Last year, HP made news by demonstrating a practical application of memristance. At the time, I was astonished that the development had occurred so soon after HP's announcement that it had discovered a way to build memristors. It turns out, though, that Hewlett-Packard's press relations may be better than its patent claims. This, in turn, could be great news for investors.
HP may have been only the first group to recognize what it had on its hands. Researchers from such varied institutions as the University of Parma in Italy to UC San Diego have also built prototype memristors from polymers and metallic oxides. They too are exploring applications for this exciting new technology and could end up holding important memristor intellectual property.
Multistate Computing and Neural Networks
Nearly all existing commercially available transistor-based technology is capable of assuming only two states per element, either 1 or 0. So by necessity, all calculation is done in binary.
Memristors, because they can assume different states corresponding to different levels of resistance, are multistate elements. This facilitates a much higher data density. Memristor storage density will be at least 10 times that achievable using current transistor-based technology.
Imagine the storage capacity of a large hard drive on the head of a pin. Moreover, memristor memory is nonvolatile. It retains its state even when no power is applied to the circuit. This has tremendous advantages over current memory technologies that lose their data when the power is switched off.
All these unique properties put off the eventual physical limitations imposed upon Moore's law using current transistor technologies. Unlike current transistor technology, memristance becomes more pronounced and efficient the smaller the element is. In transistors, small size and high density lead to greater power loss and heat production. The opposite is true with memristors. Nanotech-level scaling actually amplifies the memristive properties of the individual elements.
It's not only computer scientists who are excited by these developments. Biologists are beginning to realize the potential memristors have to mimic organic or biological computing.
Because many of the properties of memristors are so similar to brain cells they may be used to imitate brain functions. If, as scientists believe, they can be used to mimic synaptic function, they could bring true artificial intelligence much closer.
Recently, researchers have been able to model the learning ability of the amoeba with a simple memristive circuit. According to HP, these circuits can "remember and associate series of events in a manner similar to the way a human brain recognizes patterns." In other words, the circuits learn.
While HP has grabbed the headlines, such devices are currently being developed for use as nonvolatile resistive memory by various companies. Some, like HP, are probably too big to be breakthrough technology top stocks for 2010. Samsung is one of the giants working on the technology. On the other hand, Micron Technology and Unity Semiconductor apparently have some patent rights, at least, to memristor technologies. If they're significant, these companies could be small enough to experience transformational profits that rival Intel's historic growth.
It is only a matter of time before this new technology begins to break through. We will be monitoring this area for developments and will keep you informed.