STEVE BALLMER THREW DOWN GAUNTLETS to multiple members of the Xbox team. He directed Robbie Bach, through Rick Thompson, to make a business out of a game console. From J Allard, he demanded "another Microsoft," but in entertainment. Allard was inclined to accept. Ballmer believed any volume softer than a bellow was considered whispering, and had a habit of lurking at the backs of conference rooms, scowling at speakers and smacking baseball bats against his palm.

To accomplish their missions, Allard and Thompson staffed up. Cameron Ferroni and Jeff Henshaw, both friends of Allard's within Microsoft, transferred to the Xbox group to help develop system software for the console. Thompson brought Todd Holmdahl, an executive on the hardware team who had overseen the Microsoft Mouse, over to Xbox to manage hardware integration.

From the "beauty contest" in the summer of 1999 through the rest of the year, the overarching goal for every member of the Xbox team was to build a console with hardware powerful enough to topple Sony. Including a hard drive seemed a strong move in that direction, but remained a point of contention within the group all the way through the console's launch. "I am probably most credited or blamed, depending on the point of view, for having it," says Ed Fries, general manager of the games group.

Fries thought including a hard drive made sense. "What was the point of building a PlayStation clone? If we're going to invest in one new thing, a game console, we need a hard disk. We could use it to stream game content so we could make larger levels, better music, better graphics--swapping stuff in and out of the limited memory in the machine."

Cameron Ferroni agreed. Offering developers a hard drive to work with expanded their options. They could, for example, allow players to rip music from CDs and create their own soundtracks for games. "I was fighting to beat Sony in a head-to-head game war. That's what I woke up every day to do," Ferroni explains.

Rick Thompson acknowledged those benefits, but pushed back: As someone who had made a career out of analyzing hardware, he knew Microsoft could never build the drive for less than $50 over the average four-to-five-year lifespan of a console. Typically, hardware declines in price as newer, better parts supplant it. Hardware also decreases in size. That was why console manufacturers like Nintendo and Sony could reduce the price of their consoles over time. Including a hard drive would make the Xbox immutable in a way that would hurt the group's financials.

"As we got several years into the project," Fries admits, "we could get a much bigger hard disk for $50, but we couldn't get the same hard disk for $25. That $50 anchor made it so you could only reduce the price of the machine so much."

Ultimately, every Xbox included a 10 gigabyte hard drive. Two gigs went to the operating system, leaving players with four-fifths for save-game data, custom music, and extra game content. Ferroni admits he fought for internal storage wielding passion like a blade. His other weapon, though he wouldn't realize it until years later, was ignorance. "I'd only ever worked at Microsoft. I didn't truly understand the value of a P&L."

Every hard drive added $50 to manufacturing cost. Ferroni waved that away, he says, because he focused more on the component's additive value. "It all came down to, was it worth burdening every box with the cost? I was blissfully ignorant and willing to keep fighting for faster loading times, additional content, ripping music--it just made so much sense that that's the way we were going to go."

Ballmer and Bill Gates must have agreed, he realizes. "You realized that they were looking at their P&L, their balance sheet, and looking at whether it would make enough of a difference to be worth it."

Unlike Ferroni, Ed Fries knew a great deal about P&L reports and balance sheets by the time he got involved in Xbox. Despite the financial hit Microsoft took from the hard drive's cost, he believes the battle over including it was worthwhile. "If you're talking to me, you're talking to one of the world's biggest fans of hard disks. You look back at it today: Was it smart or not? All game consoles have hard disks now, or SSDs. I like to say predicting the future is easy, but predicting when is hard. I thought it was clear that hard disks would be in consoles. It was just a question of, was it the right time or not?"

Few developers took advantage of the Xbox's built-in hard drive. Those who did tended to be western studios. Tecmo's Team Ninja, the division responsible for 3D fighting game Dead or Alive 3, was the exception. Fries speculates that most Japanese developers hesitated to embrace boons like faster load times out of fear of upsetting Sony. "That would have been a political issue for them. Japan had a big influence in the market. They were the number-one game creators in the world. But it would be a problem for them in their relationship with Sony and their home market if their game looked too much better on the Xbox. There were some rebels that didn't mind stirring things up, but a lot of studios tread carefully. But I was a fan and was glad we did it."

The Xbox team found other ways to reduce manufacturing costs. After Sony announced that PS2 would play DVD movies, Microsoft had a choice to make. Xbox would run games on DVD, like the PS2, but supporting DVD movies was a touchier matter. "We'd have had to pay four dollars a box, to Sony of all people," explains Robbie Bach, head of the consumer group.

In the 1990s, Sony had entered a format war to determine the successor to VHS tapes. After years of legal battles, the parties united under digital versatile discs, or DVDs. Any company that manufactured the optical media or built a DVD player had to pay a licensing fee to the holders. Unwilling to cede money to their competitor, Microsoft's team voted against allowing playback of DVDs on the console.

Instead, they exploited a loophole. For around $30, Xbox owners could buy a remote control that came with a dongle. Inserting the dongle into the console's controller ports enabled DVD playback. "We paid the DVD video playback license fees on the sale of the dongle and remote, not on the sale of the console," says Xbox co-creator Kevin Bachus. "We saved a significant amount of money. People who wanted to watch DVD movies could buy a remote. It wasn't that expensive, and a handful of the price was the cost of the cheap plastic remote and dongle, and the rest went to licenses."

THROUGHOUT THE DEVELOPMENT OF XBOX, Seamus Blackley and Kevin Bachus stayed laser-focused on games. Every decision made for the console should put developers and players first: The easier Xbox was to make games for, the more developers would sign on to make games, the more players would support the console--rinse and repeat.

Arguably the most important component, they believed, was the GPU, or graphics processing unit. "There was a fight internally to use a custom graphics chip that used a very different graphics architecture instead of polygon-based that Microsoft was pushing," Bachus recalls.

Around the time Blackley joined Microsoft, the company was toying with developing its own graphics solution. Called Talisman, Microsoft's developers saw it as a way to improve 3D graphics by reducing the memory needed to render them. To do this, Talisman would render graphical objects into layers, then composite those layers onto the final scene. Only areas of the screen would need to be updated, thus mitigating demands on memory.

"We let it be known within the industry that we were thinking about moving away from 3dfx, ATI, and NVIDIA toward Talisman, and we just stood back and watched as the emails flooded in," Bachus adds.

Few of the emails were supportive. That worked in Bachus's and Blackley's favor. Disrupting the GPU market would force developers to learn new ways to create and render graphics. Luckily, Intel continued to boost processor and memory speeds. As first-person shooters grew more sophisticated, more players bought GPUs to play them at optimal settings, and companies like NVIDIA and 3dfx packed more video memory into their GPUs. Memory also dropped in price, compelling players to upgrade or buy new machines. In a few years, the problem Talisman had set out to solve, solved itself.

Bachus and Blackley sided with developers against any decision that would make writing games for Xbox a poorer—or, arguably worse, different—experience than writing games for the PC. Their job was to champion developers in any decision involving their console. "There was a sounding board for everything, and that was currency," Bachus explains. "Every time we went into a meeting where Microsoft people surmised that 'X' might be the right way to go, we could say, 'Here's actual evidence that the response to that is not going to be good.'"

As more people from within Microsoft joined the Xbox effort, Blackley grew concerned. He wanted to head off any situation where a majority could overrule a decision he believed would be instrumental to the console's success. He created a group within the team called ATG, Advanced Technology Group, and hired Rob Wyatt, a graphics expert and former colleague on Jurassic Park: Trespasser, to work alongside him. Jason Rubinstein, formerly of Intel, and graphics pro Michael Abrash, who'd worked at Microsoft before going to id Software to co-design Quake's 3D engine with John Carmack, had returned to Microsoft and enlisted in Blackley's ATG.

Knowing that striking early could help him persuade Rick Thompson to see things his way, Blackley wanted ATG to find the right GPU manufacturer for Xbox. He reached out to Dave Kirk, an engineering friend at NVIDIA, and shared plans for Xbox. "We were graphics people and we knew NVIDIA was the clear winner. Everyone who knew anything about graphics knew that," says Blackley. "Because I had worked on Trespasser, and because Rob Wyatt had, and because these other guys we'd hired were really good, Dave Kirk at NVIDIA would talk to us. He'd fly to Seattle and have dinner because he knew we knew what we were talking about."

Kirk confided in ATG's engineers that NVIDIA's upcoming chipset, GeForce, would revolutionize 3D graphics. When Kirk shared details of the conversation with NVIDIA CEO Jen-Hsun Huang, Huang's eyes lit up. NVIDIA was struggling financially in their war against competitors like 3dfx; a contract with Microsoft would go a long way toward stability.

Over time, engineers from Intel and AMD learned of the console. The discussion over which manufacturer to support turned into another of many battles between factions on the Xbox team. Wyatt and Blackley pushed for NVIDIA because more developers wrote games for it. That was when another faction of developers, many of whom had championed WebTV's console, spoke up. A new company called GigaPixel was developing a cutting-edge GPU and receiving help from WebTV's engineers. Rick Thompson supported their suggestion. NVIDIA made great products, but Huang refused to mass-produce and sell graphics chips at a loss. As much as he wanted the deal, he wouldn't risk his company by selling premium goods for next to nothing.

Thompson's thinking was that Microsoft could invest in GigaPixel and own the chipset outright, giving it more control over costs and total agency in manufacturing. Blackley took one look at the architecture GigaPixel proposed for its graphics chip and made a sarcastic quip. Bill Gates looked at him sharply. Blackley had earned his respect over meetings where Blackley's sharp intellect and sharper tongue reinforced to Gates that the former game developer knew his stuff. This time would be different.

"I said something like 'That's not even wrong,'" Blackley remembers, "and Bill said, 'Seamus, you're not going to get by just on reputation with this. You've got to send me an email and show me.' I was like, Fuck. I got reprimanded by Bill. Fuck!"

Unbeknownst to him, Blackley struggled with the same conflicting emotions Alex St. John had experienced years earlier, when he believed his offhand remarks to the press about Microsoft's dearth of experience in publishing would get him fired: Fear, overwhelmed by growing indignation. "I was thinking, Why doesn't he reprimand these idiots who don't know anything, but he's down on me? But this wasn't even a discussion. There are two products. One is obviously superior. The two experts you have who use these products say it's superior. But one guy is trying to say you'll do better with the other one."

Although Robbie Bach had put Thompson in charge of piecing together a business around Xbox, Bach was the one who had to make final decisions. Rendering verdicts forced him to confront his weaknesses. "I didn't play games. Didn't like games, wasn't intuitive about games," he admits.

That blind spot was compounded by what Bach recognized as a second flaw: He was not a technical person. NVIDIA, GigaPixel--to him, both made graphics chipsets. They were identical for all he knew, distinguishable only by price and popularity. He preferred to gather data from members of his team and draw conclusions from there. That didn't help when his team reached a crossroads and looked to him to tell them which way to turn. "Because I wasn't a technical person, and because I wasn't a gamer, and because I tried to absorb too many decisions--that combination led to a process where issues would come up, people would disagree, they'd have different incentives, they'd ask the person up the chain, that person would then ask me, I'd tell them my opinion, they'd go back, people would disagree, and it would come back to me again a second or third time. That doesn't make for good decision-making, frustrates people, and doesn't lead to great outcomes."

From everything Bach had heard and read, GigaPixel seemed promising. Microsoft would invest $10 million in the company and cut a check for another $15 million to help its engineers finish its graphics tech. Should GigaPixel go public, Microsoft would recoup its investments and then some. "At one point, we had reached an unofficial decision that that's what we were going to do," Bach explains. "It would have put us in control of the chipset architecture and enable us to customize and cost-reduce. The problem is, it creates huge challenges on deliverability and predictability: Could a startup company that was unproven get this done in time?"

Huang learned of the imminent deal and called GigaPixel founder George Haber with a warning: Go through with this deal, and Microsoft would own him. Haber ignored him and moved GigaPixel's 33 engineers into the WebTV division's building in Mountain View, California. Haber called Huang back to gloat.

Blackley was crestfallen, but refused to give up. Neither did his friend Rob Wyatt, who pushed back against GigaPixel until he got fired. Work continued on both chipsets; NVIDIA still planned to manufacture GeForce cards for PC regardless of whether it signed a deal with Microsoft. As January 2000 approached, Blackley fought back panic. Bill Gates was scheduled to announce Xbox at the Consumer Electronics Show (CES). When it became clear the team was way behind on finalizing hardware, the announcement was rescheduled for the Game Developers Conference (GDC) in March.

Thompson used those extra two months to reevaluate the deals he had negotiated for hardware. That was when Huang dropped a bombshell: His team could fuse the GPU and processor chips together. Thompson reported this to Bach, who hesitated. Someone had leaked word of Microsoft's deal with GigaPixel and caused a backlash among game developers: As Blackley predicted months ago, they preferred to work with NVIDIA's proven technology, not an unknown, upstart company.

Thompson welcomed Huang back to negotiations. Microsoft would pay $48 per chip for the first five million chips, plus an upfront fee of $200 million. Bach took a few days to consider the terms, then decided on NVIDIA. "The decision we made had cost and far-reaching implications, but if we'd made the other decision, I can't promise you the product would have shipped," he admits of GigaPixel.

The next day, Thompson paid GigaPixel's engineers a visit and informed them their deal was off the table.

Microsoft's last-minute switch to NVIDIA sealed GigaPixel's fate. Two years later, on January 2, 2002, 3dfx Interactive acquired GigaPixel for $186 million, a pittance compared to what they might have been worth had their technology been linked to Xbox.

A similar situation played out as higher-ups deliberated over manufacturers to provide a processor. Microsoft warmed to AMD at first. The night before Bill Gates was scheduled to announce Xbox at GDC, an executive at Intel wooed Bill Gates with a promise that it could give him a better deal. Executives from AMD learned they were out of the running when Gates broke the news at the announcement. "I was intuiting decisions, and on some level, some of these decisions had to be intuited," Bach explains. "But the real problem it created was a lack of permanence to decisions."

Neither decision was without its shortcomings. NVIDIA hit speed bumps in manufacturing its chipsets, and the electronics manufacturer contracted by Microsoft to mass-produce the console's hardware scrambled to set up assembly lines. "We lined up a hardware manufacturer called Flextronics that had three factories: one in China, one in Mexico, and one in Hungary," Bachus says. "They'd build consoles based on our specifications for, respectively, Asia, the Americas, and Europe."

Perpetual turbulence took a toll on the Xbox team. "In my world, there were so many times when I was the guy who understood games and what needed to happen," Blackley says. "Everybody else I talked to was enmeshed in Microsoft politics. I felt like I had responsibility for everything and authority over nothing."

Worse, Blackley encountered resistance or outright mockery when he tried to recruit talented personnel within Microsoft to the Xbox group. Brett Schnepf, a friend of Blackley's, confessed to him that people around the office didn't think much of the project's odds of success. "Brett told me, 'Behind your back, people call it coffin box. The reason people don't return your calls, Seamus, is because it's the coffin box. You're this guy nobody knows, and you're going to wreck their careers.'"

According to Blackley, he approached a newer Microsoft hire with an offer to join up. The recruit went to Thompson and asked for his opinion of Blackley. "I don't think Seamus knows how to get things done at Microsoft."

The recruit went to Blackley and admitted he was having second thoughts because of what Thompson said. "Let me think on it overnight."

"Okay, that's cool," Blackley says. He went out to the parking lot and broke down. Why, he asked himself repeatedly, was he even bothering? All the managers and executives had been pals for 10-plus years. They golfed or played basketball together at clubs. Even those who seemed genuinely engaged in creating Xbox wanted to do it according to organization charts, the Microsoft way. "Ted Hase and Otto Berkes, like Ed Fries, had big careers at Microsoft. They were on a career path. They had seniority levels and all this shit. On days when it looked good for that, they would be really supportive. On days it wouldn't be good, they wouldn't return emails. That's just how it was. And that's okay."

"It wasn't personal, but it was corporate-personal," Bachus agrees of the disappearing-reappearing act some executives pulled. He also expressed frustration with what he felt were avoidable setbacks like flipping between GigaPixel and NVIDIA. "You can't buy your way to success in the console market, but there are table stakes. If you're not willing to spend Billions of dollars with a 'B,' you're not even going to get into the game."

"I think it's totally fair, and I would agree with Seamus if he said that to my face," Fries says of Blackley's belief that Fries and other senior execs at Microsoft could set Xbox aside if needed. However, that didn't make him 100 percent right, either. Blackley could commit all his time and energy to Xbox--pushing for NVIDIA, courting developers, weighing in on adjacent products like the console's controller--because Xbox was his full-time commitment. Fries couldn't do that. His job, the reason Microsoft paid him, was to run the games group.

"It didn't have to do with whether I was committed to Xbox," Fries explains. "I really made Xbox the clear priority for my time once it was approved, and you could really see that, unfortunately, in our PC business. We moved teams off of those projects and released fewer PC games because we were trying to do this impossible thing in 18 months."

"It wasn't really a project yet," Bach agrees of the console throughout most of 1999. "It was an exploration that Rick Thompson was trying to get approved. When it got approved, things got rolling. Go back to May, June, or July, and there was nothing for anybody to be 100 percent committed to, other than a few people who had been asked to do exploration work. When the project gets approved, people either need to say they're in or they're out."

Blackley and Bachus weren't the only ones upset by political warfare among members of the team. One of the problems, Bach knew, was that the group lacked incentive to function as a unit. Leaders of Microsoft's business units had to deliver numbers according to their P&Ls. "All the groups rolled up to one P&L, sure, but everybody sort of had their own little silo. When group A needed something from group B, that meant group B had to sacrifice something in its P&L and in its deliverables to lend help. Like most things in life, people follow the money, so many of our groups followed the incentives."

The larger problem was that the team lacked cohesion. Some held to personal philosophies, such as Blackley's and Bachus's desire to create a developer-friendly platform. But the majority hired from outside Microsoft or recruited from other groups worked toward the often-vague goal of build a console to compete with Sony. "All-Star teams are great because they have lots of talent, but they don't always work super-well together," Bach admits. "It takes a long time for them to build chemistry. That was the case here. I think there was a fundamental structural flaw in the sense we were hiring quickly to get stuff done."

RICK THOMPSON SPENT MOST OF 1999 on a shopping spree. "Part of his exploration in the fall was to figure out a way for us to not have to do this," Robbie Bach explains.

While Thompson was authorized to study the console business and learn how Microsoft could compete, he also had approval to float the possibility of acquisition to Nintendo and Sega. Bach and Thompson dropped buying Sega as quickly as they had considered it. "Microsoft owning a Japanese company in the pachinko business was not a really good place for us to be," he says.

Though Bach was the first to cop to his ignorance of games and hardware, his decision to move on from Sega was prescient. The Dreamcast console launched on September 9, 1999. Eighteen months and a resounding beating from PlayStation 2 later, Sega announced that Dreamcast was its last console. From now on, it would develop and publish software for other platforms.

Response from Nintendo's management to the possibility of selling to Microsoft bordered on outright laughter. Sony's PlayStation had dealt a beating to the Nintendo 64 console, but Nintendo had billions in the banks thanks to continuing sales of Game Boy software like Pokémon and the relative success of N64 titles such as The Legend of Zelda: Ocarina of Time and Rare's GoldenEye 007. "They were polite about it in a perfectly acceptable, appropriate, Japanese way," Bach says, "but they were like, 'What the hell? Why would we work with you?' And they were right. If I'd been them, I don't know why I would have worked with Microsoft."

Thompson made overtures to Electronic Arts and SquareSoft, the Japanese studio hailed for Final Fantasy. Buying EA would secure bestsellers like Madden NFL as exclusives for the Xbox. EA's leaders declined, explaining that they stood to make more money by releasing games on as many platforms as possible rather than being confined to one. They did, however, root for Microsoft to succeed. More competitors vying for market share meant more options for publishers.

SquareSoft executives generously offered to sell Microsoft 40 percent of the company for $2.5 billion. Thompson could have convinced Bach to agree; Windows and Office brought in $1.5 billion a month, and the company had $30 billion in cash, compared to Sony's $5 billion and Nintendo's $7 billion. Ed Fries voted against buying Square. As the manager ultimately responsible for building a catalog of titles, he preferred to grow that catalog more organically.

Thompson delivered reports on each visit to Bach. Microsoft would start with no top-tier developers on their side. The team would have to continue going to developers one by one, as Seamus Blackley, Kevin Bachus, and a few others had been doing for most of the year. "What Rick concluded by December 1999 was, there's a business here. It's not a very good business, and I think if Rick was honest, he'd say, 'It's not even a business I like.' But he did say, 'There's a business here, and this is how we would do it,'" Bach says.

In December, the Xbox team held another meeting with Bill Gates and Steve Ballmer to bring them up to speed. One stop on Thompson's shopping spree had been Dell, Microsoft hoped would manufacture its Xbox. Founder Michael Dell evaluated the business and turned the group down. The news hadn't come as a surprise to industry-savvy team members like Bachus. In a key way, consoles and console games were like razors and razor blades. Gillette lost money on razors, but charged a premium for blades to make the difference. Likewise, console makers like Nintendo and Sony expected to take a loss on their hardware, but made that up by publishing exclusive software.

"Michael Dell said, 'Listen. Every time Sony drops the price on PlayStation, their stock price goes up because people know they're going to sell more of them,'" Bachus remembers. "'Every time I drop the price on a PC, my stock goes down because I lose money.'"

Other manufacturers arrived at the same conclusion as Dell: Why should they lose money building razors for Microsoft when Microsoft would reap profits from selling blades? "We realized nobody was going to build this for us. We were going to have to do it ourselves," Bachus adds.

Not only that, but Microsoft would need to build the sturdiest, sleekest razors if it hoped to compete. "Competing with Nintendo was never about the specs," says Cameron Ferroni, who worked with J Allard and others on system software for the Xbox. "It was about game quality, licenses, and memories. Competing with Sony was always about specs. A lot of developers go where the best specs are. If you want the best games, show them a platform that gives them that."

Gates and Ballmer had reservations, but as long as Xbox delivered on the goal of bringing Windows into the living room, the project could continue. By mid-February, the team realized that dream was impossible. Allard, Ferroni, and the rest of the system engineers were aware of Windows's reputation as a resource hog. After hours of analysis and experiments, they concluded that Xbox and Windows were a bad fit.

Robbie Bach met with Steve Ballmer to update him on Allard's findings. They could still use DirectX, but it would have to be woven into a cleaner, sparser OS--not Windows, but a custom system that gave developers a level of access to hardware that Microsoft's flagship OS did not. "It would take control of the box," Bach says of the custom software, "and there was no way for the operating system to interrupt the game. These things were antithetical to Microsoft's DNA."

Ballmer grew wary. "You mean there'd be two development environments?" he asked.

"Yes, Bach answered. "One for Windows, and another for Xbox."

Ballmer narrowed his eyes. "So, when we have the next Game Developers Conference, how are you going to talk about Xbox and developing for Windows?"

"Well," Bach replied slowly, "it's pretty clear there's not going to be one Game Developers Conference."

Ballmer informed Bach that they would need another all-hands meeting and set one for February 14. It would be what Fries and other leaders at Microsoft called a go-or-no-go meeting. "We'd either go forward or cancel the project, and everybody knew it. We knew that coming out of there, if the project was approved, we would have Bill Gates on stage at GDC in March to announce the project. We'd done a bunch of work to get ready for that--if it happened."