Symposium MT07-Materials Research Opportunities for Energy-Efficient Computing

Between 2012, the advent of AlexNet which enable high quality image recognition and the first large language models (LLMs) in 2019, the computing power required to train the models has increased by more than 300 000x. Newer generative AI models use even larger sets of parameters resulting in further increase by more than 10 000 x compared to those first LLMs. This spectacular increase in computing power stems from advances in semiconductor technology both from scaling and packaging and has been enabled by adoption of new materials and processes as several levels. And while the models build-up has been enabled by increased efficiency of the computing system, the wider integration of AI and generative AI in daily lives, continues to increase world energy use bringing an unprecedented need for energy efficient data movement and computation. Data centers consume more than 200 TWh each year, exceeding the total energy consumption of entire countries. Making computation more energy efficient while enabling blazing fast data rates has become the semiconductor technology. There has been a collective effort among academia, industry, and government to explore multi-faceted approaches for advancing energy efficient computing. Making computation more energy-efficient saves money, reduces energy use, and herald the advent of entire systems built into high-performance.

New materials make possible compute hardware stack elements needed for this revolution. Starting at the smallest scale, there are switching elements that comprise logic and/or memory. These elements are put together at the package level with passives and thermal management solutions by employing fast, low-energy connections to form the most advanced systems-on-a-chip. As materials touch every one of these components, the goal of this Symposium is bringing to front and center the challenges and opportunities that will accelerate materials innovations that enable energy efficient computation. There is already a substantial body of knowledge in this field spanning multiple approaches. The Symposium co-organizers have decided to will focus concentrate specifically primarily on the areas described below and will not cover quantum computing-related materials research and development.

• Materials for Beyond Moore's Law Computing and Information Technologies Materials for Energy: Conversion, Harvesting, Storage, and more

• Wen-Hao Chang (National Yang Ming Chiao Tung University, Taiwan)
• Ya-Ping Chiu (National Taiwan University, Taiwan)
• Vina Faramarzi (ASML, Netherlands)
• Ronald Gull (Synopsys, Switzerland)
• Boubacar Kante (University of California, Berkeley, USA)
• Tsu-Jae King Liu (University of California, Berkeley, USA)
• Pierre Morin (imec, Belgium)
• Tomas Palacios (Massachusetts Institute Technology, USA)
• Eric Pop (Stanford University, USA)
• Geoffrey Pourtois (imec, Belgium)
• Sayeef Salahuddin (University of California, Berkeley, USA)
• Abu Sebastian (IBM Research–Zurich, Switzerland)
• Dina Triyoso (Tokyo Electron Limited, USA)
• Dries Van Thourhout (Ghent University, Belgium)
• Wei-Yen Woon (National Central University, Taiwan)
• Gong Xiao (National University of Singapore, Singapore)
• Huili Grace Xing (Cornell University, USA)