Year/Course: 2023-2024, Easter 2024
Project type: Innovation

Inventors: Professor Judith Driscoll & Dr. Markus Hellenbrand, Materials Science & Metallurgy

8-10% of global electricity useage is used by computing and communications technologies, and this may rise to 20-30% within the next few years. This means that the progressing digitalisation of the world, the emergence of the Internet of Things and recent rapid progress in artificial intelligence applications present immense challenges to the global electricity supply.

Large parts of this electricity requirement is due to data centres, and the way in which data is stored and processed in the currently-used von Neumann computing architecture. Memory and processing are separated, requiring data to be moved back and forth between them with a significant energy cost. This is especially pronounced in emerging AI systems, which are based on hundreds of billions of model parameters, all of which are essentially memory entries.  

Various new memory technologies are being researched to overcome these problems. One of these is based on a phenomenon called resistive switching, but this new type of memory faces challenges of uniformity and reproducibility. 

The inventors in this project have developed a type of resistive switching memory which shows great promise in terms of these challenges of uniformity and reproducibility, and a corresponding patent application is currently at PCT level.

The new type of memory has a range of possible applications in the computer sector. It could be used to develop replacements for flash memory (current types of flash memory will not work as embedded memory as silicon technology moves below 40nm ), or to develop new neuromorphic computers which are optimised for the requirements of AI techniques. 

The challenge for the i-Team is to investigate the market potential for this new material, identifying the key players in the field of emerging AI technology and memory, and recommending the next steps for the inventors to ensure that their materials meet the required metrics for this next stage of the computing industry.