”We are the Windows company, after all,” Microsoft CEO Satya Nadella told me.
I was at Microsoft’s campus in Redmond a week before the Build developer conference, and I wanted to know what was going on with Windows after a reorg split the team into different divisions. Was Microsoft really preparing itself for a world without Windows? Nadella was ready to tell me that Windows isn’t going away — of course Windows isn’t going away — but he also wanted to explain his latest buzzwordy vision for the future of the Microsoft: AI, Intelligent Cloud, and Intelligent Edge.
Windows might still be here, but after talking to Nadella, I did get the sense that Windows is no longer as central to the company’s future plans as it once was. Instead of trying to make everything run on Windows (as his predecessor Steve Ballmer was trying to do), Nadella wants to ensure that everything can work with Windows.
It’s a big shift following Nadella’s rise to CEO in 2014. Nadella’s Microsoft is fighting fewer losing battles, and instead, it’s trying to be just a little more open. New features in Windows 10 — like a timeline that works across multiple devices and a “Your Phone” system that lets users text and access phone data from their desktop — show a measure of pragmatism, if not humility. Of course, openness has its limits: Microsoft is still aggressively pushing users toward its Cortana assistant and Edge browser in Windows.
Microsoft is a different kind of company under Nadella than it was just a few years ago. It’s working to get its software in IoT devices like drones, support companies with cloud services, and improve its enterprise software. Retail consumers aren’t exactly an afterthought, but Microsoft is starting to look more like IBM than Apple.
The demos that Microsoft will be showing onstage at its Build developer conference are a good example of this shift. I got an early look at a few of them in Microsoft’s “Batcave,” a room filled with a tangle of cables, laptops, drones, and wildly expensive Surface Hub displays. It’s located in the same building as Nadella and the other executives, and it’s where new tech is tested and honed into something that could be shown onstage — and eventually (hopefully) shipped to businesses and consumers.
In one demo, a DJI drone had Microsoft software loaded on it so it could recognize faults in an oil pipeline without the need for an internet connection. An off-the-shelf consumer drone can stream video to a Windows laptop to do the same. In both cases, the idea is that Microsoft thinks its customers will be better off if they happen to have Windows around to augment the gear they’re already using. Windows is important, but it’s not the only thing.
But if you are entirely in a Microsoft corporate setup, the company is hinting at big things ahead. It’s integrating Cortana into its Slack competitor Teams. Cortana can be “invited” into your chats and automatically suggest stuff based on your conversation, like a meeting or document you need.
I also sat in on a demo of a futuristic meeting: a speaker topped with a 360-degree camera recognized me and greeted me as I walked into the room. As we spoke, it transcribed our conversation live using far-field microphones. (Microsoft also says it could also translate between different languages live.) If one of us said something like “I will follow up with you on that,” a separate box logged an action item, correctly assigning the right person with the context of the conversation.
Of course, HoloLens was involved. Microsoft is researching ways to make its augmented reality headset more useful, and visualizing data in 3D is a good use case. As one person used the HoloLens to manipulate a 3D map of a building’s temperature sensors, I was able to see it on the screen in real time. We were getting the same data displayed in multiple different ways: in AR on the HoloLens, on the display, and remotely to somebody else using a Surface tablet.
The demo was just a bunch of existing Microsoft products working well together in a surprising way. And there’s no better metaphor for Nadella’s Microsoft: instead of naïvely hoping it will take over everything, it’s just working on the stuff it’s good at. And instead of promising a future that we know isn’t going to happen anytime soon, it’s building on what it has now.
That’s a different kind of Microsoft than what we’re used to thinking of. It’s a little less flashy, yes, but it has the benefit of being a lot more likely to succeed.
This interview has been edited for clarity and length.
I just reread your book, and I’m curious about your effort to remake the culture of Microsoft. As you said, the C in CEO stands for culture, and I’m wondering how that’s going.
One of the things that I think about for any company, including ours, is that you really need to anchor yourself on that sense of purpose and identity, a culture that allows you to express that identity and sense of purpose with changing technology.
In fact, [the Build conference] is actually a good place to start, because, if you think about it, our sense of identity is all in the story of our creation itself. We first were a tools company. Bill [Gates] and Paul [Allen] created the BASIC interpreter for the Altair. So that means in the very genesis of this company, it’s creating technology so that others can create technology.
This notion of empowerment of people and organizations is very ingrained in our mission. And the best place to start that is by talking about developers.
Then, on the culture side, you need a culture that allows you to learn and to be open to new ideas. Otherwise, it’s lethal in this business. If you think about tech and tech transitions, [you need to have] what we describe as this “growth mindset.” You have to really confront your fixed mindset.
The feedback cycle from the developers to platform companies is probably a great place for us to keep renewing our culture.
So what sorts of fixed mindsets does that feedback loop help you break? What sorts of changes are happening within the culture in that way?
We built operating systems all our life, but what is an operating system? In a world where every person is going to use multiple devices in their life, they’re going to collaborate with many people in their family or at work.
What does it even mean to build a platform? […] Now, we need to reconceptualize and build something for the person, not something that is about the device. So that’s what Microsoft 365 is all about, [to] allow anyone to have an experience that’s multidevice.
We are the Windows company, after all. Windows and multi-device experiences are not at odds. In fact, “let’s make every Windows application a multi-device application” is something that we’re going to talk a lot about [at Build].
Similarly, take Azure. We’re having amazing momentum. We’ll talk about customers using it. Azure is available across 50 regions. Having started in that group, it’s stunning to see the momentum and the scale.
But guess what? Right as the cloud has become mainstream, the edge of the cloud is where the action is. In every factory and every hospital, you now need edge computing.
And so you’ll hear us talk about everything from the microcontroller that runs Azure Sphere to Azure Stack, to the DJI drone, to the Qualcomm camera. These are the things that are exciting.
There’s this swing between the cloud and the edge, but it’s basically local computing. Do you think of it that way? Do you think of it as a pendulum?
[It’s fair to say] it’s swinging back, but it’s a very different form of swing-back. It’s not like, “Okay, let’s throw back to client-server computing.” This is very different event-driven, distributed computing, which will support the needs of the experiences.
When you have an autonomous drone with the ability to see things, the amount of data that’s being generated, the amount of compute power that needs to be local, and how those events are processed, how inference happens, it’s a very different compute architecture. [It’s] not something that we did when you had either a phone or a PC connected to a server.
So yes, it is swinging back, but it’s swinging back because the insatiable need for more compute and more data at the edge is what’s driving it.
You just did a reorganization of Windows, and your memo did a really good job of explaining it from a Microsoft insider context. I think it really explained it to your staff and your team really well. But can you simply explain to an average consumer who walks into a Microsoft Store to buy a Surface what’s coming next for Windows?
Take the latest update to Windows we just did. It’s all about being able to recognize that every Windows user also happens to have a phone as well. So that means they already have multiple devices. They have a Windows device and perhaps a non-Windows device. How do we make sure that both those devices can work in concert to help the user get the most out of their computers? For example, the Timeline feature is a fantastic feature for continuity between devices.
So, from a Windows user perspective, we want to continue to innovate in the computers we build like the Surface Book or the Surface or the HoloLens, give it new features like the Timeline or My Phone, and then on top of it, also make sure that we recognize the fact that they have multiple devices and that there’s task completion and task continuity for the user across all of these devices.
That shift sounds like you are planning to not have Windows be as central to Microsoft. We were talking earlier about what an operating system is. It might even be a reasonable thing to ask: What is Windows? What is Windows in a year or two? Is it still how I think of it?
If you look at the capability that we have — whether it’s from innovation in the silicon itself to the natural user interface that depends on some silicon innovation to all the way to the cloud — we now have more of that capability than we ever had.
So what is conceptually Windows: it’s always about managing a bunch of hardware resources, whether on the server or on the client, and creating an application model on top of it. We have that now a lot more than ever.
The thing that we also want to recognize is just like on the cloud side. We did not say we’re building Azure only for Windows Server, but we’re going to welcome Linux as first class, Java as first class.
When it comes to the users, we want to recognize that the only thing that will matter for Microsoft is us serving our users well across all of their devices.
So this is not about taking away anything from Windows. If it’s anything, it’s adding to Windows the capability to span more devices.
And that’s why we talk about Microsoft 365, so we’re not confused that our job is to serve our customers who have multiple devices.
You wrote in your book that you see the world as going from a Windows runtime to a web runtime to an AI runtime, and that people are going to be building on top of it. But what exactly does that mean? What kind of stuff is being built on that?
Xiaomi, for example, built a device using our cognitive capabilities, both speech recognition and machine translation. Chinese travelers worldwide are going to be the largest population of travelers. Just imagine if they could have a far-field device that’s just there in their pocket. And they can just put it in front, and the two of us could be conversing, one in Mandarin and the other in English, and in real time have language translation. That’s a real application of AI.
The DJI drone where you now have a drone that’s flying over oil pipelines. We have a machine learning model that can recognize any breaks in the pipeline that’s being downloaded to run on the DJI drone itself. That means Azure Edge is in the DJI drone.
Or Uber using our face recognition for their driver identification, Twitter using our AI capabilities to do machine translation. These are all practical applications.
AI has gone sideways for a bunch of companies. We all know how badly it went for Facebook with the last election. Uber does not have a great history of treating its customers well. What sort of responsibility do you have to build protections into the AI capabilities you’re providing to developers so that it won’t cause completely insane sideways effects?
I would say we have a very broad responsibility as tech companies and tech platform companies, in particular.
I think, when I look at it, there are three issues on which you have to have strong principles guiding you. One is on privacy. Fundamentally, where the world is going, you’ve got to recognize that privacy is a human right and that you need to treat it as such.
Second, you have to treat security and cybersecurity as important for everyone. And third, you have to not just ask, “What can computers do?” But you should ask, “What should computers do?” So those are the three issues.
For example, take privacy. In our case, [we embraced] some of the new regulation. Take GDPR. Let’s make sure that we’re doing all our work so that we’re ensuring GDPR compliance. And this is the start of our journey, but we take that very seriously.
Or even what we did with the Cloud Act, right? Cloud Act is a breakthrough because it creates, for the first time, a framework of law that governs public safety and privacy. And I hope that the world comes to an equilibrium around it.
Similarly, when it comes to security or cybersecurity, some of the work that we led, which is the Tech Accord, is making sure that tech companies will come together … ensuring that nobody attacks anybody, and especially civilians and small businesses don’t get impacted. That’s a huge breakthrough. [Editor’s note: The Tech Accord includes a pledge to not help governments — including the United States — “launch cyberattacks against innocent citizens and enterprises.”]
On ethics, I think that there’s a lot that we need to do, but there’s a choice. Long before regulation, we as companies, we as designers of AI, should have a set of ethical principles. And so we’ve created even an ethics board internally.
It’s called AETHER?
That’s right. It’s a diverse group of people guiding design choices so that we do not run afoul of any unintended consequences.
When you start allowing people to build on top of the AI, how much responsibility falls on you to limit their capabilities, and how much of it do you think is just encouraging people to not be terrible?
Having been in the platform business for all these years at Microsoft, I think the key is for us to be able to guide people to make those design choices.
I mean, take good UI, right? We’ve always had books about what it means to create good UI. I think we all now have to start reading about what it means to create good AI. And good AI is not just the technology frameworks. It is also the ethical principles that guide good AI, and I think that’s where we have to start.