A thumb drive that stored a terabyte of information, however, would have to take advantage of two other characteristics of nano-ionic memory, Kozicki says. First, it would have to store more than one bit of information per memory cell. Once the wire inside the cell forms, it's possible to apply a voltage again, causing more atoms to form, thickening the wire and further decreasing resistance. Successive jolts will thicken the wire yet more, and the different states of resistance could be used to store multiple bits of information per wire.

What's more, this type of memory can be stacked up in layers, since it's not necessary for each cell to be in contact with a base layer of silicon, as is the case with some other types of memory. Combining multiple bits per cell with multiple layers could make it possible to form extraordinarily dense memory, Kozicki says.

William Gallagher, a senior manager for exploratory nonvolatile-memory research at IBM Research, says that nano-ionic memory is one of several promising next-generation memory technologies. These include MRAM, which stores information using magnetic fields, and phase-change memory, which stores information in a way similar to that used to store bits on DVDs. Gallagher says that ionic memory's competitors have a head start on it. MRAM chips are already sold for some special applications, such as devices that will be exposed to harsh environments. But MRAM may also prove better for high-speed memory applications than as a replacement for flash, so it may not compete directly with nano-ionic memory. Samsung, however, could be selling a phase-change-based flash-replacement memory within a year.

Still, nano-ionic memory may not be far behind. A few companies have licensed nano-ionic-memory technology developed at the Arizona State University. These include Qimonda, based in Germany; Micron Technologies, based in Boise, ID; and a Bay Area stealth-mode startup. The startup is well on the way to producing its first memory devices, which Kozicki says could be available within 18 months. These first chips, however, won't rival hard drives in memory density, he says.

The new technology could nevertheless have difficulty winning wide adoption. Flash-type memory continues to improve and may do so for a few more generations of products. Also, the best nano-ionic-memory prototypes have been made from materials that aren't used in conventional microchips, so manufacturing could be costly, at least initially. Kozicki's group recently demonstrated that ionic memory can be built from a combination of silicon dioxide and copper--materials that are compatible with conventional manufacturing. But these materials do not perform as well, which could make them less attractive than alternatives such as phase-change memory. For the new type of memory to succeed, it may be necessary to convince manufacturers to switch to new materials.