(FORTUNE Small Business) – The headquarters of America's most infamous videogame developer lie ten miles east of downtown Dallas and across the street from a Hooters. It is noon, but all the shades are drawn, making the id software offices unusually dark and quiet. The glow of the game animators' computer monitors is the only source of light in most offices, the light tapping of keys and mouse clicks the only ambient sound.

Employees say they are immersed in id's legendary "crunch mode," a period before the release of any game during which they work 80-hour weeks and sometimes sleep on the office cot. Yet the place is so quiet you'd think the entire 20-person staff was out sick.

That air of eerie calm belies the intense work taking place in id's office: "It's like being blindfolded and running flat-out through a forest, praying you don't hit a tree," says CEO Todd Hollenshead. But that's not all Hollenshead has to worry about. For id to succeed, its next game, Doom 3--four years in the making--must be a monster hit. And even if this game sells well, as most experts expect, id, one of the last small players in a field of giants, won't be able to take its survival for granted. As in id's videogames, potentially deadly dangers--a consolidated industry, tougher competitors, and a distracted founder--lurk around every corner.

The videogame business has changed significantly since id launched the original version of Doom in 1994. Today id remains a maverick in an industry where most other independent game developers have been bought out by powerhouse publishers such as Microsoft and Vivendi Universal. (According to Hollenshead, id is profitable, with $20 million in sales last year.) Of the top 50 console games of 2003, only three--SOCOM, SOCOM II, and Star Wars: Knights of the Old Republic--were created by independent developers. Today's games, with their complex graphics and realistic effects, have become extremely expensive to produce--in some cases costing as much as $30 million. That's a hefty sum for a small firm such as id.

Moreover, most videogame companies launch multiple titles each year, hoping one will be a hit--much as a Hollywood studio releases dozens of films in search of that one blockbuster. (See the accompanying story on independent publisher Majesco, which is releasing a portfolio of inexpensive games to hedge its risks.) But id, which launched its last game, Return to Castle Wolfenstein, in 2001, is betting the company on its latest version of Doom. Period. So far, that kind of faith has paid off: Over the past ten years co-founder John Carmack's technology and the warped minds of the id creative team have run up a perfect track record. Every one of their games, from Wolfenstein 3D to Quake III, has been successful. But can a reliance on creating a nonstop stream of megahits be considered a prudent business plan?

Most entrepreneurs might express some anxiety if they were about to release a product that could single-handedly determine the fate of their company. They might appear the slightest bit concerned if technical glitches and a perpetually shorthanded staff resulted in two years of delays. They might worry if hype surrounding the launch (some analysts expect Doom 3's release to spark an increase in sales of high-end PCs specifically designed to play computer games) approached unrealizable proportions. But then, Carmack, 33, is not most entrepreneurs. When fans ask him when his new game will be released, he gives them a stock response: "When it's done." (The company says that the PC version of the game will launch this summer.) Even in an industry where delayed releases are de rigueur, that attitude seems downright haughty.

But Carmack, id's remaining enfant terrible (the other, fellow co-founder John Romero, was fired in 1996), has reason to act cocky. The first edition of Doom launched the genre of "first-person shooter" videogames, eventually took in $200 million, and changed the industry forever. (Some would argue not for the better: The game's aggressive violence horrified many parents and politicians.) And Doom's innovations were not only technological but financial as well. Id was the first company to distribute its games via shareware, releasing the first few levels for free over the Internet and then charging customers for the rest of the game. The move proved wildly successful, getting the game into the hands of players who would not otherwise have paid the full $40. What's more, it meant id didn't have to pay a publisher the standard 5% to 10% of the price of each game for distribution.

Id's next release, 1996's Quake, was another commercial success, selling $214 million. And, almost as important, it introduced a new revenue stream: licensing. Carmack had always released the source code of his games to allow users to design their own backgrounds or villains and share their modifications with friends. With Quake, Carmack decided to let other videogame developers purchase his "engine"--the core technology that allows the game to process a player's moves fluidly. Since 1996, games powered by id's engines have generated $1 billion in sales--and tens of millions in licensing fees for id. Today licensing fees account for 20% of id's revenue.

But the advantage that id once received from its business innovations has long passed. While id hasn't created a new game engine since 1999's Quake III, rival developer Epic Games of Raleigh updates its Unreal engine annually, making it an increasingly popular choice for developers of first-person shooters. (This year Eidos Interactive's Deus Ex: Invisible War and Ubisoft's Tom Clancy's Splinter Cell were both developed using Epic's engine, not id's.) Valve, a Bellevue, Wash., developer that purchased id's Quake engine to create the game Half-Life in 1998, now creates its own engines from scratch; this year's Half-Life 2 is as eagerly anticipated as Doom 3. And the id-powered Medal of Honor has blossomed into a multimillion-selling franchise for publishing leader Electronic Arts.

In an even more troubling sign for id's future, Carmack has begun to give indications that he's getting bored. Married since 2000, he's expecting his first child in August. And he has spent the past three years devoting 40 hours a week to Armadillo Aerospace, his four-year-old company that is working to develop customized rocket ships. "If I wasn't doing what I'm doing in videogames, someone else would within a year or two," says Carmack. "In aerospace things are still being done the same as they were 50 years ago. I feel that I can really make an impact in that field."

Carmack insists that he has no intention of leaving id, and Hollenshead says that focusing on Carmack's possible departure is unnecessarily alarmist. "Id and our games are about more than just John Carmack," he says.

Most of these problems would be instantly solved if id would make the same decision its rivals have, and sell. Id has received scads of buyout offers over the past decade, and--according to analyst Mike Wallace of UBS--"anyone would love to own id." Colin Sebastian, an analyst with San Francisco-based Soundview Technology Group, says that a suitor could pay $500 million for id.

But Carmack says that--despite the uncertainty and risk--independence is its own reward. "The bottom line is that we like being in the driver's seat," he says. "We can take our games to whatever publishers we want, and those publishers cough up a lot of money for them."

As long as those games continue to sell, that is.