They look like embryonic pearls or like the ball bearings for some fantastic miniature machine: two-dozen exquisitely smooth, silver-gray beads alive with reflected light in the slender glass vial that Akira Ishikawa rolls between his thumb and forefinger. in fact, the beads are the product of Ishikawa's disruptive vision -- spherical semiconductors that could cut the cost of computer-chip manufacturing by a factor of ten.

"Many people thought this was crazy," says the 64-year-old silicon wizard. "That's all right. The history of semiconductors for the past 40 years has been this way -- we have faced so many thick walls, and when we break through, there's another barrier." Craziness, after all, is just another name for crashing through more barriers than anyone else.

So here is Ishikawa's barrier: Nearly everything we know about the delicate, exacting process of building microscopic transistors on silicon is geared to making flat computer chips. Until Ishikawa, no one had tried to etch a semiconductor's tiny circuitry onto a curved surface, much less onto a sphere. Ishikawa and his team first had to invent some way to make a single crystal of silicon in the shape of a sphere, which they accomplished by dropping molten silicon through a tower that cooled the droplets precisely as they fell. They then wrote software to design circuits on spherical surfaces without distorting the physics of electrons that whiz through wires thinner than a human hair.

Finally, they figured out how to float the spheres through gas-filled pipes where slivers of material a few atoms thick were stripped away without ever allowing the spheres to touch the pipe walls -- a critical breakthrough to avoid contaminating the devices with tiny bits of dust and debris. Herein lies the enormous potential economic payoff: Because the spheres don't require immaculate "clean rooms," Ishikawa says, a factory to make them can be built for about $100 million -- compared to $1.5 billion for conventional chips.

The audacity of Ishikawa's vision has attracted skeptics -- but important allies too. Ishikawa formed Ball Semiconductor Inc. in October 1996 and has raised $80 million from such companies as Hitachi and Maxell. Bright young engineers from such countries as China, Japan, and the United States have found their way to Allen, Texas, outside of Dallas, where the company is based -- a total of 35 employees who sought out Ishikawa. "That is their choice, their way of taking a risk," he says. "I did not ask them to come, but they came."

Ishikawa's reputation as an original thinker with a good track record seems to be the one reason that Ball Semiconductor is not dismissed out of hand as a crackpot scheme. "I look at it and think, 'This is nuts,' " says semiconductor-industry analyst Steve Cullen of Cahners In-Stat Group, who does question the feasibility of Ball's approach. "But Ishikawa is not the kind of guy who would be involved in something that's nuts."

It's hard to argue with his credentials. As president of Texas Instruments Japan, Ishikawa led the company to win Japan's coveted Deming prize for manufacturing excellence. He moved to Dallas in the mid-1980s and built a thriving career in TI's freewheeling culture. But the idea for a spherical semiconductor kept gnawing at him. At first the idea was vague and formless, a brilliant abstraction about the surface area of a sphere, which is three times larger than the surface area of a flat chip. In an industry where the amount of real estate available on a bit of silicon is everything, Ishikawa's insight had huge implications -- all of which seemed impossible.