(FORTUNE Magazine) – If you're willing to set aside for a moment the gigantic chicken-and-egg problem of where to get the hydrogen to run a new generation of efficient, nonpolluting hydrogen-powered cars, General Motors has a hell of a story to tell. The automaker has recently begun revealing the depth and breadth of its cloistered fuel-cell R&D program, which is far more sweeping than most outsiders imagined.

GM is betting heavily that hydrogen-fueled vehicles will let the company thrive against a backdrop of growing worries about global warming and the new regulations on carbon-dioxide emissions those concerns will probably spawn. "Fuel cells are very important to GM's future," says CEO Rick Wagoner. "We've spent hundreds of millions already, and we're going to spend a lot more than that until we get into production vehicles."

GM is not alone in this gambit. Rival companies, DaimlerChrysler in particular, have loudly proclaimed their high hopes for fuel cells too. Yet far more than any other automaker, GM has articulated a radical vision of what freshly engineered cars built around fuel-cell propulsion systems could look like. The vehicles GM wants to bring to market by 2010 will have wild--and interchangeable--bodies with spacious interiors sculpted by designers freed from having to accommodate bulky piston engines, transmissions, and drive shafts. The cars will be "drive by wire," with braking, throttle, and steering controlled not by mechanical links but by electrons pumped through wires and massaged by an unprecedented array of solid-state circuitry. And they will come brimming with a critical component that is both weightless and shapeless: software.

Before fuel-cell cars roll off production lines in any number, though, a hydrogen infrastructure must come into being, which means that automakers and other parties need to pique the interest of the government, energy companies, and the public. That explains why GM has opened a window onto its fuel-cell program years before it produces any cars. "We find ourselves in the position of advocacy for this hydrogen-economy thing," says Byron McCormick, executive director of fuel-cell technology and commercialization. "We're talking about huge societal changes here, and we are working this globally with the agencies that regulate energy and environment and safety."

To be sure, General Motors has in the past made big claims and investments in programs that were supposed to revolutionize auto design and didn't, like the Saturn division and the EV1 battery/electric car. This time, though, GM may have not only a car that is radical but also one that people will actually want to buy. The company is now out on the auto-show circuit promoting its conceptual leap: the "skateboard" chassis that forms the spine of its fuel-cell vehicle, dubbed Autonomy. The skateboard is essentially a flat, six-inch-thick platform sprouting four wheels attached by aerodynamically shrouded suspension arms. Into this thin chassis GM hopes to cram all the other parts that make the car go, including the fuel-cell "stack," a raft of power electronics, and some sort of hydrogen-fuel storage. Wires will deliver power directly to lightweight electric motors integrated into the wheels. Customers will be able to choose any of several body styles, and GM will simply fasten the top to the chassis.

Software will give each car the feel and handling appropriate to its style. If it's a sporty body, the code will prescribe responsive steering, firm suspension, and rapid acceleration. A family car will come with software calling for slower steering, a softer ride, and less urgency to the pick-up. Drive-by-wire does away with the need for conventional steering wheels and pedals, making it possible to equip the cars with videogame-style manual controls if that's what drivers want. GM thinks three or four sizes of skateboard chassis may one day form the underpinnings of its entire product line, with bodies for economy cars, SUVs, or minivans produced according to demand. The manufacturing economies of scale for the platforms could be awesome, and help justify much smaller production runs of charismatic bodies designed to appeal to niche markets.

Let us pause now to catch our breath and point out that the Autonomy show car is only a model, not a sneak preview of anything that will be in showrooms soon. And the dimensions of its compact skateboard chassis are not yet carved in stone, but rather a target intended to inspire hundreds of company engineers and scores of supplier firms to stay focused on the ambitious performance goals of the sprawling Autonomy research and development program. Later this year a running skateboard chassis about 12 inches thick is scheduled to be ready for testing and joyrides, and in the years ahead the GM team will find out just how much it can slim down the package.

Much of GM's research is focused on cutting the cost and boosting the power output of the fuel-cell stack. Fuel cells produce electricity by catalyzing a chemical reaction that combines hydrogen with the oxygen in air. The environmental beauty is that the process has only one byproduct: water vapor. Although it makes a wonderful fuel, hydrogen is also the least dense of all elements, so fitting enough aboard a vehicle to give it decent driving range is a big challenge. "There's no dense, cheap way of storing hydrogen at the moment," laments Autonomy program director Christopher Borroni-Bird. Company engineers think that instead of holding hydrogen in gaseous form, fuel-cell cars will carry the element in solid compounds called hydrides. One candidate is a spongy metallic hydride with a huge surface area that bonds with hydrogen molecules as it cools. When heat is applied onboard, the metal-hydride compound gradually releases gaseous hydrogen to the fuel cell. But in a crash, it wouldn't leak hydrogen, as compressed-gas tanks could.

Still, hydrides aren't very efficient. Those currently available can absorb at most 5% of their weight in hydrogen; finding a hydride with double this storage capacity is at the top of GM's wish list. In the meantime, designers will try to fill empty cavities like the inside of the car's frame beams with hydride material. "We're maybe halfway there," says Larry Burns, GM's vice president of R&D.

Wherever it may be squirreled away, the hydride will need replenishing. Indeed, fuel-cell cars will never get on the road without a reliable supply of hydrogen. Engineers believe the first step may be equipping gas stations with special devices called reformers that can liberate hydrogen on demand from gasoline or natural gas while continuing to dispense gasoline to conventional vehicles. GM believes its first-generation fuel-cell vehicles may also need onboard gasoline reformers to avoid getting stranded with an empty hydrogen tank. Later on, as the hydrogen infrastructure begins to expand, cars could be built without these devices.

Another promising technology in GM's lab is the so-called reversible fuel cell, which can create hydrogen from electricity and water as well as the other way around. At night a driver would plug the car into an electrical outlet; the cell would then use cheap overnight current to crack water into hydrogen and oxygen. On the road, the fuel cell would return to standard mode and convert the hydrogen back into electricity and water vapor. Industrial-strength electrolyzers that could create a tankful of hydrogen gas from water in a few minutes might one day go into fuel stations. Given recent improvements in all these technologies, GM is optimistic that by the end of the decade, none of them will stop the show.

Once fuel-cell cars go into mass production, they could well end up being cheaper to make than today's cars. A factory producing skateboard chassis could be a lot smaller and less complex than existing car-assembly plants. In fact, chassis and body manufacturing could take place in different parts of the world, with the two big components mated only upon order by the customer, much as Dell makes computers. The electronics-rich nature of the chassis means that many of its systems can be made like silicon chips, dispensing with the need for a lot of conventional metalworking machine tools. Engineers think the four suspension and wheel-motor assemblies at the corners of the Autonomy vehicle could be quite similar to each other, if not identical, again suggesting tantalizing economies of scale through parts commonality. Although they won't elaborate, GM engineers allude to new body-manufacturing processes they say would let buyers order a customized shape that could be profitably built in a batch of one. Make mine a wienermobile!

The Autonomy program appears to have blown a whiff of freshness into fuddy-duddy old General Motors. A number of its principal participants are patently eclectic folk, some hailing from outside the car business. One manager used to work at federal weapons laboratories and Disney Imagineering, which develops technology for the Disney theme parks. Borroni-Bird, who was lured away from DaimlerChrysler a few years ago, is a Liverpudlian and author of learned papers with titles like "What Can Seamless Electro-Mechanical Vehicles Learn From Nature?" And the arrival of Bob Lutz as GM's vice chairman in charge of product development--with his exhortations that cars are fashion and entertainment and that you win the game with great designs--has visibly fanned the enthusiasm of people working on a vehicle architecture that could unleash a new wave of sculptural excitement on the roads.

GM has learned from its failures too. Even though the EV1 was ultimately a political vehicle useful mostly for soothing government agencies and environmental groups, it was the best of its kind, and building it gave the company valuable grounding in the complexities of power electronics directly applicable to the Autonomy program. And the EV1 experience drove home the lesson that a car with only two seats, a driving range of 70 miles, and a battery-recharge time of several hours is unsalable. The designers know the Autonomy will need a minimum range of 250 miles, with refueling time comparable to that of a piston-engine car.

The Autonomy program is the biggest line item in GM's research budget, and R&D boss Burns, who reports directly to CEO Wagoner, knows he must keep reaching milestones along a clear route to profitability to keep his baby funded. Most carmakers are working on fuel-cell systems too, and both Toyota and Honda will field small numbers of test vehicles next year (see sidebar). But beating everyone to market is not GM's goal; being market leader is. "That's not saying I want to be the first to sell 50,000 fuel-cell cars and commit to technology that would be obsolete by the time we get to production," says Burns. "We want to be the first to sell a million of what will be such superior vehicles that people will be willing to pay more for them." For a clean, green car program, that sounds almost businesslike.

FEEDBACK: sbrown@fortunemail.com