Listen to this essay as a podcast:
Did you ever have a box of cords in your house when you grew up? You know what I’m talking about: it was some plastic bin of full of cables and connectors for old gadgets, most of which you probably hadn’t seen in a while — but, in the event one of them should suddenly reappear, you would be well-equipped to power it up. Cameras. Video game consoles. Pagers. Computers. Or maybe you did save some of the devices, for future use. Right now, this box is probably still at the house, stowed away in the basement, or a closet, or under a bed, or in the garage under some sneakers and tennis rackets. If you’re anything like me, you don’t want to throw it out, no matter what mom says. I could really use one of those one day, I’m telling you! To throw out the old cords or, God forbid, the devices themselves, would be like emptying a bookshelf into a bonfire. Isn’t there enough room in the garage for one box of old dreams to slumber for a while longer, waiting for its Prince Charming? Maybe, that prince will be one of your future grandchildren, to whom you’ll introduce the ancient technology of the DVI cable or the Blackberry on a rainy day. Or maybe, much deeper in the future, a lucky archaeologist will find your box and marvel at the treasures inside, like Howard Carter peering into Tutankhamen’s tomb by candlelight. Cracked screens and twisted wires will glitter like golden treasures from another age. What story will your box tell? Here’s one…
✺
By personality and style, the three chief executives of the Microsoft Corporation have almost nothing in common. There was Bill Gates, the geeky but competitive founder from Seattle who made his hobby into the world’s most valuable business, and himself the world’s richest man at 40. There was Steve Ballmer, his brash Harvard classmate turned salesman-extraordinaire who sweated through countless shirts cheering the business through a long slump. And now there is Satya Nadella, the soft-spoken engineer and cricketer from Hyderabad who transformed Microsoft and took it back to the top of the world. Standing next to one another, the three men seem more like a “walk into a bar” joke than the leaders of the same company. But one thing they do share is having each expressed public regret to shareholders for the same reason: phones — or more precisely, Microsoft’s mismanagement of them. Nobody has a smartphone running Windows today, and for that they are sorry.
Correction: almost nobody has a Windows smartphone. Some people were so fond of the Windows Phone (2010-2017) with its colorful scrolling tiles and widgets, its neon bezels, that they still use them all these years later, long after updates and support have ceased, waiting for the batteries to finally die out. Rage, rage against the dying of the light!
Another sorry group are those who helped bring them into existence at Microsoft. Some of them were there at the very beginning of the giant’s foray into the new world of the mobile phone, which began long before shiny touchscreens and mobile Internet connectivity. A man called Yadhu Gopalan was one of them. And the box in his garage tells the story from the start, with every dusty old device its own chapter. For twenty years, the lifelong engineer kept the phones and other gadgets he helped create at Microsoft. His long career filled the box. And on a rainy day in March, Gopalan brought the box from his Bellevue garage the offices of his startup, Esper, where I met him. On a long conference table, we emptied out the box, one device at a time, and spoke about what went wrong, what went right, and everything in between. The story came to life. It’s not just the story of a tinkerer who wouldn’t throw out the box in the garage. It’s the story of how a great American company fumbled the ball – and then picked up a different ball to win the game anyway (after Microsoft exited the phone business, the explosion in its cloud operations brought the company back to the top of the world).
Microsoft, the company that had understood and capitalized on the personal computer wave to become the most valuable business on Earth, failed dramatically to capitalize on the mobile computer (or mobile phone) wave. Bill Gates thinks it cost Microsoft hundreds of billions – at least. But of course, the story isn’t all failure. Millions of people adored the Windows Phone and were sorry to see it go. It felt, so many, ahead of its time. But with marginal sales by the final years in 2016 and 2017, Microsoft pulled the plug. The End.
Now, to the beginning:
Once upon a time in 1994, Yadhu Gopalan was a PhD student in computer engineering at Auburn University. He’d moved to the United States as a teenager and studied electrical engineering at the University of Maine in sleepy Orono, population 10,000, a hundred miles from New Brunswick. His parents studied for doctorates while he was an undergraduate. As a PhD student in Alabama, he dreamed of working for NASA. But NASA in the 1990s was a place, Gopalan learned, where engineers played poker at lunch. It would have been too slow. He wanted whatever the opposite of that was. He got the job at Microsoft, dropped out of the PhD, and moved to the Seattle area.
For a young engineer in 1994, there were few jobs more exciting than Microsoft. The company had gone public in 1986 and its stock had run up nearly a hundredfold in eight years, with plenty more to go before the millennium was done. It was an ascendant empire of nerds. Satya Nadella, the current CEO, had joined the company as an engineer in 1992, just two years before Gopalan. Those were fun years at the Redmond campus. The unveiling of Windows 95 in July 1995 has been seared into the Internet’s memory (even those who never used it, or in your author’s case, weren’t even born yet!) thanks to the video of Gates, Ballmer and others dancing to the tune of “Start Me Up” at the highly anticipated launch. They paid the Rolling Stones $3 million to use the song as part of a larger advertising campaign to sell Windows to the masses. It worked.
In 1980, Gates had written a corporate mission for Microsoft: “A computer on every desk and in every home.” Windows 95 helped make that mission manifest. In twelve months, at a retail price of $209.95 ($109.95 if you already had a version of Windows) Microsoft sold forty million copies of the software — ten times the sales of Windows 3.0 in the same period five years earlier. Its intuitive graphical user interface and upgraded specs (32 bits!) were a revolution for both consumers and developers. Features that were first introduced in Windows 95 are now synonymous with Windows itself, like the Start Menu and ‘Close’ button. For millions of people, Windows 95 was also their introduction to the World Wide Web via the pre-installed Microsoft Network. Consumers loved Windows 95 so much that it remained the most popular operating system into the early 2000s, past multiple new Windows releases like Windows 98.
Soon after he joined, Gopalan got to work on a shadowy project called “Pegasus,” the codename for what would become Windows CE. Pegasus was part of a bold plan to take Windows beyond computers on desks, and to put it in your hand, in your car, and everywhere else it could fit. As the first mobile version of Windows, Windows CE was the patriarch of half a dozen other systems and devices like Handheld PC, Pocket PC, Windows Automotive, Portable Media Center, the Zune, Windows Mobile, and finally, the Windows Phone. Not one of those products survives today. Windows CE was where it all began, and it’s where Gopalan’s career began, too. For a decade, he was a software design engineer, contributing to the architecture of the entire CE platform and the many devices it supported.
Pegasus was released to beta in 1995, and in 1996 was released to manufacturing as Windows CE 1.0. Six manufacturers participated and released devices in 1996 and 1997: Casio, NEC, Compaq, HP, LG, and Philips. When Gopalan and I met at his Bellevue office, the first device he showed me from the box was a handheld PC made by NEC. The interface looks… like Windows! It still runs Pocket Word, Pocket Excel, and Pocket Internet Explorer. It has a stylus and a monochrome touchscreen. It has marginal storage but great character. It was the first device he helped bring to life, and he treated it with commensurate respect.
But folding handheld PCs were just the beginning of the Microsoft mobile story; they didn’t even have cellular connectivity! Emptied out on the table, the box tells the rest of the story. CE 1.0 devices gave way to CE 2.0 devices. Those gave way to Pocket PCs, and Pocket PCs gave way Windows Mobile, cell phones with mechanical keyboards and color touchscreens. There are devices from Original Equipment Manufacturers (OEMs) like Compaq, HTC, LG, Samsung, HP, Philips, Qualcomm, Motorola, Dell, and Nokia.
In some instances, Gopalan even kept the 3-D printed prototypes the engineers would use for testing before the OEM models arrived. OEM delivery day was always exciting, he reports.
But the first real Windows smartphone (with features akin to the iPhone) was Windows Phone 7, which was released in 2010, a full three years after the first iPhone. Windows Phone 7 was the last of the phones to be based on Windows CE (in this case, CE 6.0). From 2007 to 2011, Gopalan helped architect this new operating system, having spent thirteen years engrossed in the world of CE. By the end of his time at Microsoft in 2013, Gopalan had worked on software design and architecture, led OS development, and overseen a hardware team that worked with numerous manufacturers to make Windows Phones.
By September 2013, Microsoft decided to acquire Nokia’s phones and services business outright. The last Windows Phones were made via that partnership and takeover. In July 2015, less than two years later after the acquisition, Microsoft wrote off $7.6 billion related to the Nokia purchase and laid off thousands in the phone division. Windows Phone never cracked 4% of the smartphone market. The last phones rolled off the line in 2017. At that time, they represented less than 1% of smartphone sales, and support was phased out over the next several years. Users were told to find an iPhone or an Android.
On March 21st of this year, the Department of Justice announced a landmark antitrust lawsuit against Apple. The government argues that Apple is maintaining an illegal smartphone monopoly with iPhone and using the iOS App Store and “green texts,” among other things, to do it. At one point in the complaint, the government alludes to the Windows Phone, saying “Many prominent, well-financed companies have tried and failed to successfully enter the relevant markets because of… entry barriers.” So, did Apple’s alleged monopoly power kill the Windows phone? Gopalan answers: “No. We had an inferior product… we couldn’t get out of our own way to come up with a successful product.”
A novel product? Yes. A loved product? Yes. But a successful product – no. This is how.
Developers, Developers, Developers, Developers
Steve Ballmer’s now famous chant at a developers’ conference in 2000 underscored Microsoft’s understanding of a truth in the software and platform world: a platform is only as strong as its developers. It is its developers. And in the case of most platforms, the lifeblood is third party developers. In that instance he was talking about “.net” – but this understanding served Microsoft well in the development of Windows, too.
“Why is Windows so successful?” Gopalan asks. “It’s all the apps. It is all the wonderful Microsoft applications, and those that others develop. And Microsoft takes a painstaking effort to keep things compatible as Windows changes. Your DOS program that was built thirty years ago will still run well today.” This backward compatibility was a cornerstone of Windows’ success, making it easy for developers to update apps as the platform changed, providing users a vast library of software. It was markedly different in Microsoft’s mobile development.
“When we finally did have an app store, we changed the model three times, making the previous versions of apps invalid,” Gopalan recalls. “We spent three years rewriting the underlying operating system,” he remembers, “because we thought people would care and changed the application over the same UI as was three years with no UI changes, no user experience changes.”
“I’m an operating system guy,” Gopalan admits, “But there is nothing special about the operating system. Other than us geeks, consumers don’t buy the OS. People buy what’s on top of the OS.” People don’t line up around the block for kernel updates. But an App Store – that they’d line up for. And not just consumers, but those fabled developers, developers, developers.
“What is really valued at Apple is the application framework, and the user experience,” Gopalan observes of the old rival. From the launch of the App Store in 2008, a year after the first iPhones, Apple provided a stable, consistent platform for developers – though that was not always the plan, and Jobs originally intended to restrict iOS applications to those made by Apple itself. But he ultimately listened to developers, and the rest is history.
Between Windows Phone 7 (2010) and Windows Phone 8 (2012) Microsoft upended the entire architecture, rendering year-old two-year-old phones obsolete for development. Windows Phone 8 and all subsequent versions moved off the Windows CE kernel onto Windows NT, which had been the main kernel of Windows proper for years. Windows Phone 10 and Windows 10 look nearly identical, and the application framework was shared between mobile and PC. But it took too long to get there, and developers had moved elsewhere.
“We could have learned from the iPhone and the strength of the iOS App Store,” he says. “And Microsoft did do it right, but it was way too late. It was after the iPhone came out and then they scrambled and figured it out. Some people may or may not like what they actually came out with toward the end, but it was interesting, and a lot of people loved it.” Still, the inconsistent application framework remained a frustration to many users.
Gopalan, for his part, thinks that Apple was too good to beat, but that Microsoft could have owned the space that now belongs to Google via Android. All over the world, hundreds of manufacturers produce Android phones at a variety of price points. “Google did a really smart thing after they bought Android,” he said. “They decided to open source all of it and then they close-sourced just the stuff that is required to make a phone, a phone.” And they close-sourced, in particular, the Google Play Store and Google Services. “But you can do anything, you can build any other devices with your own Android fork” – more on that later – “but in order to build a device with Gmail and other important apps, you have to work with Google.”
Microsoft was notoriously antagonistic to open-source software (“open sores”) with Steve Ballmer referring to Linux as a “cancer” in 2001. Did that lead to a missed opportunity? The open-source to closed-source funnel served Google well, and over the years the company put more and more behind its walls. Developers and manufacturers loved Android. “They ran to Android,” as Gopalan put it. At the high end, Androids are as expensive or more expensive than iPhones. Yet some Androids cost as little as $50, and the Android Open-Source Project remains available for anyone, anywhere, to build a device. And as a result, the vast majority of phones worldwide run some form of Android. “That’s a space we (Microsoft) should have owned,” said Gopalan: hundreds of different devices with different features at different prices, united by an app store. “But we just didn’t cultivate it; we didn’t know how to build an app store.”
Bill Gates agrees, saying this in 2014:
“In the software world, particularly for platforms, these are winner-take-all markets. So the greatest mistake ever is whatever mismanagement I engaged in that caused Microsoft not to be what Android is. That is, Android is the standard non-Apple phone platform. That was a natural thing for Microsoft to win. It really is winner take all. If you’re there with half as many apps or 90 percent as many apps, you’re on your way to complete doom. There’s room for exactly one non-Apple operating system and what’s that worth? $400 billion that would be transferred from company G to company M.”
If Gates thought that having only 90% as many apps as iPhone would lead to doom, having only 25% guaranteed it. In 2014, the Windows Phone App Store had just 300,000 apps versus 1.2 million on the iOS App Store. It was never enough.
First, do no harm — to Windows.
But perhaps more fundamentally, Microsoft feared creating something that would cannibalize Windows, its golden goose. “Windows CE, Windows Mobile, and Windows Phone were always hamstrung. They sat outside of Windows itself, so never got the attention they needed,” recounts Gopalan. The problem with that, as is easy to see in retrospect, is that Microsoft could not have known what the future would look like if it assumed that its successful desktop product was the end-state, and that the personal computer would be the dominant form-factor forever. The different “fiefdoms” of Microsoft, as Gopalan calls them, never got along. Some were engineering-driven. Others were sales and “partnership” driven. It was the fiefdom of Windows that ruled supreme over all others. A cartoon by the French blogger Manu Cornet parodies the corporate structure:
In his July 2007 keynote at Macworld, Steve Jobs said, “Every once in a while, a revolutionary product comes along that changes everything.” And among the things that the iPhone changed forever was Apple’s business model. Apple leaned into it, and that made all the difference. In 2006, Macintosh was the biggest segment of Apple’s business by a considerable margin. $2 billion of Macintosh sales made for a blowout quarter back then. Today, annual iPhone sales top $200 billion — over 50% of all Apple revenue. Its services segment alone is now bigger than Mac. In FY 2023, Mac revenue of $29.3 billion was just $1 billion greater than iPad revenue.
“A lot of decisions we made were for Windows phones to be a paired device to a PC,” Gopalan recalls. The little things started to add up. “You could never write a version that was smart enough, or independent on its own,” Gopalan said. “You did your mail on Windows and then you synchronized your mail to the phone where you could look at it, but you couldn’t compose really well. The thing was always crippled.”
In 2017, Satya Nadella reflected: “The decision I think a lot of people talk about – and one of the most difficult decisions I made when I became CEO —was our exit of what I’ll call the mobile phone as defined then. In retrospect, I think there could have been ways we could have made it work by perhaps reinventing the category of computing between PCs, tablets, and phones.”
Indeed, Microsoft did not reinvent the categories of computing between PCs and phones, in large part because they were unable or unwilling to make a phone that would elide the PC, or make the PC a subordinate category. Today, “my life is on my iPhone” is a common idea, even for busy professionals. That single-device strength didn’t “hurt” Mac sales. In other words, cannibalizing Macintosh (on a relative basis) was one of the best things that ever happened to Apple. Could Microsoft have imagined that winning the phone war would be more valuable, by far, than Windows itself?
“The killer app is making calls,” said Jobs at the 2007 iPhone launch. To Jobs, iPhone was a computing platform that happened to have cellular connectivity, not the other way around. And it showed in the way he and Apple dealt with cellular businesses. “Jobs told the cell companies, ‘Here is the iPhone, this beautiful thing. If you want it available on your carrier, please let us know.’” Carriers sought out Apple, not the other way around.
“Apple made people jump,” Gopalan said.
***
“Don’t ask me for advice on applications” Gopalan joked to me when we met just a few miles south from the Redmond Microsoft campus. As it turns out, the Microsoft job in 1994 would be the only job for which he ever applied. But in 2013, after the launch of Windows Phone 10, Gopalan left one Seattle giant for the other. He was recruited to Amazon – to work on the company’s own mobile OS efforts. Fire OS, which runs on Fire tablets and ran for a short time on Fire Phone, is Amazon’s customized fork of the Android Open Source Project. Amazon’s mobile phone experiment lasted just a year; Fire Phone was announced ten years ago in July 2014, and discontinued in August of 2015. Gopalan moved into working on software and hardware for Amazon grocery stores, and finally, on Amazon Web Services.
After just shy of 25 years at Microsoft and Amazon, Gopalan left AWS to start his own company in 2018: Esper. He was working in familiar terrain: harmonizing embedded software for lots of unique devices. Esper is not just a device manager, he told me. “We’re a software infrastructure deployment company.” Esper is a layer beneath device management, for which companies sometimes employ other services.
They help device managers deploy software and keep devices running smoothly. Where are those devices? They’re the screens on exercise bikes, the kiosks and point-of-sale tablets in restaurants (Taco Bell is a major customer). They’re devices for delivery drivers, they’re signs in airports. Devices running Esper are in hospitals and schools, in hotels, restaurants, and gyms. Customers pay a monthly fee for each device, and Esper provides ironclad security and software deployment. “Esper is invisible,” Gopalan says of his product.
He had worked with his co-founder and COO Shiv Sundar on Windows Phones when they were both at Microsoft. Esper represents, in a way, the realization of the work that began long ago with Windows CE – to put wonderful software on any device you can imagine. The irony, after all this time, is that it had to be done with Android. Esper Foundation is the company’s own customized operating system, a fork of the Android Open-Source Project. This year, Esper started supporting iOS devices, too.
The company has raised $101 million in four venture rounds, most recently its 2021 Series C, led by Insight Partners. The company’s offices are in Bellevue, just a few miles from the Microsoft campus in Redmond. The offices are unmistakably those of a modern software startup, with standing desks lining an open floor. Gopalan wore a gray Esper t-shirt, as did a number of his engineers and salespeople. Tucked away down the hall, in a carpeted room was the hardware lab – the tinkerer’s lair, where devices are tested and stored on metal shelves.
Gopalan has worked at just three companies in his career: Microsoft, Amazon, and now Esper. Being a startup CEO is his first experience as a salesman and marketer, but that it has been a good experience. I asked him how being CEO has changed his view of the days he spent at the two Seattle giants. He paused, and looking back at me, asked, “Why didn’t I do more? Could I have done more and taken ownership of a bigger project? I was hanging out with engineers, I probably could have learned more about the sales side.”
His direct reports are in the office every day, as is he. He says he’s skeptical of remote work cultures — something he learned as a young engineer at the university-like Microsoft campus, where Saturdays were for work and office conversations. I asked him if he still had connections at Microsoft nowadays. “Fewer and fewer as the stock has gone up,” he replied. Gopalan represents a new American archetype: the immigrant dad who put in his years building products that millions of people use, without much fanfare or credit.
The most famous person in this mold is Satya Nadella, who became CEO in 2014 and led Microsoft back to the top of the world after the “lost decade.” He’s better known as just “Satya.” In our conversation, Gopalan’s admiration for Satya was clear. To many, his tenure represents the rebirth of an elite engineering culture at the company. And he’s strategic, too. Of the CEO’s management of the OpenAI board fiasco last fall, Gopalan said, “Satya didn’t say anything. He was quiet. He facilitated but didn’t take credit. It was perfect. Nobody knew.”
Other things are remarkably different. Gopalan never worked on Microsoft’s cloud division (he worked at AWS), but its ascendance amazes him all the same. “They figured out the cloud better than others. And it’s the killer. And it’s ton of Linux servers. Running Linux at Microsoft would’ve been a heresy before!” Today, Azure is Microsoft’s largest revenue segment, representing nearly 40% of revenue. The lesson to him is clear: “You need to be relevant long enough to pivot.” And pivot they did.
In his own startup, there have been a few pivots, too. “There have been difficult decisions,” he said. “I think we grew way too fast after the second fundraise. Everybody had to scale down their growth numbers. “It’s getting harder and harder to do startups,” he told me. “But we’ve been lucky so far. It’s still a struggle, a grind. But you work hard, and at some point, you hope to hit escape velocity.”
Last year, the Windows CE saga came to a quiet end just shy of 30 years since it secretly began as Pegasus. On October 10, 2023, Microsoft ceased extended support for Windows CE 8.0, or Windows Embedded Compact 2013, the final version. In a sign of the times five months earlier, the company repurposed the original CE codename for a new program; the new “Microsoft Pegasus” awards AI startups with $350,000 in credits from Azure, LinkedIn, and GitHub. Another sign of the times: on October 13, three days later, Microsoft closed its $68 billion purchase of Activision — its largest ever, nearly ten times the size of the ill-fated Nokia purchase a decade earlier.
There are still devices out there running CE; some of them are in the wild, working past retirement, wherever they are; some are in boxes like the one in Gopalan’s garage. To hobbyists, the old devices are beloved artifacts of a bygone age. But to people like Gopalan, they’re dreams from their younger selves. He lamented to me that having left Microsoft a decade ago, he doesn’t have access to the CE code anymore. “I wish Microsoft would open source it, just so that I can go look at the code. I’d love to see it and find the lines that I wrote… this thing, that line, that component but not that one.”
Is anyone at Microsoft listening? I hope so. The company whose second CEO once trashed open-source software as a cancer owns, under its third CEO, the largest open-source repository in the history of the Internet. It would be trivial for Microsoft to upload the original Windows CE code to GitHub. What’s to lose, its competitive edge in a market that’s been dead for years?
For now, Gopalan’s box will head back to the garage, and he’ll keep plowing at the Esper office (he reports that Mrs. Gopalan wishes it’d stay at the office). These days, sales and partnerships consume his time and attention, but he still looks forward to demo days with his engineering team, and technical problems that require his personal involvement. Toward the end of my visit, I asked him what he does for fun, suspecting I knew the answer. “I go tinker. To prove to myself that I can still do it.” Now over fifty, Gopalan says that he hopes it’ll be the same when he’s eighty. And he hopes, still, that one day Microsoft will open source the code to the first project he worked on as a young engineer, set the Pegasus free, and let old dreams live once more.
