Programming Windows: The Product (Premium)

For years, Microsoft had worked closely with its biggest partners to maintain and expand a symbiotic PC ecosystem based on Windows. This relationship was so successful that it catapulted Microsoft to fame, fortune, and infamy. But it required work from both sides, with Microsoft and the PC makers meeting regularly to discuss upcoming technological advances to ensure that they were supported in a timely fashion, and to assess upcoming technological disruptions to ensure that Windows PCs remained the most popular personal computing platform on earth.

The relationship may have reached an apex of sorts during the development of Longhorn, when Microsoft’s biggest PC maker partners worked diligently to incorporate that system’s incredible advances in new hardware designs. HP, for example, partnered with Microsoft on several generations of “Athens” PCs that included built-in telephone handsets and other futuristic features. And ASUS created a prototype laptop with a small external smart display for viewing notifications without having to open the display lid.

Neither ever came to market. And there was always an uneasy feeling within Microsoft generally, and the Windows team specifically, that the PC makers could and should do more. But thanks to the smaller margins inherent in building hardware, PC makers were always trying to cut costs and find new revenue models based on added value services. They started adopting new technologies, like new versions of the Universal Serial Bus (USB), more slowly, or not at all in cheaper models. They piled more and more software into their products, making Windows boot and run more slowly, and less reliably.

Most gallingly to Microsoft, the PC makers didn’t always support technologies that were provided by new Windows versions at launch; HP’s initial Media Center PC in 2002, for example, was a basic tower PC that looked out of place in a living room and not the stereo component-like design that Microsoft had requested for Windows XP Media Center Edition.

By the time the vindictive Steven Sinofsky had seized control of Windows and fooled the Senior Leadership Team into believing he knew the right way forward, Microsoft’s relationship with PC makers had already become strained. He responded to the drama by making it worse: he didn’t brief PC makers about the new technologies that Windows 7 would provide until it was too late for them to support them in time for the launch. And so the first generation of Windows 7 PCs were just uninspired Windows Vista PCs. Both sides blamed the other.

Thanks to the Apple envy that was spreading throughout Microsoft, Sinofsky quietly launched a plan to push the PC makers out of the equation: Microsoft, he argued internally, should make its own PCs. And though CEO Steve Ballmer eventually championed succumbed to the alure of copying Apple’s unilateral strategy as well, this request was shot down repeatedly by the company’s Board of Directors. And so Sinofsky briefly turned to a plan B to prove how inept these partners really were and how they could not be trusted to make Windows look as good as it should.

There were two major initiatives that came out of this work.

The first, codenamed Velocity, was designed to streamline the Windows 7 boot-time as much as possible, but it required PC makers to buy in and, more crucially, stop overloading their PCs with crapware that would counter Velocity’s benefits.

The second was called Signature: a small team at Microsoft would resell third-party PCs, and only through Microsoft’s retail stores, that had been stripped of PC maker’s inefficient software images and replaced with a cleaner, more efficient Microsoft image.

The idea behind both initiatives was similar: Microsoft would use definitive data to prove that PC makers were their own worst enemies and were harming the customer experience.

Both initiatives would eventually fail, as Sinofsky no doubt planned. Lenovo was the only major PC maker to embrace Velocity in the early days, and then only on high-end ThinkPad offerings that should never have been burdened by crapware to begin with. And Signature limped along for a few years having made little difference with PC makers, who by that time had managed to lower PC margins even further by relying on low-end and low-cost netbooks to gain marketshare.

Armed with this data, Sinofsky again made another appeal to Ballmer and the board right as Apple was set to announce the iPad: to beat Apple and push its own mobile-first, touch-first agenda, he argued, Microsoft would have to design, manufacture, and sell its own PCs. And because the software giant had, by that point, its own retail stores, mostly in the United States—the first had opened around the same time that Windows 7 launched in 2009—it could blunt the impact of this competitive affront by limiting the distribution of its PCs to its own stores.

The timing was right: by 2010, it was clear that the iPhone and other smartphones would transform personal computing, and the pending iPad launch had triggered a deep feeling of fear and unease in Redmond. What if Apple ran away with the entire market, and supplanted Windows?

This time, the argument was right, too: Microsoft’s PCs could potentially revive the Tablet PC, a pet project of co-founder Bill Gates that had failed badly despite his backing. By creating touch-first tablet form factor computers, Microsoft could take on the iPad directly and differentiate its offerings from the traditional form factors that its PC maker partners preferred. And it was a relatively low risk: if the PC line failed, it wouldn’t harm Windows too much. Probably. Hopefully.

The board gave Ballmer and Sinofsky the greenlight.

“[January 2010] was right when we started floating the idea to make a PC with Windows 8,” Sinofsky later admitted. “This was extraordinarily sensitive so almost no one knew, not even most of the Windows team. Competing with the [PC makers] was what had stopped previous efforts to make hardware (note: the top 7 [PC makers] were and remain Microsoft’s largest customers in the entire company).”

Sensing that the shift away from Windows would also take down Intel and the x86 architecture with it, Sinofsky had already started work on porting the next Windows version, Windows 8, to ARM, the mobile-oriented architecture that powered the iPhone and Android devices (and, eventually, the iPad as well.) Coincidentally, Microsoft’s hardware labs successfully booted Windows on ARM for the first time the same month that the board had signed off on the Microsoft PC plan. And so there would be both Intel and ARM versions.

With the form factor decided, Sinofsky and his hardware ecosystem lead Mike Angiulo assembled a team to begin work on a go-to-market plan. The team was led by Panos Panay, a heretofore unknown who had joined the software giant back in 2004 to work on devices like keyboards and mice and was by then in charge of Microsoft’s hardware efforts. Among the many key decisions that needed to be made, perhaps none were as crucial as the brand name. Microsoft had, over time, ushered many popular brands to market, Windows and Office among them. But it had also stumbled badly at times, and it was often mocked for its terrible naming choices. Microsoft needed to get this one right.

Fortunately, it already owned the right brand, Surface, which it was using for a struggling line of so-called table computers that resembled the early sit-down video arcade machines of the late 1970s and early 1980s.

Envisioned as a solution for casinos and hotels, the Surface tables offered multitouch capabilities and early mixed reality (MR) capabilities that allowed them to interact with real-world objects that were placed on their—wait for it—surface.

Oddly enough, Panay pushed back against repurposing the brand.

“You have no idea how hard it is to get your head out of the brand,” Panay told Business Insider of this decision years later. “When you’re working on something like the table, that’s what a ‘Surface’ is.”

Sinofsky and Ballmer overruled him, and Microsoft rebranded the second-generation Surface tables as PixelSense, the name of a new technology the bulky devices included, freeing up the Surface brand for the new line of PCs. This ended up being the right decision: the second-generation Surface—er, PixelSense—table failed to find an audience and Microsoft ending up killing the product line a few years later.

But Microsoft’s investments in the table computer weren’t for naught: Panay partnered with Surface table co-inventor Stevie Bathiche to create the Surface PCs and related accessories. These products were developed secretly—even most C-level executives at Microsoft were unaware of the effort—alongside Windows 8 with the idea that the core features in each would complement the other.

By 2011, Sinofsky was itching to show the world—or at least hint to the world—what his team had been working on, with the hopes of dampening enthusiasm for the iPad. So he went off script and appeared at the D9 Conference with Julie Larson-Green that June, ostensibly to show off Windows 8 and its multitouch capabilities for the first time. But the prototype-like “development station” they used for the demo was a preproduction Surface tablet with the same 10.6-inch display that would later ship in the final products. Given how primitive it was, few suspected a thing.

Working in concert with Sinofsky and Angiulo, Panay’s new Surface team planned two Surface PCs, which they called devices to differentiate them from the traditional PC form factors of the past and help promote the notion that Windows wasn’t tied to fading legacy technologies. (That terminology persists at Microsoft to this day despite the fact that Windows only runs on PCs.) Both utilized a tablet form factor with a 10.6-inch widescreen multitouch display and an integrated kickstand. Both could accept clip-on covers with integrated keyboard and touchpads to create a laptop-like experience. And both offered minimal expansion capabilities, with proprietary cover and power connectors, a single USB port, and either HDMI or DisplayPort for video-out.

But where one Surface utilized traditional Intel-based innards and offered complete backward compatibility with the wide range of Windows software and peripherals, the other would run on the ARM platform and offer much more limited compatibility. This version of Surface would also be thinner and lighter, and because the ARM chipset required no active cooling, it would be silent as well. Sinofsky and the Surface team believed that this Surface version, which went on to be named Surface RT after the Windows 8 variant—called Windows RT—that it ran, would be the volume seller, and that it was uniquely positioned to take on the iPad. The more PC-like version, called Surface Pro, would be aimed at power users.

From a hardware perspective, Surface RT did offer some advantages over the iPad. Its integrated kickstand allowed users to position the device on a table or other surface and use it hands-free. Its single USB port was limiting compared to most PCs, but the iPad didn’t have any expansion capability per se, and Surface RT would provide basic support for many popular PC peripherals automatically. The iPad didn’t support keyboards or mice at that time, so Surface RT’s two cover options—customers could choose between a Type Cover with a traditional keyboard or a Touch Cover with a more streamlined look—meant that users didn’t have to cede two-thirds of the display to a virtual keyboard, as was the case on Apple’s tablet.

But Surface RT wasn’t a no-brainer. The biggest problem, of course, was Windows 8. This coming OS would go on to be the biggest disaster in the history of Windows, neatly eclipsing Windows Vista. But the Windows RT variant somehow managed to be even worse. In addition to all the issues with the standard version of Windows 8, Windows RT couldn’t run any Windows software that users downloaded from the web or tried to install from a disk. Instead, it only ran the software that came with the system—including a limited version of Office 2013 with no Outlook app—and the lackluster selection of Metro-style apps that could be downloaded from the Windows Store.

There were other issues, of course. Surface RT ran very slowly, making the device seem pokey when compared to traditional PCs. And while it supported many hardware peripherals with basic functionality, customers would not be able to install the custom drivers and their associated software utilities, making printers, scanners, and other devices less useful when paired with the device.

But Windows 8 was a runaway train barreling towards its destiny, and Sinofsky was hellbent on delivering it—and his new Surface PCs—within the arbitrary three-year timeframe he had promised, by late 2012. And so Surface would arrive with many issues, issues that Sinofsky and Panay figured they could fix in future iterations. Only one of them would remain at the company long enough to see that work continue.

By early 2012, Sinofsky was ready to reveal Surface to the world. In keeping with his insular way of doing things, he wanted to control the messaging for this announcement, “own” the launch event, and keep it secret from the rest of Microsoft. But with Sinofsky then known to be making moves on the CEO position, Steve Ballmer intervened and demanded to be part of the announcement, and to be the person who first revealed this secret new product to the world.

To be fair, there had been rumors by that point. In December 2011, DigiTimes reported that Microsoft was prepping to release its own PC hardware. But this seemed so far-fetched, and the publication’s record with rumors was spotty enough, that few believed it. And to maintain secrecy, Microsoft didn’t alert the press about the coming launch event until four days before it was scheduled; worse, the invitation was so vague that many of the invited press—including me—declined to travel across the country on such short notice without knowing why they were doing so, or even where exactly this secret event was being held.

I of course contacted my PR representative to ask for more information, but all I was told was that I’d definitely want to be there. I disagreed and told Microsoft that I could keep a secret and would not fly to Los Angeles on such short notice unless I knew more. With no more information forthcoming, I declined the invitation.

I did not regret this decision.

Microsoft announced that it was entering the PC market with Surface on June 18, 2012 at Milk Studios in Los Angeles. As the event opened in the small venue, it was Steve Ballmer and not Steven Sinofsky who greeted the small cadre of press corps that had taken a chance and traveled to southern California. Not surprisingly, he was careful to frame what Microsoft was doing in a way that students of the industry would quickly recognize as a take on industry seer Alan Kay’s oft-quoted dictum that “people who are serious about software should make their own hardware.” And to remind people that Microsoft, the software giant, had been making hardware since its earliest days.

“At our foundation, Bill Gates and Paul Allen made a bet. A bet on software. But at the same time, it was always clear that our unique view of what software could do would require us to push hardware sometimes in ways that even the makers of the hardware themselves had yet to envision,” he began, somewhat confusingly. “That’s the nature of the dynamic between hardware and software, pushing each other and pulling each other forward. In fact, our number-one revenue product, actually, the year I joined Microsoft 1980 was a hardware product, something known as the SoftCard.”

Sidenote: The Microsoft SoftCard was a plug-in card for the Apple IIe that allowed that system to run CP/M.

Ballmer then queued up a video to remind people of Microsoft’s many hardware products over the previous 30 years. It noted that the software giant had over 3200 hardware patents. That Microsoft had bundled its first mouse with an early version of Microsoft Word. That it still sold best-selling keyboards and mice. And that the Xbox video game console was, of course, a hardware device, as was the briefly successful Kinect add-on. There were webcams. And that crazy Surface/PixelSense table computer.

“We believe that any intersection between human and machine can be made better when all aspects of the experience—hardware and software—are considered and working together,” he argued as he returned to the stage, seemingly channeling Steve Jobs. “With Windows 8, we did not want to leave any seam uncovered … We wanted to give Windows 8 its own companion hardware innovation. What is this innovation? It’s something new. It’s something different. It’s a whole new family of computing devices from Microsoft.”

And with that, Ballmer took a step back and the room went dark as a second video played. This one was … confusing. Instead of showing a hardware product, it showed a metal object rotating over an alien landscape, one that slowly transformed into the Surface logo. This strange imagery, which evoked H.R. Giger’s “Alien” design, would be expanded over the next few years in various misguided Surface marketing bits.

“This is the new Microsoft Surface,” Ballmer said, holding a Surface RT tablet as the lights came back up. “It embodies the notion of hardware and software really pushing each other. People really do want to create and consume. They want to work. And they want to play. They want to be on their couch. They want to be at their desk. And they want to be on the go. Surface fulfills that dream. It is a tool to surface your passions, to surface your ideas, to surface your creativity, and to surface your enjoyment.”

And having surfaced the ire of Steven Sinofsky by stealing away the Surface reveal from him, Ballmer then then said that we could learn more from Sinofsky and the Microsoft Surface team. Sinofsky, to his credit, did not reveal the deep disappointment he must have felt from this one-upmanship, and he bounded on stage and proceeded as if he had been the first to appear before the audience.

Microsoft was reimagining the tablet, just as it was reimagining Windows with Windows 8, Sinofsky started. It was a “great PC that’s a great tablet, a new type of computing device, and a stage for Windows.” He quickly rattled through some specifications, some nonsensical—it weighed under 1.5 pounds and was just 9.3 mm thin, he said, and yet it had a full-sized USB port (unlike an iPad), and had edges that were bezeled at 22 degrees “so that the PC itself fades into the background.” What?

It was the first PC with a “full magnesium case” that involved liquid metal and a “physical vapor deposition process” that supposedly made it scratch and wear resistant. (The first two generations of Surface products went on to be infamous for their horrific scratches.) And it was “the first tablet to incorporate dual 2×2 MIMO antennas, that means it provided “the very best Wi-Fi reception of any tablet today.” (Surface was infamous for poor Wi-Fi reception for its first several years in the marketplace.)

Sinofsky highlighted the built-in kickstand, which he said should be “integral” to any tablet—take that, iPad—and how its hinge design was like something from “the finest luxury car.” Like the iPad, Surface could be outfitted with a cover—he showed a Touch Cover but did not name it—that attached securely with magnets, and Sinofsky seemed particularly proud of the audible “click” sound it made when the cover connected with the device.

The cover was “made from a fine northwest polar tech,” he bragged, and just 3 mm thick, and it felt like a book in your hand when you closed it over the Surface.

Unlike the iPad’s covers, however, the Surface Type Cover provided “a full multi-touch keyboard” on its flat surface and “a modern trackpad with left and right buttons.” That keyboard had special keys related to the Windows 8 Metro-style UI, he said, and its typing experience was “twice as efficient as typing on glass” and “more comfortable.” How efficient it was compared to real keyboards would, of course, go on to sink the accessory.

Moving to a second Surface RT tablet that was connected to the on-stage displays via HDMI, Sinofsky noted that the rear camera was tilted at 22 degrees to match the angle of the device’s kickstand and “perfectly framing” the audience in front of him.

This Surface, he finally admitted, was based on an NVIDIA ARM SoC, was called Surface RT, and would run Windows RT. But Microsoft was making a second Surface, based on Intel processors, that would run Windows 8 Pro. And to show off that device, he called for Mike Angiulo to come up on stage. And not Panos Panay.

This Surface model was aimed at the “millions of professional desktop users out there.” Like Surface RT, it was a stage for Windows, but this one was heavier (just under two pounds), thicker (14 mm), and had a higher resolution Full HD display. (Left unsaid for now, it provided much worse battery life and required active cooling with fans because of the less efficient Intel parts.) Microsoft hadn’t yet arrived at a marketing term for what was in effect a high-DPI display, given it small 12.6-inch size, so Angiulo noted obtusely that the display was “a combination of a very specific pixel geometry rendering and an optical bonding process that together create the effect that your eye can’t distinguish between the individual pixels at normal viewing distances, in this case 17 inches, less than arms-length.” Woof.

Sidenote: When Microsoft killed the PixelSense table computers, it would repurpose that brand for the Surface PCs as well: future Surface PCs would have so-called PixelSense displays. This name is an obvious ode or retort to Apple’s “Retina” displays.

The Touch Cover that Sinofsky had previously revealed would, of course, work with this other Surface. But Angiulo introduced a second cover, called Type Cover, that included a real hardware keyboard instead of the virtual keyboard on Touch Cover. Despite its short key throws, this keyboard could “rival those of the finest Ultrabooks that have ever been announced,” Angiulo claimed. He then claimed that this Surface would keep cool with its “perimeter venting,” despite its powerful 3rd-generation Intel Core processor.

Like the original Tablet PCs, these two Surface PCs supported Digital Ink, and each would ship with a bundled Surface Pen that offered 600 dpi of “sub-pixel accuracy.” The pen could attach to the side of either Surface magnetically, but as future customers would soon discover, this connection was weak and would result in many lost Pens.

Surface Pro, as this second Surface model would come to be known, had a DisplayPort for video-out, as opposed to the HDMI port on Surface RT. And it featured a much faster USB 3.0 port where Surface RT made do with USB 2.0. Angiulo demonstrated how fast it was with a file copy, noting that this process was five times faster with Surface Pro.

The entire Surface event lasted less than 45 minutes and Panos Panay, the leader of the new Surface team, wasn’t even brought on stage until almost 30 minutes had elapsed. His first appearance—and mention—was when he was asked by Angiulo to come out and basically hand-model the Surface Touch Covers while he and Sinofsky posed triumphantly with Surface Pro and RT.

And it seemed that that would be it. But with Panay seeming to walk off stage for good, Angiulo suddenly revealed to the public that this hunched little man was, in fact, somehow the leader of the Surface team. And he had “some great stories … about the product and how it came to be.”

After an awkward co-grope with Angiulo, Panay faced an audience alone for the first time at a Microsoft event. A strange on-stage presence with an unusual delivery, he spoke for less than 13 minutes. But that was all it would take for most onlookers to form an opinion.

Some really enjoyed his presentation style and, I suspect, some still do. But most were likely unsettled by Panay and his strange speech patterns, which involved lots of awkward pauses and repetition. And it is perhaps perfect that he somehow uttered the term “the product” (or some variant of it) over 30 times in just under 13 minutes.

“You heard Steven and Mike both say that this was built as a stage for Windows 8,” he said. “That was part of our core vision for the product. It was very important for us that we had the hardware fade to the background for this product. It was important so the Windows software could rise to the surface.”

Panay stumbled through a bizarre list of the “perfect” attributes of Surface, from its 22-degree chamfered edges—“perfectly comfortable in your hands”—to the over 200 custom parts that were built just for Surface. The team had iterated the sound that Surface’s kickstand made “over and over again”—a “critical point”—to get exactly the right sound “so you get that visceral feeling, that emotional attachment to your product when you open this kickstand and close it.” That kickstand was exactly .77-milimeters in thickness—again, “perfect”—the same thickness as a credit card.

And then it got weird. Really weird.

“But nothing, nothing, stirs me more, nothing gets me more excited, than Touch Cover,” he said. “This is an important technology that came out of our group.”

It would not be an important technology: Microsoft would quickly discontinue Touch Cover because customers preferred the more traditional Type Cover by a wide margin. It is perhaps instructive that the man who led the Surface team—and, incredibly, still does—was so focused on such a meaningless part of the initial product line.

That said, Type Cover would be the least of Microsoft’s problems when it came to Surface.

As Sinofsky had explained at the end of the Surface announcement, Microsoft would ship “Surface for Windows RT,” as Surface RT was briefly known, alongside Windows 8 in late 2012, in 32 and 64 GB configurations. It would be priced like “comparable tablets,” which presumably meant the iPad since there were no other tablets. “Surface for Windows 8 Professional”—Surface Pro—would ship in 64 and 128 GB configurations three months later and would be priced like comparable Ultrabooks.

Microsoft had decided to launch Windows RT first, and it would make millions of more units available than was the case with Surface Pro, because Sinofsky and the Surface team believed it would be so popular. But this was a terrible miscalculation, and a great example of insular group-think. Launched in New York City that October alongside Windows 8, Surface RT faced limited availability thanks to its exclusivity at Microsoft’s retail stores. But customers stayed away in droves even when they could get it. And as the sales figures came in over the holiday selling period, it became clear that Microsoft had a disaster on its hands, one that complemented, if you will, the disaster that was Windows 8.

Microsoft had manufactured 4 million Surface RTs, but by early 2013, over 3 million of them remained unsold and were sitting in warehouses, losing value alongside the Surface covers no one was buying either. According to the book Beneath a Surface, Microsoft on some days sold under 100 Surface RTs each day, leading to a change in strategy, with the firm making the device available at Best Buy, Staples, and other electronics stores to widen availability. But it wasn’t just too late. This was a product few wanted.

In mid-2013, Microsoft was forced to write down $900 million related to the declining value of the millions of Surface RTs it had sitting in warehouses. By legally declaring that these assets were now worth considerably less than before, Microsoft could dump them in a series of fire sales. And as prices fell and fell, finally to just $199, it was able to sell off its remaining stock and move on.

Surface Pro did sell better than Surface RT at the original price points, but that was a low bar. And Microsoft would for some reason release a second-generation Surface RT, now called Surface 2, alongside an updated Surface Pro 2 one year later. But those devices sold poorly, too, and Microsoft killed off its ARM-based Surface offerings by the time Surface Pro 3, with an updated design that proved more popular, appeared in 2014.

Surface also predictably alienated all of Microsoft’s PC maker partners, despite the fact that Sinofsky had described “the top 7 OEMs [as] Microsoft’s largest customers in the entire company” in another bit of not paying attention to telemetry data and feedback. Each of those top 7 PC maker partners would go on to embrace Google’s rival Chrome OS platform and build and sell their own Chromebooks, which would compete with their own Windows PCs.

That strategy says a lot about the real relationship and lack of trust between Microsoft and its PC maker partners, and it is yet another example of Sinofsky’s divisive political thinking undermining the products he and his team were building. This, and his CEO ambitions, were of course what led to him being dismissed by Steve Ballmer just ahead of the Windows 8 launch.