(FORTUNE Magazine) – I am fascinated with a new product just announced by AT&T. The company has developed it in great secrecy over the past three years under the code name "Project Angel." To be accurate, AT&T is calling Angel a technology rather than a product, which allows the company to avoid giving the thing a name, a price, and a release date. Angel is a radio technology intended to let AT&T provide local telephone service and high-speed Internet access without having to run a wire to anybody's house. I suspect that Angel will have unintended effects, and may even be the catalyst for the next leap forward toward the information highway.

Getting enthusiastic about any new AT&T product is dangerous for a sober analyst concerned with maintaining his reputation because the company has been in the pits recently, even in its core business of connecting long-distance calls. But getting excited about AT&T is particularly dangerous for a computer analyst. It's not that AT&T is genetically stupid; it's just that the company's experience in the computer business is such a sordid tale.

AT&T never succeeded in its attempts to supply computers for general use, going back to the mid-Eighties, when it contracted with Olivetti to design and build a line of personal computers. Six years ago AT&T bought a computer company called NCR and lost nearly $4 billion on it before spinning it off again. AT&T invested in a company called Go, which helped make personal digital assistants, but Go soon folded. AT&T also invested in an online gaming network, which is now owned by America Online. And it tried to build a microprocessor business but withdrew before it really got started.

So when it comes to computers, AT&T has pretty much messed up everything it's touched. To be fair, AT&T has had much more success with its recent WorldNet venture, which provides access to the Internet for individuals and business customers. The company has signed up more than 850,000 subscribers in less than a year, making it one of the top Internet providers. Many analysts would say AT&T's success here isn't so surprising, since WorldNet involves hooking people up to the telephone network, AT&T's specialty. The rub is that most of the telephone calls WorldNet generates are local ones by people logging on to the system, and AT&T doesn't carry local traffic now. But the company can hardly stop drooling over the prospect of reentering the local telephone market, which is much bigger than the long-distance business and (so far) much less competitive.

So it's against this backdrop that AT&T announced Angel. AT&T calls Angel revolutionary because it will allow the company to cram tons of phone calls and data into a tiny piece of the radio spectrum--and more efficiently than its cellular phone systems can, even its newer digital ones. That's because AT&T developed some clever new algorithms, and because houses don't move around, so AT&T can use directional antennas instead of ones that send signals in every direction.

The part of Angel you'll actually see is an 18-inch-square box AT&T slaps on the side of a house or small business. AT&T then connects the box to the same wires that run indoors to your phones today. When you make a call, a bunch of radio stuff inside the box sends a signal through the air--look, Ma, no wires!--to an antenna close by, which in turn connects directly to the AT&T network. Once you have this box, you can actually cut the copper wire that goes from your house to your local telephone company and have all the services you do today. That's right: For the first time, you'll have a choice about how to get your local telephone service.

Angel isn't perfect. First, the box won't work if you have a power failure. The local phone company has always provided its own power through the phone wire so that the telephone system works independently of the power system. (Angel will have a battery backup. If the power goes down, it will last for 24 hours on standby or for four hours of talk time.)

Second, AT&T has to make sure you can call your next-door neighbors if they don't have an Angel slapped on the side of their house. That means AT&T will still need to buy access to the local phone system, making it dependent on the phone company. And actually, I exaggerated a little when I said you've never had a choice for your local phone service. In some areas, cable companies have started to provide phone service too, though the idea of trading your local phone company for your local cable company is not necessarily appetizing.

And speaking of having radios attached to your house, I never did get that satellite dish to replace my cable channels. It turns out that the hill behind my house is high enough so that the dish couldn't see the satellite. I hereby ask TCI for forgiveness and hope they won't send snow down my cable for all the nasty things I said about them--at least until I move to a new house, without a hill behind it.

Those problems aside, the idea of being able to thumb your nose at your local phone company is powerful enough to give Angel a good chance, if the technology works. But Angel offers another benefit that for computer lovers is pretty compelling. Angel can connect your computer to the outside world at high speeds and at very low cost.

In fact, it's this breakthrough in Internet access that turns out to be Angel's secret sauce. The fastest modems widely available today run at 28.8 kilobits per second. Local phone companies are increasing the availability of ISDN, a digital service with an Internet connection that's four times as fast. But they charge a fairly high price and have difficulty installing the technology.

AT&T's Angel box reportedly provides a modem connection at the same speed as ISDN. But Angel will let AT&T provide high-speed Internet service much more cheaply than the local phone company can today. A lot of AT&T's cost lies in installing banks of modems all around the country. They are the entry points to AT&T's network, allowing you to reach the Internet with only a local call. Without going into a lot of boring technical reasons why, Angel's digital signal means that AT&T won't need all those modems.

Being primarily interested in the telephone business and particularly in competing with local telephone companies, AT&T appears to be costing out its system so local phone service alone will pay for its capital investment. AT&T has yet to determine its actual costs. But let's say the Angel box on the side of your house costs AT&T around $300. A neighborhood antenna serving 2,000 houses might cost $200,000, or $100 a household. You'll need a modemlike device to attach your computer to Angel, as you do with an ISDN line. ISDN adapters cost about $200 these days. Even with some back-office expenses, and figuring about $50 per household for the license fees AT&T had to pay for its spectrum, the company should be able to install Angel service for far less than the $1,200 or so a local phone company must pay to install a conventional line.

You might argue that the Baby Bells have already paid for the phone lines that run to homes and offices today, while for AT&T the capital costs of Angel are still looming. But a lot of the Baby Bell costs aren't depreciated yet. Besides, the phone system isn't static. Local phone companies are replacing lines and adding second and third lines to homes all the time. And when they do, they often have to string new copper wire on a pole, or bury it.

Bottom line: You pay AT&T $10 a month for Angel service (a number reported in the press but unconfirmed by AT&T), and the company slaps an Angel on the side of your house. You then make phone calls, both local and long distance, for about the same price you do now, but you get access to the Internet for less, and at four times the speed. This begins to make a fairly compelling story.

Computer snots will instantly see the hole in this analysis: Four times the speed of a modem ain't all that fast. What we really want is a connection to the Internet as fast as our network connection at the office, which is at least ten times faster than ISDN. Cable companies have high-speed cable modems. Local phone companies are developing a technology called ADSL, which is also much faster than ISDN. A little further out are new technologies like fiber modems and low-earth-orbit satellites.

But that's just the point: All these things are a little further out. Cable modems are cool if you are lucky enough to live in places like Boston or Phoenix, which have enlightened cable companies (Continental Cablevision and Cox Communications, respectively). Otherwise, you'll have to wait for the alternatives, probably longer than for Angel, which will begin testing in Chicago later this year.

Angel has the feel of a technology that can change the competitive landscape. It may force local telephone and Internet access companies to reconsider their assumptions. The phone companies have always designed their networks and accounting systems to support the best possible service at the highest price they could justify to regulators. Faced with real competition, they might decide they can actually provide phone service a lot more cheaply.

STEWART ALSOP is a partner with New Enterprise Associates, a venture capital firm. Neither he nor his partnership has financial interests in the companies mentioned. Alsop can be reached at