Part IThe Story
Two Signatures on August 24
On August 24, 1996, Gabe Newell did two things that are usually done only after long deliberation, and he did both on the same day. He married Lisa Mennet. With Mike Harrington, a fellow Microsoft veteran who had sold his company stock to cover his half of the bet, he also founded a limited liability company in Kirkland, Washington, about five miles from the Redmond campus where the two men had spent most of their working lives. The record does not say which came first, the vows or the paperwork. The sources say nothing about the ceremony, the guest list, or whether anyone at the reception knew that the groom had become, that morning or that afternoon, an independent game developer. A profile wants that detail, and this one does not have it. What survives is the coincidence. Gaming lore has repeated it so often that it no longer sounds strange. Thirty years later it reads less like a romantic gesture than like a man putting all of his risk into a single day.
The company needed a name. The partners rejected anything that suggested, as Newell put it, "testosterone-gorged muscles and the 'extreme' of anything," which in the mid-nineties ruled out most of the industry's naming conventions. The shortlist reads like a box of rejected band names: Hollow Box. Fruitfly Ensemble. Rhino Scar. They chose Valve, a plumbing word for a device that regulates flow. It said nothing about guns or speed or adolescent fury. It said something about pipes, and about who controls what passes through them, and it is hard not to read that, after the fact, as the most accurate corporate name in American software. Nobody at the time could have known how accurate. Newell may not have known himself.
He was thirty-three. He had spent thirteen years at Microsoft, had produced the first three releases of Windows, and by any reasonable measure could have retired. He said later that he chose instead to work with "other really smart, motivated, socially orientated people to create product that would affect millions of other people," because that seemed like more fun. He also said, closer to the bone, that Microsoft "was missing the opportunity" the internet offered and that he "was willing to sort of put my money where my mouth was." That was literally true. Newell and Harrington paid for Valve's first game themselves. No venture firm ever owned a slice of the company, and none does now.
He had a phrase for what he saw coming in those years. He called it a sea change. He meant it as a figure of speech. It did not stay one.
Three decades later, Valve's revenue was estimated at $5 billion a year, at a 40 percent margin, from a company that employed 336 people in 2021, which by most accounts is the size of a mid-sized law firm. Newell was estimated to own a little more than half of it. Forbes put his net worth at $11 billion, which made him the richest man the video game business has produced. He had also spent much of the previous fifteen years trying, with mixed success, to get out of his own company's way. He had left game design because people kept agreeing with him. He had left Seattle for New Zealand and then for open water. The man who founded Valve on his wedding day had become someone the company seemed to orbit rather than someone who ran it, and that may have been the plan from the start.
By the Numbers
The Valve Ledger
$11BNewell's estimated net worth (Forbes, 2025)
50.1%Newell's estimated stake in Valve; employees hold the rest (Forbes, 2024)
$5BValve's estimated annual revenue in 2023, at a 40% profit margin
336Valve employees in 2021
13Years Newell spent at Microsoft before founding Valve
50–70%Steam's share of the downloaded PC games market in 2011
$2,391,932Arbitration award Valve won from its former publisher, Vivendi
A Telegram Boy in Redmond
The Aspen Times recorded the birth on page four of its November 9, 1962 edition, in a column called "Around Aspen": a son, Gabe Logan, born November 3 at Aspen Valley Hospital to Donald Newell. The ski town did not keep him. He grew up in Davis, California, a university town in the Sacramento Valley, and attended Davis Senior High School. Kotaku would later publish a photo of him from his first year there, which the internet treated as an archaeological find.
Two jobs from those years deserve more attention than they usually get. He delivered newspapers. Then he became a telegram messenger for Western Union, carrying typed messages by hand from the wire to the door. That is the last mile of a distribution system, done on foot, at the end of the era when that system meant anything. Within a generation, the boy who had walked telegrams to strangers would own the most important pipe in PC gaming, a pipe with no doors and no messengers.
He started programming in high school, "when programming was not an established career path," and he assumed he would become a doctor. Decades later, asked on Reddit what had pulled him into computing, he did not offer a mission statement. "When I started programming, it wasn't really an established career path," he wrote. "I did it because it was fun." Asked what got him interested in games, he gave seven words: "Playing Trek on a mainframe using punch cards." It was a Burroughs mainframe version of the 1971 Star Trek game, the first game he ever played, and in 2011 he still named it as one of his three favorites, alongside Doom, which convinced him games were the future of entertainment, and Super Mario 64, which convinced him they were art.
In 1980 he enrolled at Harvard to study programming. He did not stay. His brother worked at Microsoft, which was then a small company that had not yet become what it would become, and Gabe visited. In the version Newell gave PC Gamer in 2025, Steve Ballmer got irritated at the kid hanging around the office and told him: "If you're going to be hanging out here, why don't you do something useful?" Ballmer suggested he drop out. He did. Newell later called Microsoft the best place in the world to learn to program at the time, and the record suggests he meant it. He stayed thirteen years as a programmer and then a technical executive, and he produced the first three releases of Windows, which is a sentence about operating systems that also describes a place in history.
The break came from outside. In late 1995, Doom, a first-person shooter made in Mesquite, Texas by a company called id Software, was estimated to be installed on more computers worldwide than Microsoft's brand-new Windows 95. Newell led the effort to port Doom to Windows 95, work credited with making Windows a viable gaming platform at all. But what struck him was not the game. It was how the game had traveled.
— Gabe Newell, on id Software and Doom[id] ... didn't even distribute through retail, it distributed through bulletin boards and other pre-internet mechanisms. To me, that was a lightning bolt. Microsoft was hiring 500-people sales teams and this entire company was 12 people, yet it had created the most widely distributed software in the world. There was a sea change coming.
He had another irritation, more technical and more particular to him. Game developers kept building bespoke interfaces for graphics acceleration, each studio writing its own, so that the PC was less a platform than a pile of private arrangements. For a man who had spent a decade building the standardized layer between programs and hardware, this looked like waste.
Then a colleague showed him the way out. Michael Abrash was a Microsoft graphics programmer of near-legendary reputation who left Redmond for Texas to work on id's next game, Quake. He was the first man Newell watched trade a large company's security for a small company's ambition, and he would turn up on Valve's own roster years later, which tells you something about how loyalty works in this story. Abrash didn't recruit Newell. He just left first, and Newell paid attention.
The Partner Who Took the Money
Mike Harrington was two years younger than Newell, born in 1964, and had followed a parallel path: programmer at the game studio Dynamix from 1985 to 1987, then nine years at Microsoft as a designer on Windows NT. He was the more conventional of the two founders, in the sense that he had already worked in games and then chose the safer job. In 1996 he sold his Microsoft shares to fund his half of Valve. His wife at the time, Monica Harrington, became the company's first marketing strategist. In 2024 she told PC Gamer something that sounds like a ghost story about the industry: "There's an alternate universe where Half-Life disappeared after release."
That alternate universe was close. Valve licensed id's Quake engine, which it rebuilt so thoroughly that it got a new name, GoldSrc, and set out to make a first-person shooter with elements of horror. Then it struggled to find a publisher. Sierra On-Line eventually signed it, on terms Valve would spend years regretting. Half-Life shipped in November 1998 and was a critical and commercial success of the kind that reorganizes a genre. In 2014, IGN wrote that the history of the first-person shooter "breaks down pretty cleanly into pre-Half-Life and post-Half-Life eras." Scripted sequences, a seamless narrative with no cutaways, a silent physicist with a crowbar: the vocabulary of the next fifteen years of shooters started there.
There is a detail about the game's ending that the newlyweds of August 1996 supplied themselves. Half-Life's final boss, a monstrous infant-like creature floating in an alien world, was inspired by the birth of the Newells' first son, Gray, in the late 1990s. The couple considered childbirth the most frightening thing they could imagine at the time, so the climax of a game full of soldiers, aliens, and radioactive collapse was built around a baby. Gray Newell grew up to become a professional race car driver, which suggests his parents' fears were well calibrated.
For Harrington, the success settled something.
— Mike Harrington, on Half-LifeAt Microsoft you always wonder, "Is it me being successful or is it Microsoft?" But with Half-Life I knew Gabe and I had built that product and company from scratch.
The question was answered, and he took the answer home. On January 15, 2000, Harrington sold his stake in Valve to Newell and left to spend time with his wife. Newell's account was simple: Harrington did not want to risk another project. Afterward Harrington built a respectable second career. With his friend and former colleague Darrin Massena he co-founded Picnik, a photo-editing service that Google bought in March 2010. He left Google a year later, started Catnip Labs with Massena, served as CTO of the Committee for Children, and then as CTO of a biotech startup called Amplion. It is a good life. It is also exactly the life of someone who stops after proving the point.
Newell's response to the same question was to make sure nobody could ever ask it about him. He bought Harrington out and kept buying. Forbes estimates of his stake have shifted over the years: at least a quarter in 2021, 50.1 percent in 2024, "at least half" in 2025, with the rest held by employees. But the direction has never been ambiguous. The two founders stood at the same fork and went different ways. One wanted to know the success was his. The other wanted the success to stay his.
Leverage Is the Willingness to Leave
The publishing contract Valve signed with Sierra in 1997 gave Sierra the Half-Life intellectual property and exclusive rights to publish future Valve games. For a developer, this is a standard trap. Valve kept doing Valve things inside it. In 1998 it acquired TF Software, the group behind the popular Team Fortress mod for Quake, and remade the game on GoldSrc. It released its engine's development kit to the public, which produced a flood of user-made mods. One of those, Counter-Strike, was so popular that Valve bought its developers and turned it into a standalone product. The company was already acting as if anything the community built could become part of its catalog. It had not yet realized the catalog could become a store.
The first attempt to control the pipe failed. In January 2000, Valve and Cisco Systems announced PowerPlay, a set of router-level protocols meant to shrink online latency so far that a player with 1,000 milliseconds of ping could supposedly compete with someone on a local network. The plan required internet providers to buy approved Cisco hardware and give PowerPlay gaming packets priority over everything else. Newell was frank about who would pay. "The ISPs are going to need to spend a fair amount of money to be compliant with PowerPlay," he said. "But how they get that back is up to them." The planned dial-up service never launched. About twelve months after it was announced, PowerPlay was quietly abandoned. The lesson, if Newell took one, was that you cannot own somebody else's pipes. You have to build your own.
In 2001 Valve went back to Sierra to renegotiate, and its strategy came down to one threat: Valve would stop making games. A former executive later described the pitch as Newell and the team pivoting away from game development entirely. "Valve would never ship another game." The threat had backing. Amazon had offered Valve a partnership to build a digital storefront, and Valve used the offer as a bargaining chip. Sierra, faced with the possibility that its most valuable developer would simply go do something else, amended the deal. Valve got the Half-Life IP back, along with online distribution rights for its own games.
The second clause turned out to be worth more than the first. At the Game Developers Conference in March 2002, Valve unveiled Steam.
The fight that followed nearly killed the company. In August 2002 Valve sued Sierra and its parent, Vivendi Universal Games, for distributing its games to internet cafés against the contract, withholding royalties, and pushing Counter-Strike: Condition Zero past the holiday season. Vivendi countersued, arguing that Valve had misrepresented itself in the 2001 renegotiation because it was already building Steam to get around the publishing agreement. Vivendi wanted the Half-Life IP and a ruling barring Valve from using Steam to distribute Half-Life 2. Scott Lynch, Valve's chief operating officer, later said the countersuits, if they had succeeded, would likely have bankrupted the company.
They did not succeed, and the reason is almost too neat. Discovery produced a batch of documents in Korean. A new Valve intern, from South Korea, could read them. The translated documents indicated that Vivendi had directed the destruction of evidence. When the material reached the court, the judge moved to block Vivendi's countersuits, and the case turned. On November 29, 2004, Judge Thomas Samuel Zilly of the U.S. District Court for the Western District of Washington ruled that Vivendi and its affiliates were not authorized to distribute Valve's games through cyber cafés, and that Valve could recover copyright damages regardless of the contract's limitation-of-liability clause. The parties settled on April 29, 2005. An arbitrator awarded Valve $2,391,932.
The story had a coda that reads as both petty and principled. Vivendi's games business ended up inside Activision Blizzard, which paid $1,967,796 and refused the remaining $424,136, claiming it had overpaid in earlier years. In April 2009 Valve sued for the difference. Four hundred twenty-four thousand dollars was nothing to a company already dominating digital distribution. Valve sued anyway.
There is a symmetry in all this that the participants probably did not notice. Steve Ballmer had once told a young visitor loitering at Microsoft to do something useful. Twenty years later, the company that visitor founded was saved by an intern doing something useful. Valve has always given unusual weight to whoever happens to be in the room.
No Deadline, Unlimited Budget
Valve began Half-Life 2 about six months after the first game shipped, on a new engine built from scratch called Source. Newell set no deadline. He gave the project a "virtually unlimited" budget and promised that if the money ran out, he would pay for it himself. Geoff Keighley chronicled the result in his digital book [The Final Hours of Half-Life 2], and the shape of the story is familiar to anyone who has watched an ambitious creative project: five years of advanced physics, more story, more character, and a sequel to a genre-defining game that managed, when it shipped in 2004, to define the genre again. By 2011 it had sold 12 million copies.
During those same years, Newell spent months of his own time on Steam. This detail gets lost in the mythology of the game, and it shouldn't. While his studio built the most anticipated sequel in PC gaming, its founder was working on plumbing. Steam launched in September 2003, at first as a way to deliver patches and updates to Valve's online games. Then Valve made it mandatory: you could not install Half-Life 2 without it. Players resented this loudly at the time. The resentment is forgotten now. The install base is not. Steam started with only Valve's games and later became a publisher for everyone else's, and by 2011 it controlled between 50 and 70 percent of the market for downloaded PC games. Forbes estimated Valve grossed $70 million in 2005. Within six years estimates of the company's value ran between $2 billion and $4 billion. The game made Valve famous. The pipe made it rich.
The project cost something human. Afterward, Newell said he became "obsessed" with improving Valve's work-life balance, which is the kind of thing founders say after watching their people burn out on something they asked for. Around then his own body was also asking for repairs. He had Fuchs' dystrophy, a congenital disease of the cornea, and it was cured with two cornea transplants, in 2006 and 2007. The man who had financed an engine built to render light and surfaces more convincingly than any before it had his own lenses replaced while the engine shipped.
In May 2003, in the middle of all this, the Seattle Times ran a home-section feature about Newell's beach house on the Long Beach Peninsula. The headline: "Tuned To The Dunes: A beach house honors the magic of surprise, the gift of time." Nobody at the newspaper was thinking about game releases. Players were, and by then they had a name for Valve's relationship with the calendar: Valve Time, an industry joke for the gap between when Valve said something would ship and when it did. Valve adopted the term. It tracked the discrepancies on its public developer wiki, used the phrase in delay announcements, and blamed the gaps on its preference for team-driven decisions over corporate deadlines. Valve Time also included things that shipped early, a point defenders tend to make and critics tend to skip.
After five years on one game, Valve announced it would stop doing that. It would release shorter games more often: episodes. Half-Life 2: Episode One came out in 2006, the first of a planned trilogy. Episode Two followed in 2007, packaged with Team Fortress 2 and Portal, a puzzle game grown from a DigiPen student project called Narbacular Drop, in a bundle called The Orange Box. It was arguably the strongest single release in the company's history. Episode Two ended on a cliffhanger.
The third episode was canceled.
A Company Without Bosses
When Valve moved to Bellevue in 2003 and reincorporated as Valve Corporation, it had 60 employees. Up to then it had used the hierarchy most game studios used, and for a practical reason: physical games went through publishers, and publishers demanded deadlines. Once Valve published itself through Steam, the hierarchy started to look like a cost with no benefit. After Half-Life 2, Valve flattened. Outside executive management, it has no bosses. It uses open allocation, which means employees can move between projects whenever they like, voting with their desks.
Doug Lombardi, Valve's longtime marketing head, described the result in a 2011 interview: "Nobody writes a design doc and hands it to somebody and says, 'you go build this'. It's the teams that are coming up with the ideas and pushing in the directions that they want to take the product." Newell's version was blunter. The structure lets people work on whatever interests them, he said, but it requires them to own the product and the mistakes. It works well for some. "There are plenty of great developers," he said, "for whom this is a terrible place to work."
This was not a utopia, and it was not quite flat either. Seniority and relationships created gravity. De facto project leads became, in Valve's own vocabulary, "centralized conduits" for information. A process called Overwatch collected feedback from senior members, and teams could use or ignore it. Compensation ran on stack ranking, with employees evaluating one another. In June 2012 Valve hired the economist Yanis Varoufakis to study the economies inside its games, and he spent part of his time there writing publicly about why Valve's management structure made sense in a world of corporations. The company had enough confidence in its own peculiarity to hire an economist to explain it.
The culture's real engine was empirical, and it was the part that worked best. Valve playtests constantly, from the earliest stages of development, and iterates on what it sees. Its stated belief is that "all game designers are, in a sense, experimental psychologists." The writer Chet Faliszek described what that meant for a designer's ego. When testers missed what a level was trying to teach them, he blamed the testers, until enough of them failed the same way. "By the third or fourth time, all of a sudden you're realizing, 'I'm an idiot. This is pretty obvious this doesn't work. It's not their fault, it's our fault.'" On Left 4 Dead, a texture change made every single tester miss a ladder and get stuck. When something exciting happened by accident in a test, the team tried to make it happen for every player.
Newell drew the economic conclusion. He contrasted Valve's method with that of Warren Spector, whose open-ended games were built to be replayed with different outcomes. "You spend all of this time to build stuff that most players will never ever ever see," he said. "If only one per cent of your customers see this cool thing that takes five per cent of your development budget, that's not a good use of resources." This was the operating-system mind, the Windows mind, applied to art: build for the median user, measure, iterate.
By 2011 Valve employed about 250 people, and Newell claimed it was more profitable per employee than Google or Apple. Its experiments outside its core sometimes ended quickly. In January 2008 it bought Turtle Rock Studios, renamed it Valve South, shipped Left 4 Dead and its sequel, and let it spin out again in March 2010. In December 2012 it acquired a two-person studio to open a San Francisco office and closed the office in August 2013, having decided it was of little benefit. In July 2013 it launched Pipeline, a project putting ten high school students to work on game content, partly to answer teenagers' questions about the industry and partly to find out whether Valve, which admitted it was "not good at teaching people straight out of school," could learn to do so.
The costs piled up quietly. Between 2012 and 2021 Valve's headcount hovered around 350, and its output of finished games slowed sharply. In January 2023 the investigative outlet People Make Games published a report, based on interviews with current and former employees, concluding that the flat structure and stack ranking had produced a weak release record and a lack of diversity. Freedom to choose your work turned out to include the freedom to avoid hard work. Some projects died because no one could recruit anyone to them.
The Problem of Being Agreed With
Portal 2 almost didn't have portals. After the first game's success, Valve made the sequel a standalone product, partly because developers inside Valve were pushing to work on anything with Portal in the name. The team grew from the seven or eight people who made the original to 30 or 40. Early on they built around a new mechanic called F-Stop: an "Aperture Camera" that could photograph an object, store it, and put it back down rotated or scaled, in a 1980s-set prequel without Chell or GLaDOS, with Aperture's founder Cave Johnson uploaded into an artificial intelligence commanding a robot army. Playtesters liked F-Stop. They were disappointed there were no portals. Around October 2008, Newell told the team to reconsider.
It was one of the last times his direction set the course of a Valve game, and possibly the best. The redirected Portal 2 brought back Chell and GLaDOS and recovered Cave Johnson as a set of recordings voiced by J. K. Simmons. Stephen Merchant was cast as Wheatley, after the writers assumed he'd be unavailable and asked The IT Crowd's Graham Linehan to help them reach Richard Ayoade. The game had 13,000 lines of dialogue. Kim Swift, who designed the original Portal, left Valve for Airtight Games halfway through development. That was its own small verdict on the system: the person most responsible for a hit, free to work on anything, chose to work somewhere else. Before release Newell called Portal 2 "the best game we've ever done." It shipped in April 2011, won BAFTA's Best Game, and had sold more than four million copies by May 2012. In a 2017 Reddit AMA, asked for his favorite Valve game, he still said: "I think Portal 2 is our best single-player game."
Then he stepped back. Josh Weier, Portal 2's project lead, explained why years later. Too many developers agreed with Newell's ideas simply because they were Newell's. He wanted, in Weier's account, "to be part of the team and come up with ideas," and found he couldn't. In a company designed to have no bosses, the founder had become a boss by gravity alone. His solution was to leave the room.
That decision says more about him than anything he built. It also produced a lost decade. Valve shipped fewer games through the 2010s and bet on hardware. Newell wanted Valve to be more like Nintendo, a company that develops games and hardware together, which, he argued, is how something like Super Mario 64 gets made. In 2013 he laid off staff to move the company's hardware focus from augmented reality to virtual reality. The canceled projects make a grim list: several Half-Life projects, including Episode Three. Left 4 Dead 3. A Soulslike. A voxel-based game called A.R.T.I. A VR project called SimTrek with members of the Kerbal Space Program team. A headset code-named Vader, judged too expensive for consumers. The designer Robin Walker said that the pile of projects that never gained traction, with no shared vision connecting them, damaged morale. Walker said a Half-Life 3 had been a "terrifyingly daunting prospect."
VR offered a way back. Valve found that Portal's portals were disorienting in a headset, so it chose Half-Life. Full development of a VR Half-Life began around late 2016 with the largest team in the company's history. Half-Life: Alyx shipped in March 2020, was widely called VR's first killer app, and forced Valve to confront the premise of its own culture.
— Robin Walker, Valve designer, on the making of Half-Life: AlyxWe sort of had to collectively admit we were wrong on the premise that you will be happiest if you work on something you personally want to work on the most.
The fix was not a return to hierarchy. It was short-term, studio-wide goals: a shared target big enough to pull a self-directed company in one direction. Two years before Alyx, the opposite approach had produced Artifact, a digital card game based on Dota 2 with a design by Richard Garfield, who created Magic: The Gathering. It launched in November 2018 with unusual pay-to-acquire mechanics and lost 95 percent of its players within months. Newell later called it a "giant disappointment." Valve abandoned the reboot in April 2021.
He has been open about the regrets. Asked in 2017 why he didn't count Half-Life among his favorites, he wrote: "The issue with Half-Life for me is that I was involved in a much higher percentage of the decisions about the games, so it's hard for me to look at them as anything other than a series of things I regret." Pressed, he described game-making as a system of opportunity costs: "Anything in a game is a sacrifice of things not in the game." Then, deadpan, two more words: "And Xen."
In the same session a moderator compiled the community's most-requested questions. The first asked about the status of Half-Life 3. Newell answered: "The number 3 must not be said." When a commenter predicted he would dodge, he replied: "See how I threaded that needle. Pretty slick."
Service, Not Price
At a technology conference in Seattle in 2011, Newell set out a view of piracy that the rest of the media industry spent the next decade arriving at slowly. You don't beat pirates with anti-piracy technology, he argued. You beat them with a better option. His example was Russia, where piracy was rampant and Steam was succeeding anyway. In his view piracy was a service problem, not a price problem. The argument was self-serving, since Valve owned the service. It was also correct.
The same logic explains his public attacks on everyone else's platforms, and also his reversals. In 2007 he called developing for Sony's PlayStation 3 "a waste of everybody's time." In June 2010 he appeared on stage at Sony's own E3 keynote, acknowledged what he had said, praised the PS3's openness, and announced Portal 2 for the console with Steamworks support, which he said would make it the best console version anywhere. PS3 owners could link a Steam account and play the PC version for free. Underneath the reversal was Valve's frustration with Microsoft, whose Xbox 360 policies made patches and new content difficult. Newell called Xbox Live a "train wreck." In July 2012 he said Windows 8 was "a catastrophe for everyone in the [personal computer] space," worried that Microsoft was closing the PC the way Apple had closed its products. By September 2013, at LinuxCon, he said Linux and open source were "the future of gaming." The man who had produced the first three Windows releases had started building an exit from Windows.
The exit took ten years and nearly failed. SteamOS, a free Linux-based operating system, was announced in September 2013. Steam Machines, living-room PCs from Alienware, ZOTAC, and CyberPowerPC, arrived in 2015, and by June 2016 fewer than half a million had sold. The original Steam Controller, delayed over its own design problems, stayed on sale until late November 2019 and ended in a $5 clearance. It looked like an expensive vanity project. It turned out to be groundwork. In February 2022 Valve shipped the Steam Deck, a handheld PC running SteamOS with a compatibility layer called Proton that let most Windows games on Steam run on Linux without modification. That spring Newell personally delivered some of the first units to buyers. The founder of a company valued in the billions went door to door with packages, like a telegram boy.
By 2023, Forbes Australia reported, Steam accounted for about 60 percent of Valve's $5 billion in revenue, double its 2019 contribution. The figures underneath are almost abstract: more than 3,400 games by July 2014, more than 150 million registered accounts by January 2018, 67 million monthly active users in August 2017. Valve's games, by contrast, had become nearly optional to its finances. The company itself acknowledged that Steam's success meant it no longer depended on its games succeeding, which was liberating and, by some accounts, corrosive.
Owning a pipe this big brings scrutiny. On August 1, 2012, Valve changed the Steam Subscriber Agreement to ban class-action lawsuits from users. The same announcement revealed Valve S.a.r.l., a Luxembourg subsidiary that for a while let it sell games to British customers without paying the full 20 percent value-added tax. In December 2016 an Australian court fined Valve A$3 million over its claim that consumers had no right to refunds, and Valve's appeals ran all the way to the High Court of Australia, which dismissed them in April 2018. In May 2021 Wolfire Games, the developer behind Humble Bundle, filed an antitrust suit accusing Valve of using its "gatekeeper role" to take "an extraordinarily high cut from nearly every sale." In November 2024 the case was certified as a class action covering any developer affected by Valve's revenue share. In June 2024 a British children's digital rights campaigner filed a claim seeking £656 million on behalf of 14 million UK users. In September 2024, facing a wave of mass-arbitration claims that would have forced it to pay filing fees case by case, Valve reversed its earlier position and required disputes to go to court in King County, Washington. In February 2026, a jury sided with Valve on every count in its suit against companies tied to the patent holder Leigh Rothschild.
That summer Valve launched a redesigned Steam Machine at $1,049, sold through a randomized reservation system because of global memory shortages. Then came a strange event. In late summer 2026, about 12 terabytes of data were pulled from old Steam servers that had been left unprotected, including games, betas, and prototypes from 2003 to 2013. Among the material was more footage of F-Stop, the Aperture Camera from the Portal 2 that never shipped, and Cave Johnson dialogue about the nature of Companion Cubes. The archive of abandoned ideas had, in the end, been distributed too.
Learning From the Customers
He has a nickname, and it came from the plumbing. In the gaming community, Newell is Gaben, a name derived from his work email address. An internal routing string became a cult figure. The customers built the persona, and Newell, to his credit and somewhat against his nature, has tried to live up to it.
"[Fans] hug me when they run into me," he told the Washington Post in 2014. "I'm not a hugging person, but that's what they want." The first time it happened in public, his kids were with him. They were fine with it. He wasn't. "'Dad, roll with it.'"
He rolled with it. He was, by multiple accounts, a fan of My Little Pony: Friendship Is Magic. For his fiftieth birthday, in November 2012, 4chan sent him a present, as Polygon reported. In 2018 he recorded a deadpan voice pack of himself for Dota 2, the game that came from Valve's 2009 hiring of IceFrog, the pseudonymous developer who had been maintaining the Warcraft III mod Defense of the Ancients. Valve later spent part of 2012 settling with Blizzard over who owned the word "Dota." In the 2017 AMA, asked how he liked his steak, Newell answered: "Medium rare." Asked for his favorite non-Valve game of 2016: "My co-workers just told me it's Plants vs Zombies, and then they laughed at me." Asked about a rumor from an anonymous source that Half-Life 3 had been canceled: "I personally believe all unidentified anonymous sources on the Internet."
Under the jokes was a philosophy of communication that Valve stated in the same session. The company does not announce games the way others do because, as Newell put it, "our decision making is way more conditional than most other companies. The one thing we won't do is waste our customers time and money, which means we will cancel or change stuff much later in development." A Valve colleague added that Valve prefers "to communicate through our products," shipping updates instead of blog posts, and accepts the cost: from the outside, the silence can feel like an echo chamber, a company that isn't listening.
When Valve does talk, it talks in puzzles. In March 2010, a patch to the original Portal added an achievement that required players to manipulate in-game radios, which led to Morse code, slow-scan television images, and a BBS phone number spitting out ASCII art. Adam Foster, the Valve developer behind it, ran the BBS from his own home phone line because Valve's offices were too modern to support the protocol, and he estimated the whole alternate-reality game cost under $100. The following week, Newell accepted the Game Developers Conference Pioneer Award. His speech ended with a fake Blue Screen of Death on the screen behind him, signed by GLaDOS. The man who had produced the first three releases of Windows took his industry's lifetime honor and ended it with a staged Windows crash. In April 2011 players solving the Potato Sack puzzles unlocked Portal 2 about ten hours early.
His peers' recognition came in the same register. In 2013 he entered the Academy of Interactive Arts & Sciences Hall of Fame and received the BAFTA Fellowship, an honor previously given to Alfred Hitchcock, Charlie Chaplin, Stanley Kubrick, Steven Spielberg, and Shigeru Miyamoto. That year he shared it with Alan Parker and Michael Palin. Later that winter, at the D.I.C.E. Summit, he announced that he and J. J. Abrams were collaborating on a Half-Life or Portal film. Asked about the films in 2017, he wrote: "Yep. They're coming."
One exchange from that AMA belongs in any account of the man. A Redditor wrote in to thank Newell for something he had said years earlier, when they met at a racetrack and the fan pitched ideas for a zip-line crowbar. Newell asked him, "So what's stopping you from making a game or a mod?" Those words, the fan wrote, had become his mantra whenever he feared failure. There is now an unauthorized biography, The Valve Effect: An Independent Biography of Gabe Newell, but the truest account of Newell's influence may be scattered in posts like that one, by strangers who took a stray question as permission.
Sea Change
In March 2020, the month Valve shipped Half-Life: Alyx, Newell was visiting New Zealand with friends when the pandemic closed the world's borders. He stayed in Auckland. PC Gamer called him a "COVID refugee." He and others organized a free concert to thank the country, We Love Aotearoa, with musicians from across New Zealand and VR stations where attendees could try Alyx and The Lab. A second lockdown pushed the event from August to December. In October 2020 he applied for permanent residency. He did not bring Valve with him. By 2021 he was back in Seattle, but not for long in the usual sense. Since the pandemic, he has lived mostly at sea.
The ventures that followed look scattered until you line them up. In 2020, with longtime Valve employee Yahn Bernier, he started a racing team, the Heart of Racing, to raise money for children's charities in Seattle and New Zealand. He owns a marine research organization, Inkfish, with several ships and submarines. In November 2022 Inkfish bought the Hadal Exploration System, a private deep-sea exploration platform, from the undersea explorer Victor Vescovo, including the support ship Pressure Drop. In June 2023, while the world followed the search for the missing submersible Titan, reports noted that the only crewed submersible able to reach its depth belonged to Newell.
He also co-founded a neuroscience company, Starfish Neuroscience, in 2022 with Philip Sabes to develop neural interfaces. In May 2025 it said its first chip would arrive late that year. He had flagged this five years before it existed. "Personally I'm looking at research in brain-computer interfaces," he wrote in the 2017 AMA, between answers about steak and snack bars. In 2018 he gave $20 million to OpenAI and served as the sole member of what PC Gamer later called its "informal advisory board." Around 2013 he co-founded and funded Foundry10, an education research group that used hip-hop and auto mechanics to reach students. In August 2025 he bought Oceanco, the Dutch yacht builder constructing the roughly $400 million vessel he was set to take delivery of later that year, explaining, according to PC Gamer, that he "respects the sea."
Valve went on without him in the way he seems to have wanted. By 2024 he was less involved and spending more time on personal projects. In June 2026 he bought a 20,000-square-foot Florida mansion for $70.8 million: seven bedrooms, thirteen bathrooms, an eight-car garage, a dock, a boat lift, and a private tunnel to the beach. A man who has spent his life building passages between places bought a house with one built in. That August, at The International, Valve's Dota 2 championship, he appeared much thinner, and the internet decided he looked like Hugh Jackman. He gave the opening speech.
The sea change he predicted in 1995 was a change in how software moved: from boxes and sales teams to bulletin boards and then to the pipe he would own. He was right about that in nearly every particular. He did not predict that he would end up, three decades later, on actual water, with submarines, among the richest people in the world, still unwilling to say the number three.
One more launch remains. In 2020 Newell worked with Wētā Workshop, the New Zealand effects company, and Rocket Lab to send a small figure into space: Gnome Chompski, the garden gnome from Half-Life 2: Episode Two, the last chapter of the story Valve ever released before it stopped telling it. Newell pledged a dollar for every person who watched the launch video within twenty-four hours, with the money going to the pediatric intensive care unit at Starship Children's Hospital in Auckland. Half-Life had once ended with a monster shaped by a parent's fear of a child arriving. The next chapter never came. What came instead was the gnome, clear of the atmosphere, higher than any sequel had gone, with a dollar attached for every person watching from below.
Part IIThe Playbook
What follows is not a biography of Valve's methods so much as a list of the bets Newell made repeatedly, often against the industry's advice, sometimes against his own company's interests, and occasionally against himself. Some paid off enormously. Some cost Valve a decade. The principles below are drawn from both. Read together, they describe a founder who understood distribution earlier than nearly anyone in entertainment, and who spent the second half of his career working out how to keep that understanding from turning into control he couldn't let go of.
The principles
- 1.Watch where the software travels, not where it sells.
- 2.Make your exit credible before you negotiate.
- 3.Build the pipe while everyone is watching the product.
- 4.Own the clock by owning the cap table.
- 5.Compete with free on service, not enforcement.
- 6.Treat the playtester as the experiment, never the idiot.
- 7.Spend where the median customer will look.
- 8.Leave the room when your voice becomes a verdict.
- 9.Give a flat organization a forcing function.
- 10.Speak through shipped product, and cancel late.
- 11.Attack publicly, reverse publicly.
- 12.Buy insurance against your own platform landlord.
Principle 1
Watch where the software travels, not where it sells
Newell's founding insight was not about games. It was about distribution. In late 1995, sitting inside the most powerful software company in the world, he saw that a twelve-person studio in Texas had shipped the most widely installed software on Earth without a retail channel, while Microsoft was "hiring 500-people sales teams." The lesson he took was not "make shooters." It was that the channel had moved, and incumbents measure the old channel because that's where their sales numbers live.
The young Newell's job history gives this a nice irony: he had been a Western Union telegram messenger, the human last mile of a network about to become obsolete. He knew what a dying distribution system looked like up close. Most executives look at revenue reports, which by definition record how things sold yesterday. Newell looked at install bases, at the strange fact that Doom was on more machines than Windows 95, and acted on what that meant before the revenue followed.
Tactic: Every quarter, track one metric of how your category's products actually spread (installs, shares, mods, forks) that doesn't appear in your sales dashboard, and treat any divergence between that metric and revenue as a leading indicator.
Principle 2
Make your exit credible before you negotiate
In 2001 Valve was a hit studio locked into a bad contract: Sierra owned the Half-Life IP and the right to publish Valve's future games. Valve's leverage did not come from Half-Life's success, which only made Sierra more determined to keep it. It came from the threat to leave the business altogether. Valve said it would stop making games and build other software, and it backed the threat with a real offer from Amazon to build a digital storefront. Sierra gave back the IP and the online distribution rights.
The key word is credible. Valve had founders who had already walked away from Microsoft and paid for their first game themselves. A company with no outside investors and a founder willing to fund the next game out of pocket can threaten to stop in a way a venture-backed studio can't. Harrington's exit a year earlier showed the converse: leverage belongs to whoever is willing to stay at the table and keep risking.
Tactic: Before any renegotiation with a partner who holds rights you need, line up one concrete alternative that would make walking away rational, even if you never intend to use it, and let the counterparty see that it exists.
V
The Leverage Timeline
How Valve converted a bad publishing deal into ownership of PC gaming's main channel.
1995
Doom is estimated to be installed on more PCs than Windows 95. Newell calls it "a lightning bolt."
1996
Valve founded August 24 in Kirkland, five miles from Microsoft, on Newell's wedding day.
1997
Sierra publishing deal hands over Half-Life IP and future publishing rights.
2000
PowerPlay announced with Cisco, then abandoned. Harrington sells his stake to Newell on January 15.
2001
Valve threatens to quit games, using Amazon's storefront offer as leverage; regains IP and online distribution rights.
2002
Steam unveiled at GDC in March; Valve sues Vivendi in August.
2003
Steam launches in September; Valve moves to Bellevue with 60 employees.
2004
Judge Zilly rules for Valve on November 29; Half-Life 2 ships, requiring Steam.
2011
Steam controls an estimated 50–70% of downloaded PC games.
2023
Valve revenue estimated at $5 billion; Steam roughly 60% of it.
Principle 3
Build the pipe while everyone is watching the product
During Half-Life 2's five-year development, Newell spent months on Steam. Everyone, including Valve's own players, was focused on the sequel. The founder was focused on the delivery system. Then he tied them together. Half-Life 2 required Steam to install, which turned the most anticipated game of the decade into an onboarding funnel for a platform nobody had asked for.
Players complained. The install base outlasted the complaints. Steam started as a way to deliver patches to Valve's own games, became Valve's storefront, and then became everyone's. The strategic point is the timing: the pipe got built when the company had a product strong enough to force adoption. A platform launched without that kind of product has to beg for users. A platform launched with one can require them.
Tactic: Identify the infrastructure your flagship product depends on, build your own version during the flagship's development cycle, and ship them together so the product's demand pays for the platform's adoption.
Principle 4
Own the clock by owning the cap table
Newell gave Half-Life 2 no deadline and a "virtually unlimited" budget, and promised to fund it himself if needed. That promise only means something if the founder controls the company. Valve has never taken outside capital. Newell bought out his co-founder in 2000, and Forbes estimates his stake at 50.1 percent today, with employees holding the rest. There is no board demanding a quarterly release.
The cost is real: Valve Time, Episode Three, a lost decade. But the same structure allowed Valve to restart Portal 2 when playtesters missed the portals, to spend a decade on Linux before the Steam Deck vindicated it, and to cancel games late rather than ship weak ones. Patience isn't a personality trait. It's a feature of who owns the equity.
Tactic: Before raising capital, write down the single longest-horizon bet your company might need to make, and refuse any terms that would give an outside party the power to set its deadline.
Principle 5
Compete with free on service, not enforcement
In 2011 Newell argued that piracy is a service problem, not a pricing problem, and that the answer was a better option, not stronger locks. His evidence was Steam's success in Russia, a market where piracy was widespread. If a pirated copy is free but a legitimate one is easier to get, updated automatically, and connected to your friends, the legitimate one competes.
The same reasoning shows up in smaller Valve decisions. Flash sales, once a beloved Steam ritual, were dropped because, Valve said, "really short discounts made it difficult for many people to participate." Support staff grew about fivefold after earlier criticism. The approach treats every customer complaint as a defect in the product rather than a moral failing of the customer.
Tactic: For any behavior you're tempted to police (piracy, churn, workaround use), first list three ways the official path is worse than the unofficial one, and fix those before spending anything on enforcement.
Principle 6
Treat the playtester as the experiment, never the idiot
Valve's stated creed is that "all game designers are, in a sense, experimental psychologists," and the important corollary came from Chet Faliszek's own embarrassment: when testers fail the same way three or four times, "It's not their fault, it's our fault." On Left 4 Dead, one texture change made every tester miss a ladder. On Portal 2, testers who liked F-Stop still wanted portals, and Newell redirected a whole project on that signal.
Newell put it most clearly in a 2017 answer to a student: get into "an iteration cycle where you can measure the impact of your work, have a hypothesis about how making changes will affect those variables, and ship changes regularly." He added that the content barely matters. The loop is what makes you better.
Tactic: Write down a falsifiable hypothesis before every user test (what you expect users to do and why), and when three users contradict it, change the design before you change the users.
Principle 7
Spend where the median customer will look
Newell's critique of Warren Spector's design philosophy is one of the clearest statements of resource allocation in creative work: "If only one per cent of your customers see this cool thing that takes five per cent of your development budget, that's not a good use of resources." It is an engineer's view of art, and it runs against the auteur instinct to reward the most dedicated player.
The principle doesn't mean flattening ambition. Valve's own practice is to take the exciting things that happen to a few testers by accident and make them happen for everyone. The ambition goes into the main path, not the hidden branches. It's also why Portal 2's writers cut joke endings that "required too much development effort for little payback."
Tactic: For every feature in your roadmap, estimate the share of users who will encounter it and the share of budget it consumes, and cut or redesign anything where the second number exceeds the first.
Principle 8
Leave the room when your voice becomes a verdict
After Portal 2, Newell stepped back from game development because, as project lead Josh Weier described it, developers kept agreeing with his ideas because of his position. He wanted to contribute ideas as a peer and found that impossible. Few founders notice this happening, and fewer act on it: once a founder's suggestion is treated as an order, the organization loses the founder's best value, which is a strong idea that can still be argued with.
His Half-Life regrets make the same point from another angle. He said he can't enjoy those games because he made "a much higher percentage of the decisions," so all he sees is trade-offs. Being a fan, he said, is easier with "things that in which I was less directive." The founder who steps back gains distance and the ability to judge. The organization gets back a debate it had stopped having.
Tactic: Track how often your proposals are adopted without substantive pushback; when the rate approaches 100 percent, stop proposing in early-stage meetings and contribute only as a reviewer after options are on the table.
Principle 9
Give a flat organization a forcing function
Valve's flat structure came from a specific change: once Steam made Valve its own publisher, external deadlines disappeared, and with them the reason for hierarchy. Open allocation let people go where they thought they could "create the most value." It also produced a decade of cancelled projects with "no shared vision," in Robin Walker's words, and a company that eventually had to admit it was "wrong on the premise that you will be happiest if you work on something you personally want to work on the most."
The fix was not bosses. It was Half-Life: Alyx: a studio-wide, short-term goal big enough to give a self-directed company a shared direction. Flatness without a forcing function doesn't produce freedom. It produces drift. Newell was honest about the trade from the start: "there are plenty of great developers for whom this is a terrible place to work."
Tactic: If you run a self-organizing team, set one shared, dated, company-wide objective per year that every project must connect to, and treat projects that can't recruit people toward it as candidates for cancellation.
≠
Conventional Wisdom vs. the Valve Approach
Where Newell's choices departed from standard practice, and what it cost.
| Conventional wisdom | Newell's approach | Cost or payoff |
|---|---|---|
| Raise outside capital to scale. | Self-fund; buy out the co-founder. | Total control over timelines; no external check on drift. |
| Announce early to build hype. | Communicate through shipped product; cancel late. | Few public misfires; years of perceived silence. |
| Hierarchy and managers. | Flat structure, open allocation. | Portal, Dota 2; also a decade of cancelled projects. |
| Fight piracy with DRM. | Out-serve the pirates. | Steam's success in high-piracy markets like Russia. |
| Ship on the publisher's calendar. | No deadline; "Valve Time." | Half-Life 2's quality; Episode Three's absence. |
| Founder steers creative direction. | Founder withdraws when deference sets in. | Healthier debate; slower decisions. |
Principle 10
Speak through shipped product, and cancel late
Valve's refusal to discuss games in development frustrates its fans and protects its customers. Newell's explanation in 2017 was direct: Valve's decisions are "way more conditional than most other companies," and "the one thing we won't do is waste our customers time and money, which means we will cancel or change stuff much later in development." Announcing a game commits you to it. Valve wants to keep the option to cancel open as long as possible.
When Valve does talk, it does so through artifacts: patches, alternate-reality games, Morse code hidden in radios, a BBS on Adam Foster's home phone line that cost under $100. The company accepts the "latency" this creates (its own word), including the impression that nobody is listening. The trade is fewer broken promises.
Tactic: Don't publicly announce anything you would be unwilling to cancel; communicate progress through small shipped updates instead, and accept the reputational cost of silence in exchange for keeping the option to stop.
Principle 11
Attack publicly, reverse publicly
Newell called the PS3 "a waste of everybody's time" in 2007, and three years later stood on Sony's own E3 stage, acknowledged the remark, and announced Portal 2 for the console with Steamworks integration. He called Xbox Live a "train wreck" and Windows 8 a "catastrophe." Each attack was a negotiating position about openness: who can patch, who can update, who controls the store.
The reversal is the instructive part. He didn't quietly drop the criticism. He went back on stage and said what had changed, namely Sony's openness to Steam-like features. Public criticism kept him credible. Public reversal turned that credibility into a partnership.
Tactic: When you criticize a partner or platform in public, tie the criticism to a specific, changeable policy, so that if the policy changes you can reverse publicly and turn the critique into a deal.
Principle 12
Buy insurance against your own platform landlord
Steam ran on Windows, and Windows was Microsoft's. When Windows 8 suggested Microsoft might close the PC the way Apple had closed its ecosystem, Newell, who had produced the first three Windows releases and knew exactly how much leverage an operating system holds, started building an alternative. SteamOS in 2013. Steam Machines in 2015, which flopped. Proton. Then the Steam Deck in 2022, a Linux handheld that runs most Windows games unmodified.
The insurance took nearly a decade and included visible failures. But by the time it paid off, Valve no longer depended on any other company's operating system to reach its customers. Newell's career began with porting Doom to make Windows a gaming platform. Its second act was making sure gaming didn't need Windows.
Tactic: List every platform your business runs on that another company controls, and fund at least one long-horizon project per year that reduces your dependence on the most dangerous of them, even if early versions fail.
H
The Hardware Ledger
A decade of bets on owning the device layer.
| Device | Year | Outcome | Status |
|---|---|---|---|
| Steam Machines (partner PCs) | 2015 | Fewer than 500,000 sold by June 2016. | Flop |
| Steam Controller (1st gen) | 2015 | Discontinued after a $5 clearance in November 2019. | Flop |
| HTC Vive / SteamVR | 2016 | Established room-scale VR with "Lighthouse" tracking. | Partnership |
| Valve Index | 2019 | Higher-end headset; platform for Half-Life: Alyx. | Premium |
| Steam Deck | 2022 | Linux handheld running most Windows games via Proton. | Hit |
| Steam Machine (Valve-built) | 2026 | $1,049; randomized reservations amid memory shortages. | Pending |
Free playbook
Get The Business Model Playbook
58 business models, one visual page each: how the money flows, the metrics that matter, and who runs it. Free when you join the Faster Than Normal email.
Free. No spam. Unsubscribe anytime.
Part IIIQuotes and Maxims
In their words
— Gabe Newell, Reddit AMA, January 17, 2017The most important thing you can do is to get into an iteration cycle where you can measure the impact of your work, have a hypothesis about how making changes will affect those variables, and ship changes regularly. It doesn't even matter that much what the content is - it's the iteration of hypothesis, changes, and measurement that will make you better at a faster rate than anything else we have seen.
— Gabe Newell, Reddit AMA, January 17, 2017The issue with Half-Life for me is that I was involved in a much higher percentage of the decisions about the games, so it's hard for me to look at them as anything other than a series of things I regret.
— Gabe Newell, Reddit AMA, January 17, 2017I don't think most of us think retrospectively like that. I think we are mainly thinking about what we need to do in the future. It still seems like we are just at the beginning of what is going to be possible, and the lure of that future is way more powerful than self-congratulation.
— Gabe Newell, to The Washington Post, 2014[Fans] hug me when they run into me. I'm not a hugging person, but that's what they want. I was with my kids the first time that happened in public, and my kids were pretty cool with it. But I wasn't. 'Dad, roll with it.' Even now, I'm learning from our customers.
— Gabe Newell, Reddit AMA, January 17, 2017The one thing we won't do is waste our customers' time and money, which means we will cancel or change stuff much later in development.
Maxims
- Count installs, not invoices. The channel moves before the revenue does; watch how software spreads, not how it sold last quarter.
- Leverage is a credible exit. The partner who can walk away sets the terms; the partner who can't, signs them.
- Ship the pipe with the product. A must-have release is the cheapest way to make a platform mandatory.
- Whoever owns the equity owns the clock. Patience is a feature of the cap table, not of temperament.
- Out-serve the free option. Piracy, churn, and workarounds are product defects before they are moral failures.
- The tester is never the idiot. When three people fail the same way, the design failed.
- Build for the many, not the hidden few. Five percent of the budget for one percent of users is a bad trade.
- Exit the room when agreement becomes automatic. A founder's idea is only useful while people can still argue with it.
- Freedom needs a target. Flat organizations drift without a shared, dated goal to pull them together.
- Never depend on a landlord you can't replace. Build the alternative platform before you need it, and expect the first versions to fail.
