As Geometries Get Smaller, Keeping Up is Harder

Ally Winning, European Editor, PSD



There were two stories on the news recently that show both the potential for smaller and smaller electronics, and at the same time how hard it is to bring those designs to fruition. Firstly, researchers at Karlsruhe Institute of Technology (KIT), led by Professor Thomas Schimmel, have developed a single atom transistor. This is a major breakthrough for the electronics industry. The transistor switches the current by a controlled repositioning of a single atom. What’s more, the transistor operates at room temperature, consuming very little energy. If this was scaled up to the billions of gates found in modern processors, and the billions of processors throughout the world, the energy savings would be enormous.

For the demonstration, the researchers took a new approach, using metals instead of semiconductor material. Metals offering better conductivity, meaning that the device operates on low voltages and consumes very little energy. The biggest test will obviously be taking the discovery out the lab and scaling it up into a useful product for manufacturing.

This is where our second story comes from this month. As geometries move lower and lower in the nanometre domain, devices are proving hard to manufacture and monetize. The cost of building fabs and providing equipment for the latest geometries is exceptionally high. Too high for GlobalFoundries, one of the biggest manufacturers of ICs. The company announced recently that it was stopping all development at the latest 7nm technology node. Instead, it will concentrate on making specialised ICs at the larger 12 and 14nm nodes. It is thought that the company lacked the orders from leading edge companies willing to pay the extra money to get semiconductor using the 7nm technology node. This only leaves TSMC, Samsung and Intel developing for 7nm technology at the moment.

The electronics industry is always moving forward, and in the past there have been many early adopters willing to pay out to have ICs fabricated at the latest node. However, the 12 and 14nm nodes seem to have the right performance, power and cost metrics for many manufacturers, and those nodes are expected to have a longer shelf life than previous bleeding-edge nodes. But it does beg the question that if there are only three companies willing to pay for the ability to manufacture at 7nm, how many will be able to afford to manufacture chips made from single atom transistors when the time comes?