In his cramped cubicle at Nanomix, a nanotechnology company in Emeryville, CA, just across the bay from San Francisco, theoretical physicist Seung-Hoon Jhi peers at a computer model of a hydrogen fuel tank, carefully tracking the movement of individual molecules. As he raises the temperature of a simulated sheet of boron and nitrogen atoms from a frigid 50 Kelvin to a slightly less chilly 80 Kelvin, he watches the reaction of a handful of hydrogen molecules dotting its surface. The boron nitride sheet undulates, yet the hydrogen molecules hold fast. It's an encouraging sign in a virtual experiment that may have just saved weeks or months of painstaking experimental testing in Nanomix's effort to develop more efficient hydrogen storage materials for fuel cell cars.

It's cyberdreaming, of course. But Jhi and his Nanomix colleagues are so confident in the veracity of this computerized modeling, pieced together from precise calculations of the behavior of individual atoms, that they are using the simulations to design and test materials that have never been made before-materials whose ordering at the nanometer scale (a nanometer is a billionth of a meter) can produce properties useful in applications ranging from ultrasensitive sensors to flat-panel displays to stealthy coatings for war planes. Down the hall, less than 15 meters from Jhi's cubicle, the company's experimentalists are busy working in the lab to synthesize the most promising results of the modeling.

While Nanomix is just one of several recent startups hoping to exploit nano materials, the company is betting it has an edge: the skill to both virtually design the materials-without so much as stirring a beaker-and then go into the lab and make them. Its cofounders-theoretical physicist Marvin Cohen and experimental physicist Alex Zettl, both from the University of California, Berkeley-have been collaborating on such alchemy for over a decade. Now they are hoping to leverage that expertise as the basis for a nanotech business. "Our goal is to have the first working nano components on the market," says Nanomix CEO Charles Janac.

Designing materials on computers has tempted industrial researchers for more than a decade. In theory, at least, the idea is simple enough: using the rules of quantum mechanics it is possible to calculate the behavior of the electrons that swirl around an atom. Given enough computing power, one should be able to use such calculations to design a material atom by atom, building in desirable properties by adjusting the electronic profile. The problem is, the properties of materials result from the interactions of a huge number of atoms. And even today's most powerful supercomputers struggle with quantum calculations involving more than five or six hundred atoms, severely limiting the ability to design new materials.