Background
Billions of smart speakers, wearable devices and factory sensors are used globally, with experts predicting that 1 trillion will be in circulation by 2035. Each device depends on built-in memory, but today’s standard, Flash, is reaching its limits in terms of technology and energy efficiency.
The global semiconductor market is now worth over £400 billion, with AI hardware demand growing by 20% per year. Data centres already produce about 3% of the world’s CO₂ emissions – similar to aviation. Meeting net-zero targets urgently requires memory that is smaller, faster and more energy-efficient.
About the Project
The Royce-supported project “High-Density Embedded Non-Volatile Memristor Technology for Edge AI Applications” set out to create a new kind of memory called Resistive Random Access Memory (RRAM). Unlike Flash, RRAM stores bits by switching the resistance of a nanometre-thin oxide layer and can do so at low voltage.
Project Details & Results
Royce funding via the Industrial Collaboration Programme supported a post-doctoral researcher, wafer processing at the London Centre for Nanotechnology and advanced characterisation at Royce facilities in Manchester, Cambridge and Imperial College London. SME partner Intrinsic provided wafers, staff time and testing equipment, and has since invested in further research. Access to 300 mm silicon wafers from IMEC plus rapid feedback from Royce Application Scientists allowed quick design–build–test cycles that a small company alone could not afford.
The team removed the bulky selector transistor normally required by RRAM, producing a “self-selecting” cell that saves both space and energy. In practice this means the new memory chips use very little power, making them ideal for battery-powered devices. They store and retrieve information almost instantly, fast enough for real-time artificial intelligence on small devices. Early versions have already handled over 100,000 rounds of saving and deleting data and are expected to safely store information for at least 10 years, even in harsh temperatures. The design is also compatible with today’s chip factories, making it easier to bring this technology into everyday products.
Electron microscopy at Royce Manchester confirmed the uniform oxide layers that make the device operate, while X-ray photoelectron spectroscopy at the Universities of Manchester and Cambridge provided further insights on the device operations.
Impact
Economic & industrial
Follow-on investment ensures continued high-value R&D activity in the UK and supports the development of future licensing revenues. The new memory cell design can reduce chip area and lower manufacturing costs for microcontrollers.
Societal & environmental
By enabling lower-power, high-density memory, the technology supports local processing of AI tasks on devices such as wearables and autonomous vehicles. This can reduce reliance on energy-intensive data centres and help lower overall carbon emissions.
Knowledge & capability
The project enabled a post-doctoral researcher to gain experience in advanced semiconductor fabrication and characterisation, supported by engagement with industrial partner Intrinsic. Collaboration with Royce researchers provided valuable feedback on materials deposition techniques, contributing to more effective device development.
Collaborators
A University College London (UCL) team, working with Intrinsic Semiconductor Technologies and supported by Royce ICP, has built a transistor-free memory chip. The tiny device stores data faster, lasts longer and uses far less energy than the current Flash technology, opening the door to slimmer batteries in phones, drones and future AI wearables.
"This project addressed a critical bottleneck in the density of embedded memory by demonstrating transistor less, non-volatile RRAM devices that have the potential to dramatically improve low-power AI devices, strengthening the UK’s strategic position in the global semiconductor market."
Dr. Adnan Mehonic, Associate professor and RAEng Senior Research Fellow
University College London
