Call In The Margin-Buster

Nayel Shafei is holding an OC-48 line card in his hand. This piece of heavy metal - it's about the size and weight of a laptop computer - allows data to be transmitted at the rate of 2.5 billion bits per second. Pretty fast. Sending The Tonight Show with Jay Leno as an uncompressed stream of high-definition television consumes 1.5 billion bits per second.

To make a 2.5-gigabit-per-second connection, you need two line cards, for reception and transmission. The total cost of these two puppies is on the order of $312,000. By the ounce, they're probably worth more than gold.

But Shafei is on a mission. He wants to unleash the still-simmering broadband economy. To him, it's way too expensive to provide "end-to-end'' transmission of multimedia communications. And he's looking to destroy costs, not just tweak them.

The former Qwest Communications International executive vice president of products has struck out on his own. In about five months, he's lashed together a network as president of a New Jersey start-up called Enkido that rivals that of his former employer - at a fraction of the cost. The secret: Leasing or buying existing capacity in major markets across the country. No digging ditches or underneath streets.

But he's also driving down the cost of networking gear. Take that line card. Inside is something called a pump laser. It's the active ingredient. By itself, it only costs $15. Surrounded by a lot of protective casing, it goes for $200,000 on one end and $112,000 on the other. Nice markup, if you can get it.

Shafei's idea is to obliterate such margins. Instead of going to conventional gear providers such as Lucent Technologies or Nortel Networks, he sought out a chipmaker called SoftCom. His pitch: Put the laser on a PC card.

Now, Intel owns SoftCom - and is turning out OC-48 cards for $500 apiece. The $312,000 coupling has gone down to $1,000. Shafei is a step closer to igniting corporate and governmental use of broadband communications.

But many such steps are needed, from line cards to routers, if the promise of fiber-optic technology is going to make the delivery of all kinds of videoconferencing, video programming and multimedia information-on-demand on corporate networks commonplace.

There's now more than enough capacity going into the ground to sate foreseeable demand for visual communications. Qwest, Level 3 and friends are together putting down 80 trillion bits of capacity.

But the real obstacle is turning on the fiber. Lighting up each 10 billion bits of capacity costs $250 million. Lighting all 80 trillion bits would cost $4 trillion. At current costs, it isn't going to happen.

Don't hold your breath waiting for the Lucents and Nortels of the world to wring out that much cost from lighting up fiber and igniting the broadband economy. They have profits to protect and customers to take by limo to luxury suites at football games.

Count instead on guys like Shafei.

If you want creative destruction - if you want a low-cost, high-capacity broadband economy - find destroyers of margins. Not protectors.