(FORTUNE Magazine) – Regulars at the airport outside Mojave, Calif., have seen some odd aircraft over the years, but one plane that's been flying there for the past few months is especially peculiar: a Boeing 747 with three conventional jet engines and, inboard on the port side, a humongous fourth engine so much larger than the others that it almost drags on the runway. Early on, when the engine was spooled up before takeoff, its exhaust blew a gap in the windrow-like pile of boulders that is supposed to serve as Mojave's jet-blast deflector. On one occasion the pilots throttled back the other three just after liftoff and climbed out on that one engine.

The 747 is a flying test bed for the General Electric Aircraft Engine business. The big engine is the GE90-115B, the largest ever flown in both physical size and power. Its fan, which is what you see when you look from the front, is 10 feet 5 inches in diameter. The number 115B designates that the engine puts out 115,000 pounds of thrust. But on a stand at GE's 7,000-acre testing grounds in Peebles, Ohio, it almost topped 123,000, nearly the combined output of the 747's three other engines. Starting in early 2004, pairs of this GE giant will power new Boeing 777 models, including one able to fly 18 hours nonstop--enough for Los Angeles to Paris or Atlanta to Hong Kong if you care to sit that long.

The 115B is not a completely new design. The first GE90s went into service in 1995. But after spending some $2 billion developing them, GE is putting as much as $600 million more into this version. And that's just part of what GE's aircraft engine managers call the most ambitious product and technology development program in their history. At a time when the airlines, GE's principal customers, are nosediving toward bankruptcy, trailing plumes of burning cash, the company has a dozen new or updated engines under development. Outsiders might well wonder whether GE has jettisoned common sense. But it doesn't have a lot of choice. Engines, often sold at breakeven or at a loss, are not where this business makes its money. They are a means to an end: parts and service revenues, which will account for 40% of the business's $10.6 billion in sales this year and possibly as much as two-thirds of its $2.1 billion in operating profit.

Since the early '90s GE, with the help of a French partner, has become the world's leading manufacturer of engines for airliners. It has changed where and how it builds engines, including the recent start of its first assembly line, albeit a very, very slow one. It has dramatically compressed the time, and thus the cost, expended in developing engines. And it has brought in a crowd of new partners, including competitors, to share development costs, allowing it to take on more projects. All this has been in the interest of building a corporate annuity, a decades-long stream of very profitable parts and service sales.

David Calhoun, 45, became president and CEO of the aircraft-engine business in 2000, after running, successively, the corporate audit staff, the transport equipment and lighting businesses, and GE's Employers Reinsurance. He says the strategy in aircraft engines is the same as in GE's other "long cycle" businesses--locomotives, steam turbine generators, medical equipment. "The industries in which GE is involved are the sort that stand the test of time. You can afford to take the long look. The equipment lasts for many, many years, and you get to service it every step of the way. The model works. As long as you're secure in the fundamentals, in our case as long as we believe people and cargo will fly, then you're okay."

Even though fewer people are flying this year, GE can't wait around for recovery. Explaining what he admits is a "boatload of development," David Joyce, recently promoted to vice president of commercial engines, says, "When the up-cycle hits, we have to have the most modern product. We have to be the engine manufacturer of choice."

Even then, the payoff will be well in the future. For example, while the GE90-115B will be the only engine that airlines will be able to get on the new 777 models, Boeing will have to sell around 400 planes before GE recovers its investment. That could easily take a decade. So far, Boeing has orders for 54. But even in the current down-cycle, and even though its commercial-engine deliveries fell from 1,675 in 2001 to 1,395 this year, GE isn't doing as badly as might be expected. In November Calhoun said that his entity's 2002 revenues would be down 7% but that operating profits would be about the same as the $2.1 billion earned in 2001 and would move up a smidgen in 2003.

If you say, "Aha, military sales!" you're partly right. GE supplies engines for everything from supersonic fighters to cargo transports to attack helicopters, not to mention the gas turbines that power the U.S. Navy's non-nuclear surface ships. Sales of military engines, spares, and service were up some $520 million this year, to $2.8 billion. Except for industrial gas turbines, flat at $1.4 billion, the rest of the engine business was down. Commercial engines dropped 20%, to $2.2 billion; service and parts declined 16%, to $4.2 billion. Profits on the military business are always thin, and this year the margin got thinner as profits failed to rise as fast as military sales. One thing that did help hold up total profits was some $500 million in cost cutting, including layoffs that reduced head count from 30,200 to 25,700. However, to use a GE cliche, parts and service revenues are what "spell 'mother.'" While that part of the business declined this year, the profits it generated went up 20%.

The financial health of a commercial aircraft engine builder can be measured by its "installed base," or how many of its engines are flying. Until the mid-1990s, GE was a weak second, lagging behind United Technologies' Pratt & Whitney. In its Lynn, Mass., factory, GE built the first U.S. aircraft jet engine early in World War II. After the war it had a succession of successes with military jets, expanding its manufacturing and engineering space by taking over a formerly government-owned factory complex in Evendale, a suburb north of Cincinnati, that has since become headquarters. But all too often GE was outmaneuvered and outsold in the commercial market by Pratt & Whitney.

Not counting engines made by the Russians, just seven years ago more than half of the 28,000 engines on planes carrying 50 or more passengers bore Pratt & Whitney's flying blue eagle trademark. Well behind was GE, with about 4,000, plus slightly fewer than 5,000 from CFM International, a fifty-fifty joint venture between GE and France's Snecma Moteurs. Britain's Rolls-Royce had around 3,500. The small balance came from Honeywell and International Aero Engines, a collaboration among Pratt, Rolls, a Japanese consortium, and DaimlerChrysler's MTU Aero Engines.

Today the number of jet engines in the 50-passenger-and-up group is around 36,000. But the number from Pratt & Whitney has declined to not much more than 12,000. Many older planes with its engines have been retired. It has also lost out almost completely to GE and Rolls in the 50- to 90-passenger regional-jet business. On those planes just one engine brand is certified, as opposed to two or three for most Boeing and Airbus planes. The combined GE and CFM total is now more than 16,000 engines. Rolls, which is especially strong in engines for wide-body aircraft, has risen to about 6,000. (Those numbers don't include turboprops or executive jets, in which Rolls and Pratt compete but GE does not.) The future looks even better for GE. Still more aircraft with Pratt engines will be parked. And in the next three years GE and CFM together will deliver 50% more engines than Pratt and Rolls combined. GE claims that CFM's installed base alone will pass that of Pratt & Whitney as early as 2004.

From Calhoun down, GE executives explain time and again that their big bet on new-engine development is to make sure that they don't let somebody else do to them what CFM has done to Pratt & Whitney. Founded in the mid-1970s, the joint venture was an early effort at dividing the task of developing and manufacturing an engine. GE designed and now builds the core, or "hot section"--the compressor, combustion chamber, and high-pressure turbine that are, in fact, the jet engine. Snecma took on the gearbox through which auxiliary equipment is powered and the parts that turn a jet engine into a turbofan, a low-pressure turbine bolted on the back that drives a big fan up front. The result was the CFM56 engine.

CFM first grabbed business from Pratt & Whitney when Douglas introduced a new model of the DC-8. But the big break came in the early '80s. Boeing, which had opted to certify only a Pratt & Whitney on its 737, wanted a more modern engine for the plane's second series. Pratt didn't come up with one that was suitable. CFM did. It became and remains the only engine builder for what has developed into the most successful aircraft ever launched. More than 330 carriers, from Aer Lingus to Zhuhai Airlines, fly Boeing 737s. Since 1984, CFM has delivered 6,362 engines for the 737, plus another 3,048 for Airbus's competing narrow-body aircraft. More are on back order. And there's no sign that Boeing is giving a thought to taking its winner out of production. The sticker price per engine is $5 million. That gets discounted, but it's a nice starting point. This year GE began selling $700,000 kits to upgrade older CFM56 engines to more-economical, nearly new ones. Southwest Airlines, which flies 150 Boeing 737s with CFM engines, was first in line to buy the kits. GE projects that by 2010 service and parts revenues on its installed base of CFM56 engines will be more than $4 billion annually.

The GE-Snecma partnership could hardly have worked better. But it's an open marriage. Snecma recently teamed with Pratt & Whitney to vie against GE--unsuccessfully as it turned out--in two regional-jet-engine competitions. To be partners one day and competitors another is now a way of doing business for all aircraft engine makers. Airlines want better economy and more "time on wing." Governments demand fewer emissions and less noise. To respond, engine builders have to spend more to adopt exotic technology, such as composites and ceramics, and to take new approaches to compression and combustion and the control of exhaust gases. In the past GE has occasionally opted not to offer an engine for a new plane. It passed on Boeing's 757. Today that's less likely. Says Calhoun: "Winning on airframes is the biggest, most important thing we have to do." Still, it helps a lot, as a GE manager puts it, "to share the pain."

In addition to the big GE90s, General Electric has three families of commercial-aircraft engines, all labeled in the alphanumerics of engineer-speak: CF34s for regional jets, CFM56s for narrow-body, single-aisle planes, and CF6s for wide-body planes such as the 747 and the Airbus A330. In every category GE has new engines in development, and for every one of those it has some sort of partnership deal. On top of that, it is working with Rolls-Royce on an engine for the F-35, the Joint Strike Fighter. And it has teamed with Pratt & Whitney on an engine for Airbus's four-engine superjumbo, the 550-passenger A380. On that job both companies will outsource a lot of business. With orders from each, Germany's MTU will build about a fourth of the A380's engine.

In all the engine programs other than CFM, GE has "revenue-sharing participants" that pay a fee to be part of the project. They then design and build some piece of an engine and get a share of sales and the aftermarket orders for their parts. Three revenue-sharing partners supply more than 40% of the cost of the GE90 engine. Snecma builds most of the compressor and is a GE partner in a Texas company that builds the composite fan blades and in a French company that manufactures a structural part. Fiat Avio, a GE partner on a helicopter engine, makes the gear box, a Fiat specialty. Ishikawajima-Harima Heavy Industries builds the GE90's low-pressure turbine and is also a revenue-sharing partner on "growth" versions of the regional-jet engine. Out of about $4.4 billion in parts delivered to the aircraft-engine assembly plants last year, revenue-sharing participants provided around $900 million worth. GE buys an additional $2.5 billion in parts and material under more conventional arrangements and then adds $1 billion in value in its own plants or those of wholly owned affiliates that supply it as well as competitors.

GE upset the companies working with it on the GE90 when it lagged a year behind Pratt and Rolls in getting the engine certified on the 777 for ETOPS. That's the industry acronym for Extended-Range Twin Engine Operations, the FAA regulations under which airlines fly twins over the oceans. Pratt and Rolls had existing engines that they beefed up to meet the Triple 7's needs. GE started fresh. New designs take longer; in this case much of the delay was caused by choosing a composite material instead of metal for the fan blade. That lowers weight, always a goal in aircraft design. It also promises to be safer if one or more blades break loose when, say, the engine swallows a bird. But the technology wasn't in hand when the engine was launched. One difficulty was how to inspect the blades. By the time GE solved that and other issues, most of the first customers had gone with competitors.

But airframes grow, and that gave GE its chance. Succeeding models get stretched to carry more people or more freight. Both require more thrust. Initially the 777's engines were rated at 76,000 pounds, but that grew to 85,000 pounds, and then, fairly quickly, to 90,000, enough power to pull all the air out of Madison Square Garden in about one minute. At that point, recalls Chaker Chahrour, head of the GE90 program, the company decided "to deep-fry the competition." It had designed the GE90 for growth, convinced that Rolls and Pratt, which had already squeezed more thrust out of their old engines, would struggle to keep up. The revenue-sharing participants weren't eager to spend more on additional development but went along when GE raised the ante to 94,000 pounds. With that, the competition began to flag.

When Boeing began to talk about even more power for two new 777s--one with more payload, another with more range--GE decided to exploit its edge. It was sure that it could supply additional thrust and almost sure that Rolls and Pratt couldn't unless they developed a new engine. Says Chahrour: "We told Boeing, 'It's been brutal so far--really tough to have three competitors on every campaign. The only way we'll develop an engine for a new 777 is if we are the only engine on that plane.'" In July 1999, after a year of discussions, Boeing agreed. The 115B project got going.

By then GE was well along in changing the way it develops and builds aircraft engines. From 1991 to 1993, when airlines slumped and military purchasing declined, GE's aircraft engine revenues went down 15%, but its operating profit plummeted 43%. Something had to be done about costs. The company began to outsource more and to press suppliers to move their operations to low-labor-cost regions. It also created a network of in-house Tier 1s by moving chunks of manufacturing out of its huge, old, and long-unionized factories in Lynn and Evendale and into small plants around the country.

Today parts of Evendale, which sprawls over 6.4 million square feet, and Lynn, which has 3.4 million, are nearly empty. Both plants still make parts, and both assemble engines. Lynn does almost all the military engines. With parts from Snecma, Evendale assembles CFM56s destined for Boeing and builds "aeroderivative" engines for industrial use. But you might walk a block in an Evendale building before encountering a worker. Turbine and fan blades are now made in Madisonville, Ky., rotating parts are split between Lynn and a plant in Wilmington, N.C., small-diameter tubing is produced in Hooksett, N.H., and so on. Many but not all of the new, smaller plants were eventually unionized, but all got a chance to become profit-conscious specialists employing the arsenal of modern, lean manufacturing practices. The object, of course, was to cut engine costs and boost profits on aftermarket sales of parts, easily half of the business of many of the small plants.

While production scattered, the manufacturing organization itself was broken up into "centers of excellence" that have full responsibility for a group of parts, from design to shipment of the finished item. For example, Colleen Ahtans, a 43-year-old engineer, heads the center for rotating parts. Based in Evendale, she is responsible for two plants and has 230 design and manufacturing engineers who specialize in everything that revolves inside an aircraft engine. Ahtans says that installed base is "what I live for." In some years as much 70% of what her center produces goes out as parts and spares rather than to the assembly plants.

Bob McEwan, 49, runs the center for structural parts, the stuff that isn't supposed to move. Also operating out of Evendale, he annually buys about $400 million worth of material and parts, including all castings and forgings. But few suppliers have the equipment needed to machine a 1,300-pound forging into a 400-pound turbine frame. That's done in Evendale and takes 600 hours. Reporting to McEwan are about 350 production workers and 94 engineers, plus others under contract in Eastern Europe, India, and Mexico. Getting design, development, and manufacturing people into one organization is, he says, a big change. "In the old days manufacturing was in this big chimney, and engineering was in this big chimney, and they never talked to each other. Engineering did a drawing, folded it like an airplane, and sent it over the wall saying, 'Make this sucker.' We don't have that anymore."

GE's managers run out of adjectives trying to describe the small, non-unionized plant in Durham, N.C., where 120 employees assemble commercial engines. Says Chahrour, for whom Durham builds the GE90: "It's almost utopia." From the start, in 1993, the plant was planned to be different. Contract workers handle maintenance, receiving, and getting kits of parts to the assembly stations. Assembly is done by technicians who aren't hired unless they have an FAA Powerplant Mechanic's license. To earn the license, which isn't required in Lynn and Evendale, they have to pass the FAA's tough written, oral, and practical tests. They also need either formal schooling or on-the-job experience. Eight out of ten Durham people got their experience in the military.

Applicants go through a full day of tests, assemble a small engine, and are run through several interviews, including some with specially trained hourly workers. Get a single failing grade on anything from anybody and you're out. If three survivors get job offers out of 12 starters, it's a good day. One aim is to make sure the applicant can work as part of a team, since teams are responsible for planning how to do every job as well as who does it when. Teams even decide who's on first or second shift. There's a production-improvement leader on each team. He or she worries about schedules, quality, and cost but is not a "supervisor," insists one of them, Robin Robbins. Plantwide issues--materials, quality systems, training, etc.--are handled by councils made up of a leader and somebody from each team. The organization couldn't be flatter. Everybody effectively reports directly to Tim Scott, the facility manager.

While teams run the plant, individuals build engines. Subassemblies are put together in stations manned by one person. Only a couple of technicians assemble each engine. This isn't mass production. It's a Swiss watch factory with bigger pieces. A turbofan is built out of about 2,500 parts, some machined to such close tolerances that they can be joined on the assembly line only by chilling one of them with dry ice so that it contracts slightly. The pace looks leisurely. Everything is put together very, very carefully, for which you can be very, very thankful the next time you're back there tensely crammed in coach while the plane hurtles down the runway on takeoff.

Aircraft engines are initially assembled vertically, rising up out of a pit surrounded by a platform for workers and their tools that can move up or down. When the engine is nearly completed, it is hoisted into a horizontal position, but the assemblers and the parts still come to it. Doing it this way, Durham builds, at best, one GE90 a week, and every three days one CF6, the engine for wide-body aircraft other than the 777. The high-volume product in this business is the CF34, the regional-jet engine. Durham's target is one every two-shift day. Two years ago the plant started GE's first aircraft-engine assembly line: a moving ten-by-ten-foot platform that carries the engine and technicians. On its left it passes by subassembly cells. On the right it moves by racks holding parts and special tools. Recently the line was lengthened to 200 feet, and the speed was doubled--to three feet an hour.

In addition to changing the way it builds engines, GE set out to cut the time it takes to develop them. When military spending slowed in the early '90s, the company shed engineers, from 8,500 to 4,100 in 1995 and 3,900 today. If GE was to have a wide range of programs underway, it had to do more with fewer bodies. To cut costs, it also had to fit within the aircraft manufacturers' development window. While Boeing and Airbus could move an airframe from start to flight in three years, GE was taking five to get an engine from launch to FAA certification. Recalls Calhoun: "We had to start doing stuff even before they [the airframers] had a clue about what they were going to build. The GE90 took a lot longer and cost us a lot more than it should have."

In 1995, Mike Benzakein, head of advanced engineering, formed a team charged with reducing development time to 24 months. This was to include 14 months to get to the first engine and ten more months for testing--a target his boss, vice president of engineering Corbett Caudill, admits was "pulled out of the air." Seven subteams took on different parts of the engine, mapped all the steps to get from idea to product, and then figured out ways to whittle time out of each. Better computer-based analytical tools helped. So did the lesson of the composite fan blade. Now, says Benzakein, "we won't launch until we're sure we have the technology. We can't afford inventions." By 1998 they had it figured out. Today Benzakein and Caudill say there's no other way their stable of engineers could simultaneously work on so many commercial-engine projects plus several for the military, including a new engine for the F-16 and one for the M-1 tank as well as the F-35 project. A true GE executive, Calhoun recently said that if 24 months is possible, why not 18? That's next year's goal.

The GE90-115B was one of the first engines to be developed in the new manner, but the schedule has slipped to what's likely to be about 28 months. One completely new design, the CF34-10 for regional jets, should be on target. Given Calhoun's expectations for this market, it had better be. From 1993 to 2001 regional airlines more than doubled the revenue passenger-miles they fly in the U.S. and about tripled them in Europe. Even in the present doldrums, they've kept buying planes, all now built by Montreal's Bombardier or Brazil's Embraer. GE is the only supplier of engines for Bombardier's regional transports. Rolls-Royce builds the engine for Embraer's 50-passenger plane, but the CF34-10 will power its new 70- and 100-passenger models. The two aircraft builders have more than 2,400 engines on order at GE. More will shortly head to the Far East. In November the CF34-10 was picked to power two versions of a regional jet that a Chinese consortium intends to have flying in time for the 2008 Olympics.

"In the next couple of decades," says Calhoun about engines for regional airliners, "this will be the franchise for our business. It's the biggest of our expenditures in the current 24-month period. And it's a real winning bet." That Durham assembly line might even have to speed up to four feet an hour.