107. Click In With Surface

Happy Holiday to those in the US. This is a special double issue covering the creation and launch of Microsoft Surface, an integral part of the reimagining of Windows from the chipset to the experience. To celebrate such a radical departure from Microsoft’s historic Windows and software-only strategy this post is unlocked, so please enjoy, and feel free to share. I’ve also included a good many artifacts including the plans for what would happen after Windows 8 released that were put in place. The post following this is the very last in Hardcore Software. More on what comes next after the Epilogue.

As a thank you to email subscribers of all levels, this post is unlocked for all readers. Please share. Please subscribe for updates and news.

Back to 106. The Missing Start Menu

In 2010, operating in complete secrecy on the newest part of Microsoft’s campus, the Studios, was a team called WDS. WDS didn’t stand for anything, but that was the point. The security protocols for the Studio B building were strengthened relative to any other in the entire engineering campus. Housed in this building was a team working on one of the only projects that, if leaked, would be a material event for Microsoft.

WDS was creating the last part of the story to reimagine Windows from the chipset to the experience.

When we began the project, it was the icing on the cake. After the Consumer Preview, it had become the one thing that might potentially change the trajectory of Windows 8.

As Windows 7 finished and I began to consider where we stood with hardware partners, Intel, the health of the ecosystem, and competing with Apple, I reached the same conclusion the previous leader of Windows had—Windows required great hardware to meet customer needs and to compete, but there were structural constraints on the OEM business model that seemed to preclude great hardware from emerging.

At the same time, the dependence on the that channel meant there was no desire at Microsoft to compete with OEMs. In 2010, the Windows business represented 54% of Microsoft’s fiscal year operating income and Office was 49%—yes you read that correctly. BillG used to talk about that amount of revenue in terms of the small percentage of it that could easily fund a competitor or alternative to Windows. The “Year of Linux” was not just a fantasy of techies but a desired alternative for the OEMs as well. So far the OEMs had not chosen to invest materially in Linux, but that could change especially with an incentive created by Microsoft’s actions.

Like my predecessors, I believed Microsoft needed to build a PC.

Building PCs was something BillG was always happy to leave to other people. In an interview in 1992, Bill said, “There’s a reason I’m the second-biggest computer company in the world…. The reason is, I write software, and that’s where the profit is in this business right now.” On the other hand, the legendary computer scientist and arguably father of the tablet concept, Alan Kay, once said, “People who are really serious about software should make their own hardware.”

Microsoft was founded on the core belief that hardware should best be left to others. In the 1970s hardware was capital intensive, required different engineering skills, had horrible margins, and carried with it all the risks and downsides that pure software businesses, like the one BillG and PaulA had pioneered, did not worry about. With a standardized operating system, the hardware business would quickly consolidate and commoditize around IBM-compatible PCs in what was first a high-margin business that soon became something of a race to the bottom in terms of margins.

Microsoft’s fantastic success was built precisely on the idea of not building hardware.

BillG was always more nuanced. He and PaulA believed strongly in building hardware that created opportunities for new software. Microsoft built a hardware device, the Z-80 SoftCard, to enable its software to run on the Apple ][. Early on, Microsoft created add-in cards to play sound. PaulA personally drove the creation of the PC mouse, the famous green-eyed monster. Modern Microsoft built Xbox, but also Zune and the Kin phone.

Apple built great hardware and together with great software made some insanely great products.

To build hardware in this context meant to build the device that customers interacted with and to build all the software and deliver it in one complete package or, in economist’s parlance, vertical integration.

Mike Angiulo (MikeAng) and the ecosystem had the job of bringing diversity to the PC ecosystem, a diversity that Apple did not have. This diversity was both an enormous strength and the source of a structural weakness of the industry. PCs in any screen size or configuration one might need could always be found or even custom built covering any required performance and capacity. If you wanted something like a portable server or a ruggedized PC for a squad car or a PC to embed in 2 Tesla MRI, Windows had something for you. Apple with its carefully curated line essentially big, bigger, biggest with storage of minimum, typical, maximum across Mac were the choices. Even a typical PC maker like Dell would offer good, better, best across screen sizes and then vary the offering across home, small business, government, education, and enterprise. Within that 3 x 3 x 5 customization was possible at every step. This is the root of why Apple was able to have the best PC, but never able to command the bulk of the market.

The idea of vertical integration sounds fantastic on paper but the loss of the breadth of computing Windows had to offer was also a loss for customers. It is very easy to say “build the perfect hardware” but the world also values choice. One question we struggled with was if the “consumerization” of computing would lead to less choice or not. In general, early in adopting a technology there is less choice for customers. Increased choice comes with maturity in an effort to obtain more margin and differentiate from growing competition.

One bet we were making was that Windows on ARM and a device from Microsoft was the start of a new generation of hardware. It would start, much like the IBM PC 5150, with a single flagship and then over time there would be many more models bringing the diversity that was the hallmark of the PC ecosystem.

That is why we never built anything as central and critical as a mainstream PC, and never had we really considered competing so directly with Microsoft’s primary stream of profits and risking alienating those partners and sending them to Linux. The Windows business was a profit engine for the company (and still is today) and that profit flows through only a half dozen major customers. Losing even one was a massive problem.

Microsoft had also lost a good deal of money on hardware, right up to the $1.15 billion write-off for Xbox issues in 2007. Going as far back to the early 1990s and the original keyboard, SideWinder joystick, cordless phone, home theater remote, wireless router, and even ActiMates Barney our track record in hardware was not great. Microsoft’s hardware accessories were at best categorized as marketing expense or concept cars. It was no surprise my predecessors backed off.

Like the mouse, the sound card, and perhaps Xbox, I was certain that if we were to succeed in a broad platform shift in Windows that we would need to take on the responsibility and risk of building mainstream and profitable PC devices. We tried to create the Tablet PC by creating our own prototypes and shopping them to OEMs as proofs of concept. We repeated this motion with the predecessor to small tablets called Origami, same as we did for Media Center. Each of these failed to develop into meaningful run rates as separate product lines even after the software was integrated into Windows.

OEMs were not equipped to invest the capital and engineering required to compete with Apple. As an example, Apple had repurposed a massive army of thousands of aluminum milling machines to create the unibody case used in the MacBook Pro. Not only did no OEM want to spend the capital to do this, but there was also no motivation to do so. Beyond that, the idea of spending a huge amount capital up front on the first machines using a new technology until a process or supply chain could be optimized was entirely unappealing even if the capital was dedicated.

The OEMs were not aiming for highly differentiated hardware and their business needs were met with plastic cases that afforded flexibility in design and components. In practice, they often felt software was more differentiating than hardware, which was somewhat counterintuitive. The aggregate gross margin achievable in a PC software load was a multiple of the margin on the entire base hardware of a PC. The latest and coolest Android tablet was a fancy one made by Samsung, with a plastic case. The rise of Android, a commodity platform, all but guaranteed more plastic, lower quality screens, swappable parts, and the resulting lower prices.

The OEMs were not in a battle to take share from Apple. They were more than happy to take share from each other. Apple laptop share was vastly smaller than the next bigger OEM making Windows laptops. Each OEM would tell us they could double the size of their entire company by taking share from the other OEMs. That’s just how they viewed the opportunity. The OEMs were smart businesspeople.

Thinking we needed to build hardware, then building it, was one order of magnitude of challenge. Choosing to bring a product to mass market was another. Hardware is complicated, complex, expensive, and risky—risky on the face of it and seen as risky by Microsoft’s best customers.

The Surface team, organized within the Entertainment and Devices division home to keyboards and mice, finished the first release of their namesake computer. It was a Windows Vista-powered table like the ones popular in bars, pizza places, and hotel lobbies in the 1980s when they ran Frogger. The new platform software provided the cool demos for Windows 7 touch. Surface came about from a research project rooted in long term efforts around optics and display technology. The effort was productized as the original Surface, with Panos Panay (PanosP) brought in to help accomplish that from the peripherals group. Unfortunately, the commercial viability of the table was limited. RobbieB, the executive of the division (that also included Xbox and phones), was looking to offload the effort or better said make Windows pay for it.

We would move the team over to take on our hardware challenge if we could figure out an arrangement that would not torpedo the Windows business. It was this step that put a halt to all previous projects.

Managing what could become a first-party hardware effort from within the Windows team posed significant challenges, even obstacles. PC OEMs would rightfully become unglued if they believed the decidedly limited, information they shared with MikeAng’s ecosystem team about their plans was shared with other OEMs. If their information were to be shared with a first-party team, that was even worse. Some of the earliest concerns expressed to regulators about Microsoft had to do with the walls, or lack thereof, between different parts of the Microsoft ecosystem and competitive parts in Microsoft, for example, between Windows and Office, where Office might have an unfair advantage. I had no desire to further a conspiracy theory.

JulieLar offered her own set of challenges, but almost the opposite in nature. She was concerned about having a first-party hardware team that acted like an OEM more than a part of Microsoft. She wanted our hardware to fully embrace, not mangle, the Metro design language. I hardly wanted a team that would embrace the economics of OEM, such as crapware and expensive accessories.

Collectively we wanted a decidedly-non OEM to be our OEM.

Both Julie and Mike spent a lot of time with Panos to help us all arrive at a structure and work process that was set up for success. Panos joined Microsoft in 2006, relatively recently, after working at NMB Technologies, an expansive 60-year-old Japanese maker of electronics components including keyboard switches and skateboard ball-bearings (that I used to pay top dollar for in high school.) The complexities for Panos in jumping into the thicket of Windows were significant, not to mention the delicate nature of first-party PCs. The concerns for all parties were legitimate.

The idea of a whole hardware team moving over, rather than building one organically, was worrisome, but we did not have the time to start from scratch. By most any measure, the idea that in about 30 months we would have a new ultramodern PC built on all new components the industry had never used before by a team that had only made a table computer was crazy. Panos’s commitment to being part of the massive effort was not just significant but deeply sincere. He saw the opportunity the same way we did—a chance to reimagine Microsoft.

Working for Mike, Panos would expand the team in every dimension. Hiring and growing engineers for hardware, firmware, mechanical, plastic, manufacturing, acoustic, industrial, and safety, with designers from graphical to industrial to packaging, along with all the support functions I neglected to mention in that long list. We even had to embark on some special construction in the buildings to account for the needs of the kind of equipment required to build a PC.

WDS was formed in June 2010 just as we were beginning to code Windows 8. We did not even send out an org announcement mail, keeping with the secrecy of the project. The name “Surface” moved over with the team and would naturally stick with the device when we were ready to bring it to the public. For now, codenames and codewords were in order. Not just one codename, but an entire system of codenames. We had codenames for every part including unique codenames used for each vendor – part combination. There were codenames for every presentation. Over the course of the project, we maintained over 200 codenames. Why? The secrecy was to maintain our commitment to keeping information separate.

It is worth noting, we were piloting ARM work on NVIDIA Tegra hardware and seeding any groups across the company with ODM-style tablets housing Tegra components. No one inside Microsoft lacked hardware for testing, evaluating, or betting on Windows on ARM.

The next 18 months were the most remarkable sprint in hardware development I could have ever imagined. Panos built an extremely tight and remarkably talented team to deliver products in relatively short time. Everyone stepped it up to a new level. One of the most critical aspects of assembling this team was his direct manager, MikeAng, who created a cocoon around the team, isolating them from all the forces internally. Mike also mentored Panos on the product development methods used to create Windows while apprising him of the best ways to integrate with the software team to avoid thinking like an OEM—the key problem we set out to solve.

One of the things that Mike had to do which was super valuable for the whole effort was to be the “vault” for information about Surface and information about the PC ecosystem. The two orgs, Ecosystem and Surface, could have absolutely no information leakage between them. In many ways, the integrity of the Windows business model was in Mike’s hands. It was easy to trust Mike since I’d worked with him since he joined Microsoft out of college, and his post-Microsoft career included becoming a lawyer, so I think it is fair to say he was the perfect person for this role. As we’ll see it was just a bonus that he was exactly the right kind of engineer for this role as well.

The core of the WDS team had previously worked together, but the scale of their previous projects was small by comparison—everything was bigger in Windows. The team doubled from the early days of mice and keyboards to a broad range of peripherals and most recently Xbox consoles. Ten years in, Xbox was selling about ten million units a year (about 800,000 in 2021) and the business remained roughly break-even due to console margins. We had much grander aspirations.

The story of the two devices, both tablet form factors, that became the first Surface PCs was one of incredible design and engineering efforts going through a remarkable series of iterations in an incredibly short time and then scaling to manufacture millions. It was also a blur. Creating a PC was new for me as well. It is not only fair, but important, to say that without the collaboration across Julie’s product team, JonDe’s engineering team, GrantG’s test team, and Julie herself, we would not have Surface. Jensen Harris (JensenH) and the entire UEX development provided an attention to detail on everything from the BIOS boot screen the on-screen keyboard, sound scheme, to the out of box experience that rivaled anything Apple did. It is one thing to build a reference PC within the walls of Microsoft—something done many times before. It is entirely another thing to deliver the innovation and scale of products that we did, and then bring it to a worldwide market.

I cannot stress enough how much of a whole-organization effort Surface was.

We did not set out to build what the industry was calling a consumer tablet or a tablet for consuming information and lightweight computing needs, as many, or just Apple, characterized the tablet market relative to the PC market. The world did not need more of that. Microsoft’s heritage and business rested with productivity, so the overarching goal was to create a PC that was great for productivity, creation, and mainstream information work. The PC would just happen to fit the current definition of a tablet, which was smaller than an 11.6” laptop and usable as a screen and a traditional laptop posture.

To justify building our own hardware we needed a unique point of view.

We believed that if we could build a tablet that worked fantastically well for the web and Office, while being ultra-lightweight and ultra-reliable, we could redefine the device that constituted the bulk of PCs used in school and work. One of our most significant challenges was that the narrative for new devices was a dominated by tablet, meaning a keyboardless slate focused on consumption. This created the situation that even if we achieved our vision technically, we had an enormous communications challenge.

The first Microsoft device, code-named GT or Georgetown for no real reason, was a premium ARM tablet able to deliver on the promise of PC-style mobile productivity. The whole point of building hardware as a first party was to do what the PC ecosystem would likely struggle or simply decline to do, and ARM was first among the challenges faced. The PC makers had all been struggling to enter the smartphone market, and ARM required a significant amount of product engineering and investment that was inconsistent with the low-priced mindset. Besides, they were fully occupied with Netbooks and later Ultrabooks and seeing little success with smartphones. As described in section 101, those OEMs working with ARM were genuinely enthused and supportive, but somewhat like Detroit’s reaction to electric cars, the internal tension proved too great to create and bring to market ARM, at least for the launch of Windows 8.

Whether premium implied premium price along with a premium experience was an ongoing debate. Typically, in hardware designs, manufacturers work backward from a retail price goal, assuming a certain BOM, margins, volume to manufacture (impacting the BOM or bill of materials), marketing and sales costs, returns, and so on. As product makers, this makes intellectual sense, but it frontloads so many constraints that rarely do great products emerge. I chose to start from the user scenario and see how far we could go and what we could get for what price, knowing that as we converged difficult decisions would arise.

This was a point of constant discussion with SteveB, who—and this was not news or even unique to this project—favored low prices and high volumes. Gaining market share was always preferable and what had gotten Microsoft to where it was. I felt this time we needed to take a different approach—reimagine what a Windows device could be even if that meant we start at the high end, perhaps like Tesla was doing at the time with Elon Musk’s famously not-so-secret secret plan to iterate to a lower price with greater scale, funding that iteration with earlier high-priced products.

Early in the process, Mike and Panos set the tone for the role WDS would play in the overall strategy for Windows 8. It was extraordinarily helpful. The mantra for the joint efforts of hardware and software coming together was expressed as, “our job is to build a stage, a stage for the operating system and software.” This fit well with the whole design for Windows 8 software, which itself was viewed as a stage for apps and user creations. The idea that every layer of the system was trying to get out of the way of the other layers above aligned entirely with the Metro design language, and also exactly what no one ever did, especially OEMs. This was in stark contrast to typical PCs and even typical OS iterations of the past, which felt the need to announce themselves whether via branding (those Designed for Windows or Intel Inside stickers, for example, that don’t even appear on Apple laptops because they drove Steve Jobs crazy) or via icons, popups, or notifications.

We were not designing the hardware experience after the software experience (or before); rather, we were designing the hardware and software experience together—a reimagined experience.

I tried to keep my own role in check. I’d seen many past hardware cycles within the company where there was so much going up and down the chain, so much approval, and so much tweaking by executives. Plus, it is crazy easy to have an opinion on a device and act like some enlightened big-shot executive. Panos’s experience with executives at Microsoft rightfully made him wary of me at first. Executives often had an outsized view of their contribution while underestimating the amount of work, backchannel, and eyerolling they caused. Advertising, pricing, user interface design, and feedback on hardware were places that executives easily meddled. And for that, teams suffered.

I used a lot of different laptops. I bought nearly every iPad typing cover available (and there were dozens). I was a road warrior, so I was real-time testing. I used browsing, email, and Office a lot. I could have probably justified my own opinions, and many times I really wanted to, but had to resist the urge. I had asked Mike and Panos to pull off what by any measure would be a Herculean task. My responsibility was not to meddle.

From my vantage point in developing the product, there were five inflection points in the design process—a high-altitude view since there were thousands of choices made in developing the product. At some level these were choices to be made or constraints. The project could have easily become impossible or careened off path with so many potential degrees of freedom. Anyone who looked at the typical roundup of PC laptop reviews for back to school or holiday was left browsing a dizzying array of specs, ports, dimensions, and component choices. It would have been easy to get lost in all these decisions to be made. Each one of those decisions is related to many others in an ever-complex set of linear equations to be optimally solved.

The team picked these key points in the design process: chipset, the productivity posture, typing, materials, and peripherals.

By establishing the initial platform of ARM and a sub-11” form factor many of the traditional “spec” issues of designing the device would fall out. But this understates just how many and how fast, Panos and team needed to put constraints in place—constraints that would maintain a point of view and deliver a product.

The chipset was perhaps the most straight-forward choice for the team to make on technical grounds, though hardly without controversy or for lack of a better word politics. There was a good deal of pressure from the Windows Phone team and from some OEM partners who were still trying to build phones of their own (though on Android now) to go with Qualcomm. Qualcomm was viewed as a safe choice and the choice with the most intellectual property and lawyers to support that. On the other hand, the early hardware was far from ready to build a PC especially the graphics support which was so critical. It was the first taste of developing a product where the choices had implications beyond building the best experience. My own view from dealing with Qualcomm was that they were more concerned with the volume of devices we would commit to up front than with the quality or innovation we would deliver, even early in the process.

The teams working across Windows porting to ARM were favorable to NVIDIA given the choices we had. In particular, NVIDIA was strongest on graphics which were key to the Metro-style experience. The underlying graphics chips mapped closely to our DirectX API and made developing device drivers supporting the animations and effects we needed with the performance we required as straight-forward as could be given all that was new. Aside from that NVIDIA was a joyful partner and great at working with ODMs to supply test hardware.

Ralf Groene (RalfG) led hardware design efforts for WDS. He joined Microsoft with a classic design education and experience at several marquee design houses in Silicon Valley. If we were casting a movie for an industrial designer, Ralf would be perfect. RalfG and the design team were cycling through prototypes of sizes and overall form factors at an incredible pace. Engineering for productivity and typing fell to the design team, with a very strong collaboration with the UEX team to make sure the software could gracefully adapt to the presence or absence of a keyboard and trackpad.

The kickstand was such an amazing choice. Yet I lack a clear recollection of the genesis of it, only that the moment I saw it I was convinced it was an inspired choice. Having a kickstand solved the main problem tablets had, which is how to keep using one when you wanted to put it down or more importantly to type. The kickstand could easily have been a gimmick but in fact it became an iconic element and one that served to further the point of view centered on the dual modalities of consuming content hands-free and full-time work on the go. The kickstand would have tradeoffs in real-world usability, especially with the larger devices to come later, but the for the first generation I felt it was exactly the tradeoff to make.

The design team iterated with a kickstand, something that had been found on some early media players and phones. Should the kickstand be two small legs, one fatter leg, or something else? Should it support both landscape and portrait? The team settled on a full-width kickstand, almost a fold-out foot that provided a rock-solid level of stability. That stability came from a unique hinge design, created entirely by the mechanical engineers on the team. Opening and closing the hinge had the feel and robustness of a luxury car door, a German one of course. As the hinge closed the air dampening effect would give the motion a soft cushion with a nearly inaudible, but intentionally there, sound of luxury. I once spent time in the lab with the mechanical engineer on the team amazed at just the complexity and number of choices that were considered in the design and fabrication of the hinge. It was a great example of how every small detail is vastly more complex than readily apparent.

The beauty of the hinge and kickstand were that they reduced the thickness of the whole productivity scenario by quite a bit—the plastic cases of the iPad including Apple’s own cases were true compromises. The 2015 Folio, which added two layers of material in a complicated origami-folding design and later the 2018 Magic keyboard, which was just heavy at 600 grams or 130 grams more than the iPad itself, were both awkward. The latter keyboard added so much weight and thickness to the iPad that it was more practical to carry a laptop. Whereas Steve Jobs had to sit in a chair holding the iPad, with Surface one could hold it or flip out the kickstand and watch the move while enjoying popcorn, or having a video conference with the integrated HD webcam angled precisely enough so it would capture your head and not neck or below when placed on a table.

Important for the productivity posture was the overall size. When working, a screen can never be too big, but when traveling a screen can never be too small. Finding the balance was super important. The math would work out super well to have a screen that was 10.6-inch diagonal but with a widescreen layout of 16x9, the typical HD or Blu-Ray numbers. At 1366x768 the screen was equally optimal for the new Windows 8 snap view to show two apps side-by-side—GT as a stage for Windows 8—and also to display movies full screen without the drawbacks of letterbox modification that consumers hated.

There was quite a bit if a debate over 16:10 versus 16:9 aspect ratio and the possible screen resolutions. UEX favored the extra width of 16:9, though 16:10 offered more vertical space for productivity but could not support side-by-side apps. The availability of components and cost of using the UEX preference made it a tough choice. This was also an area where we were going against the iPad trend which picked a more traditional even old 4:3 aspect ratio used on NTSC TV and VGA monitors at 1024x786 which was also criticized in reviews for letterboxing widescreen video. The iPad was between wide screen video and the traditional trade book and Amazon Kindle aspect ratio of 3:2.

From a supply chain and manufacturing perspective, the choice was more complex. The supply chain for screens was following the volume of all manufacturers combined. No single manufacturer could simply order a size not being used by others without committing to volume. In 2011 the dominant size for tablets was a 10.1” screen with 1280x800 resolution, a 16:10 aspect ratio. We saw the 1280x800 resolution as a waypoint, picked for price by the Android tablet ecosystem. It was neither good for productivity in landscape nor good for reading in portrait mode.

We were deeply committed to having a screen that supported the side-by-side view of two apps in landscape. This came with a real tradeoff, however, in the usability as a pure tablet. In other words, even from the choice of screen we were optimizing for productivity. When used as a “reading tablet” the portrait mode orientation of 9:16 was particularly awkward. Reading with one hand holding the device was tiresome, like holding a college chemistry textbook in bed. Additionally, traditional laptops with 13” or 15” screens had moved in large numbers with Windows 7 to the 16:9 and 1366x768 resolution, indicating that resolution was an excellent design point for apps and would be around for years to come.

The screen had two other technical issues to work through with significant impact on the experience. First, the core display technology for LCD screens was in the midst of a transition from MVA (Multi-Domain Vertical Alignment) to IPS (In-Plane Switching) with the latter being the newer technology. Going with the newer technology would be preferred as a basis for applying a touch sensor, but that further constrained the potential supplies and sizes.

Second, the display panel would need an extra stage in manufacturing to attach a touch panel sensor. There were two approaches here as well. The standard approach, again for Android tablets, was known as air-bonding. In this technique the touch panel does not directly touch the underlying display panel leaving a small gap which eases manufacturing. Unfortunately, this also introduces parallax, an effect by which where you touch and where the sensor detects your touch do not always align in your brain. This is seen at check-out counters and ATMs which prefer the manufacturability and low-cost of this approach. While cheaper, it is a disaster for precision work on a PC. The newer approach, direct bonding, used by Apple and more premium tablets, directly bonded the touch sensor to the display. This was more expensive and had a non-trivial defect rate increasing costs as well.

The way you know from the supply chain that you are asking for something difficult is that the price goes up and delivery goes further out. The suppliers, all singing from the same page, strongly encouraged a 16:10 screen with 1280x800. Their view was that this was the volume tablet resolution. After many trips and many meetings, we were able to secure a high-volume supply of the 10.6”, 1366x768, IPS, direct bonded screen. This was a tough choice and added early risk, but once you lock in a supplier as a volume partner risk finds a way of decreasing. I suspect over time even for the 10.6” screen we would have gone with a more comfortable aspect ratio, but by then we would have had more answers to windowing and productivity than we had with Windows 8. I’m getting ahead of myself.

Productivity without a keyboard is not really productivity. We were not confused about this point as they were in Cupertino. Productivity without a mouse or trackpad, on the desktop, was not even possible. Because Windows RT had the desktop and desktop Office, a trackpad was required. The 10.6-inch diagonal screen would create a case with a 10.1-inch long edge, which by no accident was wide enough for a nearly full-size home row of keys for touch typists. This could be a huge advantage over the keyboard for the iPad, which was only 9.5 inches across, making a big difference in key size and spacing. Our screen size allowed for arrow keys and a row of function keys with the trackpad. These small differences bring a world of benefits when designing devices with which people have such intimate reactions. This is decidedly different than a desktop PC or even a large laptop. There’s something about a device that must be held that makes small decisions so much more important.

The keyboard presented a unique opportunity to innovate to maximize portability. Panos himself personally an expert in mechanical keyboards and part of Microsoft’s own efforts in innovative keyboards was well-positioned to develop two keyboards we would introduce with GT.

Fundamental to the keyboard was the idea that it would be faster and more accurate than touch typing on a glass screen and include a trackpad for precision, or “professional,” pointing as so many reviews of iPad noted was lacking. Since GT had a kickstand, the keyboard did not need to also serve as a redundant cover for the back but when folded to cover the screen would offer screen protection. The underside of the keyboard was covered with a layer of what I always termed northwest fleece with a soft woolen feel when carrying GT around in your hand.

Our folding keyboard could make an easy transition from productivity and typing to reading or watching while lounging around. The keyboard itself was a touch-based invention from Stevie Batiche (StevieB), who was the resident scientist/engineer leading all things touch and display, and a genius. We would also add a very thin mechanical keyboard, one of the thinnest mechanical keyboards ever released to market. Our hearts would be with the touch keyboard at launch owing to its productivity and on the go scenario.

StevieB’s ultrathin keyboard invention was essentially a touch panel with ever-so-slightly raised key outlines impressed on the panel. The electronics were created by laminating several layers of materials including a touch sensor and all the “wiring” together in a hot press machine—sort of a touch screen sandwich. The keyboard, called Touch Cover, was a mere 3.2mm and weighed only 7 ounces and under 200 grams. Given the dimensions of the screen, Touch Cover was able to incorporate a trackpad on par with many small laptop trackpads and a complete set of laptop keys including dedicated keys for Windows 8 charms along the top row, which also doubled as traditional PC function keys. In US English even the F and J keys had slight outward impressions for traditional typists to find the home row. There would be many skeptics about the keyboard when it came to typing proficiency. While there were many risks in the entire project, at the time Touch Cover seemed like the riskiest part of the project. It would be so immediately visible and open to snap opinions. The press that typed for a living were used to their preferred “feel” when it came to laptop keyboards, often a sore spot for just about every Windows laptop review.

GT had a beautiful hinge and an incredible touch keyboard, but how to attach the keyboard posed another challenge. Because of the hinge the cover could not be attached to the back of GT, but that would be dumb anyway—adding weight with no purpose was what we saw on the iPad. Once again, the mechanical engineers had a go at the problem. They put magnets to work. Not just any magnets, but a series of magnets of exactly the right strength to support the device, even swinging it from the keyboard, while also easy to remove. The magnets did not just attach the keyboard to the tablet but they were the connectors for the signal and power. Attaching the keyboard had to perfectly align the connectors or it would not work.

The magnetic keyboard was a more difficult to path to perfection and reliability. Upon first seeing it, just as most others, I was skeptical. Would it fall off? Would it properly align even when connected in a sloppy manner? Would flipping it around to use as just a tablet prove “goofy” or would it really work? Would Windows correctly enable the on-screen keyboard at the right time and get out of the way entirely when a keyboard was attached? Would it work just as well for the thin mechanical model?

The process of connecting the keyboard to the device proved to be more than just reliable but something of a signature. The first time we showed the Surface to the team that would lead creating television commercials they immediately connected with the “click” sound the engineering team worked so hard to get right. That click would become the centerpiece of the initial campaign along with the profile of GT and the keyboard. A subdued version of a magnetic connector also with a click would be used for the very small charging adapter. Ironically, Apple’s pioneering MagSafe connector that was so popular on MacBooks was not used on iPad, which took its lead from iPhone for charging. Even today Apple continues to struggle with magnets when it comes to covers and keyboard cases for the iPad. The new 2022 iPad typing portfolio with a kickstand is kind of a mess.

The removable keyboard provided several advantages. To compete in the “pure” tablet market the keyboard could be priced optionally and also be reviewed as an accessory versus required. It also permitted personalization by choice of colors since the laminating process was able to substitute any color materials for the back or front. The back of the Touch Cover could be customized in both material and look, for example corporate logos or art as we later offered. Finally, StevieB’s innovative touch panel was not restricted to keys and could be used to create any touch surface. Since it could be removed easily, we envisioned the potential to have specialized Touch Covers dedicated to specific applications. One example we showed early on was a synthesizer drum app mated to a custom keyboard. At a small, private event held one night in Los Angeles I even had a chance to watch actor/musician Zack Efron of High School Musical have fun with the prototype music generating cover.

The materials choice for inner fame and device case were almost always where the OEMs made choices that were best for the bottom line—plastic provided low cost, light weight, rigidity, ample room for cooling, and agility across component and peripheral changes. This is where Microsoft’s ability to provide a significant investment could make a real difference in the final product. The engineering challenge with case materials is the triangle of cost, weight, and rigidity. Cost is not only the cost of the material, but the manufacturing cost of making the case, with all the curves, holes, and cutouts that make it a computer. Materials can be inexpensive and lightweight but too flexible to be durable—the screen is glass and needs structure to prevent it from flexing and breaking. Materials can be rigid and lightweight but cost a tremendous amount, such as ceramics used in fighter jets. Aluminum is lightweight and relatively rigid, but to bring the cost down Apple invested a huge amount of upfront capital in order to take blocks of aluminum and use mechanical milling to turn it into a PC.

We needed the device to be as thin as possible and from the start we considered every fractional millimeter we could save.  We knew that bringing over Windows in one product cycle, including the desktop, would bring real challenges to our ability to compete with the nine hours of video playback possible on the iPad. We needed every available millimeter for battery. An innovation in materials could prove a game-changer.

In research, the materials scientists became intrigued by a relatively new process of injection molding a magnesium alloy. Such a material was expensive and the manufacturing process complex, but the resulting parts were lighter than machined aluminum, extremely rigid and strong, and created by more flexible molding. At this point, the injection molding process had only been used for small parts such as watch frames or jet engine components, but the materials partner, Mike, and Panos’s team thought it possible to use for the much larger cases. The material could be readily colored via a permanent and robust vapor deposition process, which afforded other opportunities.

MikeAng, a mechanical engineer by training and certified alloy welder, dove into the process, visiting the factory in Asia. He returned with movies of cases being injection molded that looked like scenes from 20th century industrial America—sparks flying, molten flows of metal, giant machine presses.

Betting GT on this new process was one of the more uncertain aspects of the hardware design, and also very expensive, relatively. The high upfront costs made us all nervous. With any upfront cost, called NRE, or non-recurring engineering costs, the only way of justifying them is to make a lot of the product which spread the cost across many devices. That was certainly our intent.

The material would get the name VaporMg, pronounced vapor mag. Finished in a smooth, so smooth, black-gray, the material was another manifestation of the collective efforts of design, materials, and manufacturing coming together. While we originally planned to make the entire case out of VaporMg, supply constraints resulted in a more traditional aluminum frame with VaporMg used more sparingly. While that resulted in a bit of increased weight, the resulting scaled manufacturing and cost reduction were a good tradeoff.

There’s no doubt that VaporMg was the most extravagant choice in the whole project and one I felt the most over-extended in considering. In a big company with plenty of money, it is not uncommon to see branching out into new areas take on almost comically bad cost controls relative to industry norms. Without experience or baselines to compare and with all the excitement of “those other people doing this must be dumb” it is so easy to do this. Across the product, Mike, Panos, and team had extremely good controls and constant attention to the BOM, bill of materials, and NRE. Mike was a product of our frugalness in Office and brought that to this project and team. The VaporMg choice was one I felt we should go with even though it was so uncharacteristic for me or much of Microsoft.

One thing we considered was that we could reuse the materials process and plant capacity with other OEMs and license it to them for whatever devices they wanted to build. The idea of acting as a source for various components was a step we considered for trackpads and touchscreens as well as we saw OEMs unwilling or unable to take on the NRE costs to create competitive laptops. Once again though, I probably failed to consider they also viewed such costs unnecessary when it came to their share battle competing with other OEMs more than competing with Apple.

VaporMg was so strong that the team turned a standard GT chassis into a skateboard, on a lark. As a former skateboarder this thrilled me to no end. At one of the many press events hosted on campus offering a behind-the-scenes look at developing hardware, I skated around the lobby of Studio A. Not to worry, I was wearing a helmet though it was a borrowed bike helmet we located just for the photos. We snapped some photos that continue to live on. To further demonstrate the strength of the material, we showed off the drop-test machine that simulated dropping a tablet with different forces and angles and even directly on the kickstand—each time GT performed admirably. Protected by the Touch Cover, the combination felt relatively indestructible. The screen was securely protected without any extra casing on the back maintaining the thin profile and light weight.

Finally, productivity for a premium laptops and tablets circa 2012 still depended on wires or more specifically dongles. Apple’s MacBook Air approach of wireless everything was conceptually great but practically a pain, certainly in the early days. Carrying around dongles to connect to projectors or to USB devices was annoying and error prone. As a tool for productivity, GT needed to show up at a meeting and get handed the slides on a USB drive, pop the drive in, launch PowerPoint, and then connect to a projector, or handle devices like microphones and speakers. As a stage for Windows 8, we had optimized this type of flow from a performance and user interface perspective—get to the user’s work product and get to work as quickly as possible.

Mike and Panos framed GT as a device that should connect in the way people needed it to connect. GT had a standard USB port, standard audio output, and a mini-HDMI port in addition to the dedicated magnetic-charging connector. The USB port was rather tricky as it set a minimum thickness that was not an issue on the iPad that did not have a universal USB connector. The fat USB port was a bit ugly and some on the industrial design team referred to it as a “missing tooth” in an otherwise sleek profile. At one point we were so close to the standard for a connector we risked losing the trade group designation and permission to use the logo. This in addition to strong support in hardware and software for all the current wireless protocols for Wi-Fi and Bluetooth including support for wireless speakers and displays. This array of connectivity defined the soul of the device and shouted out point of view that the device was not a peripheral but a full computer.

Power adapters had become the bane of my existence as someone who used a variety of devices at home and work and travel. It was still too early to have a standard connector—Apple was still using the wide pin connector, which was painful and finnicky despite the super nice magnetic connector. GT created an ultra-skinny magnetic connector. I was, however, much more interested in the power brick. I could not stand the typical PC that had a three-foot cord to the PC, then a brick, then another three-foot cord to the wall. It turned out people in Japan disliked those too and Yodobashi Camera came to the rescue with an aftermarket model, a single 2 meter cable with a brick at one end with folding prongs. I stocked up on those. I really wanted GT to have an adapter with simple folding prongs. I wanted this so much I didn’t even bother to ask about non-US plugs. As a constant traveler, I was also cursed with hotel rooms with too few outlets. Back then I almost always had to unplug a lamp which usually involved moving furniture around. I often traveled with an after-market phone charger that had an extra pass-through outlet to charge a USB device. I needed this because PC USB ports did not charge when the PC went into standby, which I usually discovered at 6 a.m. local time.

Okay, so I meddled once after complaining one too many times.

One day RalfG invited me over to the studio and brought me through dozens of models for power adapters. I knew as soon as I got to the studio that Panos was managing me, but in a way that was entirely okay given how close we’d become. As we walked from 3-D model to model, I opined and envisioned about what it would be like in a hotel room. They explained to me all sorts of stuff about thermals, tolerances, UL standards, and even patents. One company had a patent on prongs that folded into power bricks! I picked the smallest adapter with folding prongs and a built-in USB charger. I like to think this was another iconic choice. It was not. It was merely useful. Ultimately, GT did not have a USB port on the charger but did have folding prongs. The second device had the USB port I wanted on a slightly higher-powered charger.

From the project start we had been considering a second device. Early on, Mike had suggested to me that the team could build a desktop all-in-one using a screen with the next generation touch panel technology from the Surface table running an Intel processor. This device was the concept for a follow-on or successor to Surface table. It was a beautiful device. The prototype followed the same lines and hinge as would appear much later after many iterations as the Surface Studio all-in-one.

We really wanted to showcase desktop productivity, especially the all-in-one form factor, which had no presence with PCs. Apple made all-in-one Macs as their desktop Mac and those, as with iPads and MacBook Air, made their way into every movie, high-end retailer, and boutique office setting. An all-in-one with the GT premium aesthetic could put Windows in those spots too.

PC makers finally decided to wed the displays they were making with the laptop parts they made to create all-in-ones competitive with Apple iMac. Windows 7 saw a broad introduction of low-cost and reasonably performant all-in-ones, which was great to see.

The arrival of PC all-in-ones and the reality that the next generation screen technology was not ready, especially in a large size and at scale led to us abandoning the all-in-one product. A good deal of work had gone into the basics of developing an Intel PC, which presented us with an opportunity.

Standing around the now defunct all-in-one prototype, we discussed the idea of creating an Intel-based GT with the Surface BOM, or a derivative of it. At first, I thought this was entirely against the spirit of the project as it would simply compete with our OEM partners and cause more of a rift than not. At the same time, I’d been so happy with the progress we were making while equally aware of the limitations customers would see in ARM-based Surface. The early prototype tablets like the \build Samsung had the promise of being great tablets for using and testing Windows 8 apps while also working as platforms to build those apps or use any existing x86 Windows development tools.

Would the existence of a Microsoft-branded Intel device help address customer concerns, or would it create more confusion? Would OEMs see such a device as an inspiration, or would it further annoy them? What would Intel think?

I had many concerns about forging ahead with an Intel-based PC coming from Microsoft. It did not seem prudent. Frankly, it felt more like an extra poke at OEMs when our main goals were embodied in the ARM-based device. 

What could we offer uniquely as a stage for Windows 8? The Ultrabook investments were ongoing, but Intel’s specifications called for them to be large-screen devices for Intel-specific channel strategies. Everything that was a tablet or Netbook-like computer, less than Ultrabooks, on Intel was running low end chips that had mediocre performance on Windows. What if we mated mainstream laptop chips, Intel Core, to the GT form factor? We would have an ultraportable Intel-based Windows 8 device in an ultra-convenient Netbook-sized package. There was nothing like it in market, an ultra-portable Ultrabook.

The team took this as an opportunity to up-level all the specifications from GT. More of everything including a full HD screen, faster USB, support for dual monitors, bigger battery, more memory, and storage, and by default it would use the Type Cover with mechanical keys. We were super confident in the Touch Cover, but we saw in early usage that it did not have the throughput for a full-time professional user on this professional-oriented device. We weren’t going to be religious about it and quickly adapted.

The up-leveled GT could in fact contribute something that would help enormously to smooth over the risk of building hardware. We could offer a super high-quality pen experience, exactly what BillG wanted. Since the device would be more portable, resembling a typical paper notebook, lighter than scarcely available Tablet PCs, it might prove to be the ultimate pen-based notetaking device. The fact that we could focus an Intel SKU on the pen and handwriting and deliver a size that the OEMs were not considering provided a reasonable justification in favor of the device plan. The up-leveled GT filled an obvious gap in the whole ecosystem lineup.

Why did we not put a pen on GT? The power draw and thickness of another layer added by the current state-of-the-art pen digitizer would impact the competitiveness with iPad. Prioritizing the ARM device to be competitive with hardware specs of the iPad was a key factor, and a good tradeoff for us to make. Given my own reticence about the pen it was also an easy tradeoff, though I had to repeatedly defend it with BillG.

This was enough positive and with MikeAng we considered or gamed out how the OEMs would respond to an x86 device. Our main point was that the OEMs were focused on Ultrabooks in order to gain the advantages of Intel pricing. Our device would fall outside Ultrabook specs and thus prove more innovative and less part of the crowded Ultrabook space.

From my perspective, I viewed the device as an objection handler. It was a way to signal to the market that Windows 8, the x86 product, was well-suited to playing in this new world. Instead of a focus on ultra-mobile and all-day battery life, the focus was on power and all the needs of professionals in an ultra-mobile package. This reflected the early reviews of Windows 8 and the challenges we were having with the tablet narrative as put forth by Apple and reviewers.

GTX, or Montlake, was the code name for this second device. It looked bit like a puffed-up GT. Unlike GT, GTX would have all standard Windows PC components from the chipset to the BIOS to storage and RAM. It would also have a fan, and to mitigate that the team developed innovative cooling slots all around the device to keep internals from overheating or creating a hot-to-the-touch exterior. By virtue of it being Intel based, GTX had noise and heat that GT did not, and had about half the battery life more typical of the real-world numbers being seen for the first generation Ultrabooks, about 4-5 hours, compared to the 9+ hours we would see with GT.

GTX was about three months behind GT. I was clear that GT was the priority should trade-offs in any dimension be required. By the time we announced GT we would announce that GTX would be available three months later. For many that gap felt amateurish compared to the most recent Apple launches which had announced new devices on a Tuesday for sale on a Friday. As we know now, that was not the long-term norm for Apple nor was it how the iPad, Apple Watch, or other devices launched since. Nevertheless, Apple was doing everything right at the time so as agile as I thought we were, it was still less than was expected of us.

By the early spring of 2012, the manufacturing team started making small runs of test devices. It took months of work to scale up an assembly line, and it was expensive to make devices in any quantity for testing. Plus, everything was top secret. We had made about 1,000 test ARM devices partnering with NVIDIA, which were being used mostly in labs. They were also shared with teams across Microsoft contributing to WOA under the strictest security protocols. We never had enough of these test machines for everyone who might have wanted one to just check out, but we had enough to build the software across Windows, Developer, and Office. These devices cost over $1,000 each and were great to have but decidedly test machines. I was still using our NVIDIA test device, but, as with anyone who knew about GT, I was anxious.

Time was passing both slowly and quickly. We were closing in on finishing the Windows 8 software schedule, August 2012, which meant time was slow as each day fewer and fewer code changes were made. GT, however, was getting close to the point where it was time to commit to manufacturing. Spinning up a line means you want to ramp it to full capacity and not stop, otherwise we’d just be burning cash.

Surprisingly, there were absolutely no leaks. Even across Microsoft no one suspected anything, certainly not first-party hardware. With Windows 8 RTM approaching, the overall messaging for our reimagining point of view would start to diffuse through Microsoft. That meant talking about hardware. At a regularly scheduled Board Meeting I was asked to provide an update on Windows 8. As the code was essentially done and we had broad disclosure already there was little to do but assure the Board that we were on track and to discuss the worries we all collectively shared. The Board, however, had not yet seen the final GT hardware. In many previous meetings we discussed ARM and the strategy, including building first-party hardware, and I had previously shown demonstrations using the NVIDIA testing prototype. My sense was that the gravity of the decision to make first-party hardware had not really sunk in—the general view of “competing” with our OEM partners was still viewed as a challenge, especially because the Windows Phone Team had tried to create a phone but ended up partnering with Nokia, before later acquiring the company.

Whenever I spoke with the Board, I was always excited to show the progress of our product—so long as we were ready to, and it would not be over-promising or confusing. I would almost always drive a demo myself casually and without much fanfare—not only was this more authentic, but it also permitted a more casual interaction. With GT I did just that. At one point I just pulled it out of my bag and began to show it off just as we had been doing in incredibly limited ways. The pitch was easy because it was what we had aimed to build. Promise and deliver.

The Board and Steve, however, were much more focused on the big decision to sell first-party hardware and the bottom-line impact to the overall Microsoft P&L and earnings. I think the Board was still smarting from the Xbox write-off five years earlier. This ended up being the discussion, which seemed awkward to me as we had articulated this as the strategy from the start of the project. At the meeting, I was starting to get the feeling that I had in 2001 when we suddenly backtracked on the fully-baked Office.NET product plan. Only this time I felt I had laid all the groundwork and had discussed this at several other meetings. In hindsight, I suppose it was the reality of seeing the actual device and that we were delivering that made it all seem that much more real.

There was a tense discussion about the costs, especially the hundreds of millions of dollars to build out the final assembly capacity and the balance of creating enough units to keep the cost down by spreading the non-recurring costs over as many units as possible but not too many that we’d flood the market. One thing I had done specifically was not play any games with the budget or expense approval. Many projects at Microsoft spent a lot of money by spending a little bit at a time so as not to exceed an executive’s approval limit. My whole career, with the help of CollJ, I kept my spending authority artificially low. The spirit of my mentor JeffH was always looking over my shoulder reminding me it was Microsoft’s money and not mine to spend. This meant SteveB had to approve all the spending and we would have to have the discussion up front. At the end of the meeting when the Board asked the final approver question, there was no doubt in my mind. At the same time, strategically and from the perspective of how well we had executed there was no doubt. I was accountable for the result and was certain we should move forward and said as much. SteveB and I would have several more discussions, mostly centered on the price point and volume before he hit approve on the invoicing tool.

Before the first public showing, I wanted to show the device to PaulA. Paul was an early advocate of Microsoft building hardware and had several forays into leading-edge hardware through his Vulcan innovation company. I brought a device over to his office near the stadium used by the Seattle Seahawks. As someone who said he believed the old saying about those serious about software building hardware, Paul was excited to see what the team had accomplished. We spent most of the time in a deep discussion about hardware manufacturing, the bill of materials, and how Microsoft would scale the business. Paul’s Vulcan had built an ultra-portable PC powered by a novel Transmeta chip, which aimed to be Intel-compatible and consume less power. He was profoundly connected to the challenges of making hardware and at the same time agreed on the need to do so.

We set a date for a public unveiling in mid-June 2012. Just two weeks before at the All Things Digital Conference, we wanted to make sure one reporting team got a bit of a preview, only one team. I was using my Surface RT all the time, but not in any public places, and essentially handcuffing my messenger bag to me. D10, the 2012 conference, featured Apple’s Tim Cook, Mayor Michael Bloomberg, and a host of top-shelf guests, including SteveB, but no Windows demo, Phew.

We prepared to give a brief and private demo at the conference. We did not schedule any time but instead sort of set up a bit of an ambush outside on the terrace of the show’s green room. I messaged show co-host Walt Mossberg and asked if we could find 15 minutes between his on-stage interviews to show him and Katherine Boehret something very cool and secret. He was intrigued and they both of course agreed.

JulieLar and I would meet them and provide a surprise sneak peak of Surface. This was the first showing to anyone outside the Microsoft “tent” to see it. I grabbed my top-secret messenger bag from the hotel safe and met Julie and our audience of two backstage.

I offered a brief lead up, reminding them of the Windows 8 vision and goal of reimagining Windows from the chipset to the experience. I then took out Surface RT from my bag and said, “So, we built a computer.”

We showed them all the bells and whistles. And they were, at least by my account, speechless for a more than a brief moment. We just smiled while they took turns handling the fully functional sample.

Then they had a lot of questions. How much would it cost? What would the OEMs think? What was the battery life? When would it come to market?

Wait…would it even come to market? Katherine asked if it was a prototype or one of those PCs we built to encourage better PC design. We assured them that this was going to market as a first-party device.

Walt asked about one of his most significant issues with Windows PCs, the prevalence of crapware—that software added by PC makers to “enhance the value” by offering trials, advancing PC makers extra hardware (such as buttons to launch specific software), and other add-ons that you probably don’t want but significantly impact performance. On this topic I was extremely confident in answering. The whole point of creating Surface was to show the marketplace what the very best Windows experience could be. Little did I know just how difficult it would be to maintain this seemingly obvious point of differentiation.

While I had many fun moments described here in Hardcore Software, this surprise for a team I respected so much was one of the highlights. I wish I had a photo of the moment.

Invitations were later sent out for a mysterious event in Los Angeles with no hints of what to expect and little notice. The venue was a Hollywood production facility, Milk Studios. More than 200 press, financial, and industry guests made their way to the sound stage the morning of June 18, 2012. Some were still complaining about the short notice, the time, the location, and the lack of detail.

On stage, SteveB introduced Microsoft’s move to hardware, emphasizing that we would not let barriers between hardware and software prevent us from innovating and serving customers. At the end of his section, he unveiled Surface, not the keyboard or the kickstand or any specs. It was the CEO photo op lending all his gravitas to the event.

The stage was small but had been designed with a runway making an intimacy possible, which would become a hallmark of Surface events. The risers were beveled at 22 degrees, the same angle as the kickstand.

As I looked on, I stepped through the demo in my head for the 1,000th time. I had my Surface RT in hand with the screen polished as bright as a signal mirror, using a trick the team had of buffing with hand sanitizer and a microfiber cloth. As Steve was finishing, I touched the power button like I was supposed to and noticed the slightest little screen twitch. I brushed it off.

Rookie f*****g move.

I walked out on stage, as excited as I could ever possibly have imagined. About 10,000 people had worked on what was about to be shown for the very first time. Microsoft had never done such a surprise announcement before. We had never spoken to a room of people without pre-briefing them. Everyone was on edge.

Proudly, after making my way down the narrow runway, I announced Surface, a stage for Windows 8.

Two minutes and thirty-one seconds into the presentation, I was showing the new Internet Explorer. As I touched the screen for the audience, nothing was moving.

I muttered “Oops” and skipped backwards up the runway to the demo table and grabbed the backup. What felt like ten minutes was nine seconds, the longest nine seconds of any demo I had ever done. Who crashes the hardware during a reveal we practiced two dozen times flawlessly? Me. The guy on stage who should have swapped out the machine when the screen twitched. I could physically feel Mike, Panos, and the demo team backstage cringing.

Then the teleprompter went out too.

I kept going. In a bit of stage craft, I introduced the magnetic connector and a nice cover with smooth fabric on the outside, jokingly referred to as “Fine Northwest Polartec®” (it was just nylon), an innovation as well. I attached the cover with a memorable click—that great work of the mechanical team.

I posited to the audience: Why shouldn’t we do something more than cover the screen? To oohs and aahs, some gasps, and then applause, I extended the cover to reveal the full multitouch keyboard. At least that worked I thought.

The main reveals were complete. Phew. I made it through the crash. Backstage the team was despondent but also worried about how I would react. I was so disappointed in my failure to switch devices. That it happened at all still nags at me a decade later.

MikeAng then came on stage and flawlessly demonstrated the Intel-powered GTX, newly named Surface Pro, including writing on screen with a pen. I know that made a certain person back in Redmond very happy. Mike then introduced Panos. Panos detailed the team culture and design process: RalfG, StevieB, Pete Kyriacou (PeteK), Brett Ostrum (BrettO), and so many featured in a video Panos showed before a deep dive on all the technologies in the product.

We concluded and then opened up an area offstage where the press could experience the device. Despite my troubles on stage, letting them see the device in action proved the right call. It wasn’t a free for all, but it was something.

As the press and guests filed to the open area, one of the best in the business stopped me. By all accounts hyperventilating, Joanna Stern then of ABC News had just one question, “How did you keep it a secret?” I like to think we got more done than that, but that was a remarkable accomplishment for Microsoft. Some inside the company would be critical of the secrecy, specifically critical of me, but especially for Surface the nature of the business required it. The test hardware running essentially the same componentry was available to all participating teams.

With the product in hand, I started to use it more regularly. Just after the show I made it a point to use it on my regular places around headquarters. I had a photo snapped while eating lunch at Kidd Valley, my favorite protein and low-glycemic index spot right by campus.

One of the most useful traits of Surface running Windows RT was how even if something failed it took only a few seconds to restart the device and get back to where I left off, especially while under development. That made it more worry-free than a Windows laptop, as we expected and designed for. It was also so fast to resume from standby, and the Metro Mail app connected to Exchange was constantly fetching email even when in standby.

When AnandTech’s  Anand Lal Shimpi reviewed Surface at availability in an extraordinarily detailed and in-depth review of almost 9,000 words, the conclusion was just what we hoped to achieve.

I don't believe Surface is perfect, but it's a platform I can believe in. What I'm most excited about is to see what happens after a second or third rev of the design. [. . .]

If you're ok being an early adopter, and ok dealing with the fact that mobile devices are still being significantly revved every year, Surface is worth your consideration. If you've wanted a tablet that could begin to bridge the content consumption and productivity divide, Surface is it.

My first day of work I could not even get a PC the size and sound of a small refrigerator to connect to a network and run reliably. How far we’d come. I held in my hand a PC that could survive a 10-foot fall, rebooted in seconds, running the best software Microsoft ever created, weighed under 2 pounds travel weight, and easily connected everywhere. I could not be prouder of the work the team had done.

Over the summer as the product wound down we began planning for what would come next. This almost sounds crazy with all that was going on. We had created a machinery for delivering products. It was incredible to be part of. To think that just five years earlier the team, mostly the same team, had just struggled for six years to create one release and we created two releases on time, with none of the stress and strain relatively speaking, and with very high quality. We needed to keep moving.

Windows 8 was a 1.0 release. Perhaps by many accounts it was a 0.9 release, at least for the reimagined part. The Win32 and Intel part of the release was incredibly solid. It was indeed Windows 7 plus more. The team knew we had set out on a multi-year, even decade journey. I knew we would need to operate completely differently.

There would not be a Windows 9 next. In a very short time over the summer, we organized and began planning on “Windows Blue” which is a name I picked to signify it was not another big release. We had a lot of finishing to do. Microsoft never got anything right in the first version, or even the second. We needed to move the team to think about acting with precision in a timely manner so we could, in Microspeak, crank out a release every year.

We created a vision and a schedule for a release to finish in one year. The entire goal of the release was to refine everything from RTM. Refine the platform. Refine the user experience. Refine the core built-in applications and service connections such as SkyDrive and Outlook.com. Decide on a long-term Surface roadmap. The list was very long. Above all we’d fix the things that were so obviously not complete, such as requiring a trip to the desktop to use files or change a PC setting. And yes, we would pay close attention to the reviews as they came out. Though I can say even today, we would not have planned on a full retreat, a redo, or a reset.

Microsoft takes three versions to get things right. We grew up with that reality. The world was moving ahead with or without us. We were determined.

On August 2, 2012, Windows 8 went to manufacturing. We finished 120 days late from our original schedule. The Surface assembly line went into operation. We had a fantastic event on campus. Once again the administrative assistant team outdid themselves.

August was always a busy month at Microsoft. The field sales organization held their annual rally/strategic coordination meeting with almost all of the global staff. Microsoft usually held a company meeting around that time. There was also a Wall Street event or shareholder meeting. These events are the rhythm of a large corporation.

August/September 2012 was also when the company came together to get behind Windows 8. For all the concerns and issues raised as the product progressed, there was an electricity and a sense of excitement. People had seen the doldrums for the past decade. People had seen PC numbers declining. People had seen Microsoft lose in mobile, fall behind Linux, and not yet have a role in cloud computing. While the financials the company was putting up were fantastic by any measure, the buzz was gone. More than a buzz, the company wasn’t a leader. The company wasn’t, for lack of a better word, relevant. We all believed, and genuinely hoped, Windows 8 would offer a chance to be the industry leader again, not just the biggest company, but the most relevant company.

At both the global sales meeting and the company meeting we presented Windows 8 and Surface. The global sales meeting told the story of Windows 8 and why it could win. The field of course were worried about the missing Start menu. I shared some usage data comparing how the Start screen compared to the Start menu. Rather than trying to jam all the functionality into a tiny corner of the screen we showed how people saw the information from Live tiles and made full use of the screen, all maintaining familiar positional memory and muscle memory. It was just one point but the kind of message you send to the field that says, “we heard” and here’s what we consider to be the case. It was not spin. It was what we believed.

What I remember though was not apologizing for the Start screen, but the absolutely incredible and over-the-top reception just when I walked out on stage. I have never felt so much gratitude from 30,000 people all at once. I got to stand there, head bowing, absorbing that gratitude for the team back in Redmond. I stood on stage to applause for nearly a full minute. Humbling doesn’t begin to describe the feeling. I couldn’t say it enough at the time, but I can once again say thank you to the global field team for that welcome and appreciation on behalf of the team. It was so genuine, so unexpected, and just so wonderful.

Just a week later we gathered at the Key Arena in Seattle for the company meeting. While much of the message was the same as MGX, this was headquarters and that means it was mostly the product development groups and company staff. They wanted to see the long version of everything. We did everything to deliver. We had the biggest hardware lineup of Windows 8 PCs MikeAng had ever assembled. PanosP came out and “debuted” Surface once again. Tami Reller walked everyone through all we had done to bring the products to market. We had an amazing and relaxed time. Once again, the reception was just over the top. It was amazing.

We had one more surprise even after all of this. One of the many great joys of being part of creating Surface was the chance to work with a creative team in Hollywood who would bring Surface to television and web advertising. The team was headed by movie producer Andrew Panay, brother of Panos. He assembled a team of creative people to write, visualize, and direct the television spots, the sound stage for the launch event, a worldwide street art campaign, and a plethora of other materials. Andrew was already a seasoned film producer (Wedding Crashers), and their early ideas and pitch made it easy to decide to collaborate with the team on a whole new approach. Andrew, with his keen eye for talent, brought in the director Jon M. Chu to create the marquee television debut for Surface, called Movement. The commercial featured the iconic “Click-In” sound, the hinge action, colorful touch covers, and a troupe of talented au courant dancers. Astute viewers and frequent travelers would later recognize some of the dancers from another well-known spot from Chu, the Virgin Atlantic safety video. Chu would later direct G.I. Joe: Retaliation, Crazy Rich Asians, and In the Heights among his dozens of credits.

We debuted the spot not only for all of Microsoft HQ, but for SteveB as well. Steve did not like surprises, but I had a strong feeling he would love this commercial. It was high energy like Steve, and it sold hard on the attributes of Surface. It was cool. It was hip. It was, well, everything Microsoft was not.

Steve, MikeAng, Tami Reller, Panos, JulieLar, and some others stood by the green room looking on to the full venue of 20,000 or so employees as the lights went down and the commercial debuted. The arena and SteveB went bonkers. Score another awkward high-five moment for me and everyone within a 20-foot radius. The commercial was a moment. The campaign went on to win awards in the advertising world as well, something that Microsoft had not accomplished in ages.

Both events really demonstrated that Microsoft’s headquarters personnel were excited, optimistic, supportive, and a great many were direct contributors. Certainly, we had not experienced anything like that in the past decade of Microsoft.

For the October launch, the Microsoft Store team secured New York City’s Times Square, for a Surface and Windows 8 block party celebrating the opening of the new flagship Microsoft Store. The launch event itself was in an enormous hall downtown. It was filled to capacity. For me personally it would be mostly anticlimactic like all launch events usually were.

Walking around Times Square that night among thousands of people at our own party was crazy. But my mind was not there. I was thinking about what was going to appear with reviews and more. The plan was to only sell Surface RT in Microsoft stores. I was thinking about how we would sell all the Surface devices we built. Our plan was to sell exclusively through only the few Microsoft Stores. I did the math on that, and it was scary. I worried about the aggressive pricing of $499, or $599 with Touch Cover. I was thinking about the Start screen. The desktop. The App Store. The Windows team. Windows Blue. All those people who had done so much, given so much, done such a fantastic job.

I loved this company. I was also exhausted.

About a year before that big launch, I wrote an email that I would deliver on my last day of work at Microsoft. Unsent then, I knew it would go out in the not-far-off future. I stored it in my email drafts folder and held onto a printed copy, undated and unsigned, in an envelope in my Windows logo Timbuk2 bag along with my Surface tablet and my 16 pages of printouts of the vision, staffing, budgets, product reviews, compete info, and the dev schedule that I always had with me. I told nobody about that letter.

I wrote it after a big debate at a cross-division executive meeting over the degree to which Windows 8 would be designed to sell other parts of Microsoft—how many of those colorful Metro tiles would be pushed to customers, not because each was a good product choice but because we wanted to sell them something else. I was being excessively principled, as I was known to be, believing in the primacy of the customer experience. I believed others were being too flippant with our role as a platform provider and a significant reason why we could make a better PC for customers. “It is not the user’s PC, it is our PC,” one exec opined. “Sure, it is crapware, but at least it is our crapware” and “We fought the DOJ so we could own the screen” others chimed in.

Product cycles are emotional roller coasters. Anyone who has ever gone through one knows there are dozens of times when it just gets to be too much. The more seasoned one becomes the more one knows never to do anything precipitously. I knew it was time for a change. I intended to get through the launch. The plans were in place.

I was spent.

Equally clear to me was that the company was spent too. Microsoft needed to do things differently, without me.

On to 108. Epilogue: The End of the PC Revolution

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit hardcoresoftware.learningbyshipping.com