are you doing it against an evolving underlying platform or a static one? Because the cycles are so long, many people tend to think, “Hey, it’s just maturing. We know how to commoditize it” — whether it’s with free software, Gears, AIR or whatever.
Java was another one of these “write once, run everywhere” things. It just doesn’t work, because people don’t actually want the desktop on their phone. One is a fixed or portable experience and the other is a mobile experience. Your fingers don’t get small and pointy, you know? You can’t shrink the icons down to be tiny enough to make them work.
People pick up on [concepts like] “software as a service” or “write once, run everywhere,” and [yet] the diversity of the application, the computing platform and experience people want, fly in the face of these simplistic views of how it’s all going to collapse down into one convenient model [that will] stay there for the rest of your life.
Knowledge at Wharton: What’s the biggest challenge facing Microsoft right now?
Mundie: We have many challenges. It’s hard to pick which one is the biggest…. We’re trying to mature our products in all of these other computing devices that we’ve been investing in for five or ten years. It’s a high flux environment. Many of these are regulated industries. Your ability to prevail is not merely a function of having great technology or even a lot of money. The company has been one that perseveres over long periods of time in the things that we think are strategically important, whether it’s game machines, cell phones, the desktop, the cloud services, the advertising business — there are five big business battles that we’re in, and I think doing reasonably well at.
In the late 1990s, when we put out the first of our Windows CE handheld devices, the popular product at the time was the Palm Pilot. I remember announcing this stuff, and people almost ridiculing the company and even me personally, saying, “You guys just don’t get it. Why are you building this operating system? Why are you supporting color screens? Palm Pilots — it’s all anybody ever wants. Four buttons, PIM [personal information management software] — man, life is good.”
And we said, “No, no, you don’t get it. People will not ultimately be happy with that. It’ll have to be integrated into other things. They won’t be satisfied with a 2×2 black-and-white low res screen, and oh, by the way, it should merge with your cell phone.” You look back a decade later, and of course, everyone’s forgotten that. In the smart phone category, we’re arguably first or second by volume.
Knowledge at Wharton: Palm has even licensed your mobile operating system.
Mundie: Right. If you look back over a long period of time, we persist…. Many of the things that we set out to do are important at the time they become really big.
That’s one of the greatest challenges the company has right now: That we care about these things when they get big, but we’re held to a standard — for what we produce in terms of reliability, supportability and maintainability — that Joe Average Company or any start-up from a garage never has to worry about.
It is frustrating when you do the research and start into these things, and then everybody says, “Oh, look! These kids have done this thing. Why didn’t you do it?”
In part, we work on these things, and by the time it goes from being a very tiny business to something that’s actually material, well, we’re usually in a pretty good place.
It’s frustrating for the people at the company to be perceived as, “Well, you guys don’t ever innovate.” The amount of invention that comes out of the company is staggering. It’s just that our job is to make [sure it’s] packaged, integrated and delivered on a continuous basis. Therefore, we tend to lose the excitement that people want to accord to the guy in the garage.
That is one of the challenges we face. We’re over 90,000 people and operate in more than 100 countries. Everybody looks at Microsoft and says, “I know what Microsoft is. I use Word or Office or Windows,” or whatever it is. They have a fairly localized view.
When you try to run the company, make the products and deal with releasing simultaneously in 38 languages with means to support an unlimited number of languages — these are things most people have no appreciation for. Many of the things that we get chided for — “Why don’t you do these things?” or “Why did somebody else beat you?” — well, they launch in one country, in one language, and they don’t have to deal with the scaling issues.
Many of the things I worry about are not even the technological ones. As we continue to scale up and try to operate globally and deal with the exigencies of the world — those, frankly, are sometimes more challenging than just working on the software.