(FORTUNE Small Business) – Out in the Arizona desert, Jerry Knoblach is preparing to launch the world's cheapest "satellite." A colleague counts down: "Three ... two ... one...." Then Knoblach simply opens his hand and releases a weather balloon with a lightweight communications device dangling underneath. Filled with hydrogen, the balloon bobs in the early morning thermals, drifting skyward, destined for the far reaches of the earth's atmosphere.

This may not sound like the stuff of aerospace revolutions, but Knoblach is a genuine pioneer, at least in terms of cost savings. A typical rocket-borne communications satellite today costs upwards of $100 million to build and launch. Knoblach's company, Space Data of Chandler, Ariz., can get one of his devices in the air for only $400. Because he needs to release a balloon every 12 hours, the annual cost is around $300,000. How does he do it?

Instead of using a rocket to blast a satellite into space, Knoblach sails his transponders up to the stratosphere tied to an inexpensive weather balloon. Because it's higher than storm clouds and heavy winds, yet closer than space, the stratosphere is a tantalizing place from which to transmit signals for cellphones, pagers, and other messaging devices. If you park a kind of poor man's satellite up there, it's possible to undercut the prices of existing telecom providers or fill in their coverage gaps.

For the past decade, all kinds of stratospheric dreamers have been in the chase, attempting some truly wacky inventions, including gargantuan solar-powered dirigibles and tiny planes intended to fly unmanned for months at a time. Only Space Data has succeeded. Right now, it serves one narrow customer niche: oil companies that want to monitor wells and pipelines by remote. But telecom balloons have big-bucks potential, such as offering cellphone coverage to underserved rural areas. "Think of this as either a very tall cell tower or a very low satellite," says Knoblach. "It just happens to be a balloon."

But like all space ventures, this one is risky. So far, Space Data has a tiny revenue trickle: $50,000 in 2004. That's set against a ferocious $750,000-a-month burn rate. Knoblach, 42, is smoking through the $33 million he has raised, mostly from his deep-pocketed dad. But he's confident that he can turn an operating profit by 2006. So-called remote telemetry--providing messaging from isolated oil wells and chemical depots--is a turbocharged market, projected to grow to nearly $4 billion in 2008, from $800 million in 2004. Sign up a few big customers, such as an ExxonMobil or a DuPont, Knoblach figures, and Space Data can float into the black.

But if Knoblach hopes to strike it really rich--not merely survive--at 100,000 feet, he'll need a more diverse customer base. Cellphones are the big opportunity. Currently, 40% of the U.S. land mass lies outside the range of cellphone service. Big wireless players such as Cingular and Verizon aren't in a rush to dot the countryside with expensive cell towers because much of that land is sparsely populated. But they might be amenable to letting Space Data become a roaming partner, thereby extending their networks into unserved rural areas.

Still, convincing established cellphone players to partner with a balloon outfit may be a tough sell. Right now, Space Data's relaying of electronic messages allows some margin for error. Say a message is delayed by two seconds. No one really notices or cares. But voice data is more complicated to transmit, and consumers are far less forgiving. If Space Data is going to crack the cellphone biz, it will need to guarantee a crisp, uninterrupted signal. Knoblach says he's addressed the issue: "We know we can transmit a voice signal reliably because we've done it in tests." Some experts remain skeptical. "Their approach is so out of the mainstream," says Jeff Kagan, an Atlanta-based telecom analyst. "I think there's going to be ample suspicion until the concept is proven."

Knoblach grew up in St. Cloud, Minn., during the 1960s space race. He caught all the Apollo moon missions on television and built a fleet of Estes model rockets. In 1992, after graduating from MIT and Harvard Business School, Knoblach joined Orbital Sciences, an aerospace company based in Dulles, Va. He quickly grew disillusioned with traditional satellites, stunned by the regulatory hurdles and, most of all, the astronomical costs. Before long, Orbital Sciences transferred Knoblach to a division that built tracking devices for weather balloons--and inspiration struck. It was not exactly mission-critical at Orbital Sciences. It was not where the big money was either spent or made. But Knoblach became convinced that there was serious, unrealized telecom potential for weather balloons.

In 1999, Knoblach quit his job and set up operations in--where else?--a garage. At the outset, he faced a major R&D challenge: Unlike satellites, weather balloons aren't designed to remain aloft. Rather, they travel steadily upward until the atmosphere becomes so thin that they pop. Knoblach needed to find a way to level them off so they would bob along in the stratosphere at roughly 100,000 feet.

Basically, the weather balloon--a century-old technology--needed an upgrade. To maintain balloons at altitude, Knoblach developed a system for venting (releasing hydrogen so they fall) and ballast (dropping silicon powder so they rise). It's a staple of hot-air balloons, but nothing like it existed for weather balloons. Knoblach also helped design an in-flight computer that weighs only ten ounces. It measures altitude and wind speed and sends the balloon commands to keep it level.

Even after he solved the altitude problem, Knoblach found it impossible to keep balloons aloft for long stretches. The telecom pod and flight computer are powered by lightweight batteries that can summon only so much juice. The solution: Send up a fresh balloon every 12 hours. After the replacement arrives, the old balloon cuts loose its payload of gizmos, which drifts to earth under a small parachute while emitting a GPS signal that allows it to be recovered. It's worth the trouble, because the electronics cost about $1,000.

By 2000, space data had a working prototype. But if the company was going to provide telecom services, it needed the right to transmit and receive messages on certain wavelengths. Knoblach scored his first such spectrum rights from a bankrupt paging company. Originally, the license cost $80 million, but he got it for $3.6 million. (It expires in 2014 but is renewable.) He bought a second chunk of bandwidth at a Federal Communications Commission auction soon after 9/11. Space Data was one of the few companies that participated. It obtained a 50% discount from the FCC--dropping the price to $4.5 million--in part because it agreed to provide future services to various underserved Indian tribes in remote locations.

Knobloch's father had made a fortune twice over, first as a St. Cloud real estate developer, then as an early biotech investor. That was lucky, because nobody else was buying Knoblach's vision. "I'd meet with a venture capitalist," he recalls, "and the instant I mentioned balloons I'd get this stare like I was a complete idiot."

Space Data's headquarters are tucked away in a bland office complex in Chandler. Inside, a large digital-projection map shows the real-time coordinates of the company's fleet, a constellation of 11 balloons hovering above the south-central United States. Each balloon can broadcast to an area the size of Oklahoma. (By contrast, a cellphone tower has about a ten-mile range.) Space Data employs about 70, including balloon launchers in 11 cities, among them Sweetwater, Texas, and Tucumcari, N.M.

After the telecom pods parachute to earth, Space Data furnishes freelance wranglers with GPS coordinates. The wranglers are folk with the time and inclination to drive, for a modest fee, as far as 500 miles to recover the pods.

Space Data's customers in the petroleum industry have relied on satellites to relay information about temperature and flow from remote sensors back to headquarters. Now Space Data is taking on the satellite guys. "It's a wonderful service," says Rafael Santiago Jr., an engineer with Riata Energy, a gas company in Amarillo, Texas. "At first there were a lot of doubters in my company, people thinking, 'Balloons?' But then the data started streaming in, and the doubters were silenced."

The price is right, according to Santiago. For each remote monitor--Riata has hundreds installed--satellite service is roughly $300 a month, vs. about $50 for Space Data.

But if Space Data is going to be more than a niche player, it must figure out how to get into business with the telecom giants. Knoblach says he's in talks with a major cellphone provider and expects to be beaming its signal into rural areas within three years. Who knows? Maybe one day you'll be placing a call from the floor of the Grand Canyon, thanks to one of Knoblach's homely weather balloons.