(FORTUNE Magazine) – One of the most exciting technologies of the new millennium is about to move a few steps closer to the mass market. The technology is fuel cells, almost universally seen as an energy-conserving, low-pollution way to power millions of tomorrow's motor vehicles. What stands in the way is economics, which can be overcome only by chipping away at manufacturing costs and building plants big enough to realize economies of scale.

Though DaimlerChrysler and others still hope to sell the first fuel-cell cars in California in 2004, mass production by the auto industry isn't likely to begin until around 2008. Most of the big carmakers have, between them, committed more than $1 billion to this dream. Meanwhile, the emergence of fuel-cell products for other uses, though produced in far smaller volumes than those foreseen for motor vehicles, is significantly advancing the manufacturing know-how the automakers will need:

--Production engineers at Ballard Power Systems in Vancouver, which is working closely with DaimlerChrysler and Ford, are revving up to produce fuel cells that will be the heart of a small portable electricity-generating device. Sunbeam Corp.'s Coleman Powermate unit plans to have it in stores for this year's Christmas selling season.

--Plug Power in Latham, N.Y., which has a distribution agreement with General Electric, has just begun shipping the first fuel-cell generating units designed to make a home or small business largely independent of the electric-power grid.

--International Fuel Cells in South Windsor, Conn., which already makes big, 200-kilowatt fuel-cell generators for commercial uses, will roll out a new version in 2003 whose costs will be two-thirds lower because of design and manufacturing improvements.

Fuel cells priced like jewelry have been used on spacecrafts for decades to generate onboard power and provide drinking water for crews. The far cheaper types needed for mass production work on the same principle. They transform hydrogen and oxygen in air into electricity in a noncombustion catalytic process. It's the reverse of the water-splitting electrolysis experiment you may remember from high school chemistry class. Fuel cells are about 50% more efficient than car engines at producing power, which is delivered to the wheels by electric motors.

Though a fuel cell emits no pollution or atmosphere-warming carbon dioxide, it does cause emissions elsewhere--albeit less--in plants or devices that extract the needed hydrogen. One way to obtain hydrogen is to "reform" natural gas into its constituent hydrogen and carbon molecules in a process that releases some carbon dioxide. A few years ago most automakers believed that vehicles would have to extract hydrogen in an onboard reformer using a liquid fuel such as methanol. There may be a simpler solution, at least initially: building fueling stations similar to prototypes operating in California, where hydrogen produced on demand from natural gas is pumped under pressure into tanks aboard experimental electric cars powered by fuel cells.

About three years ago, buses and cars looked like the first big fuel-cell applications. As things have turned out, stationary and portable generators are the earliest arrivals on the scene. Engineers at Ballard Power Systems, one of the main players in fuel cells, are tooling up now to make them for Coleman Powermate. A leader in gasoline-powered portable generators, Coleman is tightlipped about its new product, which is said to be usable indoors, where the carbon monoxide in exhaust fumes is verboten.

Making it, however, moves Ballard closer to the day of mass-produced fuel-cell cars. "Developing proprietary manufacturing technology is an important part of our business plan," says Firoz Rasul, Ballard's CEO, "and making these units for Coleman helps us refine processes that will apply to stationary-power and automotive products as well."

In 1998, Ballard entered a partnership with Ford and DaimlerChrysler, which want to assure themselves of a future supply of fuel-cell "engines." That venture is now called Xcellsis GmbH. The three companies also have a joint venture called Ecostar Electric Drive Systems, which is developing and commercializing the non-fuel cell components, such as motors and power-handling electronics. In the meantime, Ballard has been collaborating with engineers from Mercedes-Benz on developing production methods that apply just as well to what Coleman needs.

The good news from a manufacturing standpoint is that fuel cells are full of components that are flat and thus amenable to automated handling. The fuel-cell type everyone's banking on for the near future uses a "proton exchange membrane," or PEM, a thin film of plastic material coated with a platinum-based catalyst. The membrane sifts electrons--a.k.a. electricity--from a flow of hydrogen molecules so that they can be detoured to run a motor or some other useful gadget before being routed back to the membrane's other side, where they combine with oxygen to produce heat and water vapor.

Working with its catalyst supplier, Johnson Matthey in the U.K., Ballard is trying to minimize the amount of costly platinum required. Research has shown that only the top layer of catalyst reacts with the gases in the fuel cell, so applying too much is just a waste, says Eamonn Percy, Ballard's vice president of operations. Just a few years ago Ballard technicians were hand-painting catalyst onto membranes for prototype fuel cells, a method that took too long and used too much platinum. The latest process, using specialized machinery that visitors don't get to see close up, is the fifth generation of catalyst-application methods the engineers have devised.

Ballard, which trades on the Nasdaq, has spent about $15 million on a recently installed suite of equipment intended to meet the medium-volume needs of the Coleman program and the fuel-cell development programs under way at several of Ballard's automaker customers. The company recently won a $17 million contract, its biggest to date, to supply systems to Honda's fuel-cell-vehicle development program. About 250 engineers and production workers are employed at Ballard's brand-new Plant 1, a 110,000-square-foot facility outside Vancouver.

In the manufacturing process, catalyst-coated membranes are sandwiched and bonded between pairs of electrodes made from a conductive sheet material to form units called membrane-electrode assemblies. In turn these are sandwiched between "flow-field plates," molded flat pieces with mazelike channels in both sides. The channels carry gases to and from the membrane where the electricity-generating reaction takes place, direct the flow of cooling air or water, and remove the water droplets that form as the hydrogen and oxygen combine. Each square-shaped membrane-electrode assembly, measuring about eight inches on a side, can produce about 0.7 volts of electricity. A fuel cell system calling for 70 volts requires a stack of 100 units.

Ballard buys molded flow-field plates for the Coleman product from a supplier but will make its own plates for the automotive fuel cells. In earlier prototypes the plates were carved from thin slabs of graphite material on computer-controlled milling machines, in a process far too costly and time-consuming for volume production. Now the plant is using equipment with patterned rollers--some of it originally designed to produce laminated linoleum flooring--to emboss the channels.

In collaboration with DaimlerChrysler over the past two years, Ballard has designed a plant intended to produce automotive fuel-cell systems at a rate of 300,000 units per year. Depending on how demand in the automotive market shapes up, the high-volume plant, which would cost $250 million to $300 million to build, could be needed in about 2006.

Ballard's success to date in lowering costs, Rasul says, puts it on track to meet the auto industry's price goal of $20 per kilowatt for the fuel-cell stack at the 300,000-unit production rate, or $50 per kilowatt for the car's entire electric propulsion system. Costs for the stack were around $500 a kilowatt three years ago and since have been whittled down, though Ballard won't give a figure. A car needs about 50 to 75 kilowatts of electric power.

Plug Power, another Nasdaq company, isn't chasing the automotive market at all but focusing on stationary fuel cells that run on natural gas. CEO Roger Saillant says the company's plan is to develop a family of products ranging from one kilowatt to 100 kilowatts, built from modular components that will reduce design and manufacturing costs. Plug Power plans to ship about 150 of its five-kilowatt units this year to electric utilities and government labs that want to get experience with installing, operating, and servicing the systems, which incorporate reformers that produce hydrogen to feed a PEM fuel-cell stack.

The company calls its first system a "grid parallel" design because its mission is to meet the first five kilowatts of a building's power needs--enough to supply a modest-sized home or store--with utility power filling in the rest. Feedback from users of this first-generation system will help shape the design of subsequent models aimed at consumers who are in remote locations or have worries about the quality and reliability of utility power.

General Electric has signed up to distribute Plug Power's products in the U.S., except in four Midwestern states covered by DTE Energy, a stakeholder in Plug Power and the parent of Detroit Edison. Saillant won't say what his company's production costs are for the five-kilowatt systems but notes that to be competitive, the installed cost of residential systems will need to be between $300 and $500 per kilowatt, and between $1,200 and $3,000 for larger commercial systems, depending on the applications. Stationary systems cost far more per kilowatt than projected automotive systems because they extract their own hydrogen and are designed for high reliability and round-the-clock usage.

Most of the stationary fuel cells in service today were built in the past decade by International Fuel Cells in South Windsor, Conn. A unit of United Technologies, IFC is the grand old man of fuel cells, having supplied them to NASA's Apollo program and the current space shuttle fleet.

IFC has sold about 220 of its commercial PC25 units, which make 200 kilowatts of power apiece and serve in locations like a mail-sorting center in Alaska, where frequent power outages have caused machinery to jam. The PC25s include fuel processors that extract the needed hydrogen from natural gas and use a phosphoric-acid technology instead of PEM inside the stack. That design is soon to be retired in favor of PEM, which allows greater electrical output per given volume of stack and does away with the corrosive chemical.

The company is now in the process of tooling up for a PEM fuel-cell market that it sees growing from $1 billion in 2005 to $20 billion in 2010. With its eye on vehicles as well as stationary applications, IFC has five automotive customers for its prototype PEM systems, including BMW and Hyundai. In California, the South Korean company recently demonstrated a car using IFC's ambient-pressure fuel cell system, which its maker claims is simpler and more efficient than competing designs like Ballard's that use a compressor to move gases through the stack.

Like Ballard, though, IFC expects stationary and portable power generation to become volume markets for fuel cells sooner than cars. IFC plans to roll out the PC25 stationary power plant's PEM successor, which will put out about 150 kilowatts, in 2003. Jim Bolch, the company's vice president of manufacturing and supply management, says the new model will be dramatically cheaper, with an installed cost of $1,500 per kilowatt, vs. the PC25's $4,500.

At least in the early stages of production, Bolch believes, the company will make and assemble its own PEM stack components. IFC already uses some robots in building the stacks for its PC25 product. Eventually, Bolch says, the company may end up providing the design, assembly, and testing for its products and let the suppliers make most or all of the components. This year IFC will invest $60 million in developing fuel-cell products and manufacturing processes. Like Ballard and Plug Power, it is betting on a bright future.

FEEDBACK: sbrown@fortunemail.com. Stories from FORTUNE's Industrial Management & Technology section can be found at fortune.com/imt. Executives in manufacturing and research and others eligible to receive FORTUNE's Industrial Edition can subscribe by calling 888-394-5472.