Chinese scientists develop world's first carbon-based AI chip

Express Tribune

08-03-2025

Listen to article
In a groundbreaking advancement that could change the future of computing, Chinese scientists have developed the world's first carbon-based microchip capable of running artificial intelligence (AI) tasks using a revolutionary ternary logic system.
This innovation, spearheaded by researchers from Peking University and Beijing University of Posts and Telecommunications, marks a significant leap forward in semiconductor technology, potentially surpassing the limitations of traditional silicon-based chips.
The newly developed chip utilises carbon nanotubes (CNTs), a material that offers exceptional mechanical and electrical properties.
CNTs are tiny cylindrical tubes made from graphene sheets, and they have been used primarily as conductive additives in lithium-ion batteries.
However, due to their superior electrical conductivity, excellent stability, and ultra-thin structure, CNTs are now seen as a promising material for the next generation of semiconductors.
This breakthrough differs from conventional silicon chips in that it employs ternary logic instead of the traditional binary system, which uses only zeros and ones.
The new chip processes data not just in ones and zeros, but also in a third state, which enables faster computations while using less energy.
This ternary logic system improves the efficiency of data transmission within the same physical space, allowing for quicker and more energy-efficient computing.
The team of researchers designed a novel carbon nanotube transistor using a concept known as source-gated transistors (SGTs).
By adjusting the gate voltage, the CNT transistor can switch between three distinct current states, thus creating the foundation for ternary logic circuits.
This new design promises to overcome the limitations of current chip technologies, particularly in terms of power consumption and processing speed.
To test the capabilities of their new chip, the researchers built a neural network capable of learning and reasoning by mimicking the connections between neurons in the human brain.
Extensive experiments revealed that the CNT-based neural network achieved perfect accuracy in classifying handwritten digits, demonstrating its vast potential for AI applications, including image recognition and machine learning tasks.
One of the lead researchers, Peng Lianmao, a member of the Chinese Academy of Sciences, has been studying carbon-based chip technology for over two decades.
His team has made remarkable progress in developing high-performance CNTs and achieving precise control over nanotube arrays.
In 2020, they fabricated an eight-inch CNT wafer that outperformed similar silicon-based devices in integrated circuit performance.
This achievement cemented China's position at the forefront of global research in carbon-based semiconductor technology.
The new chip is not only highly efficient but also offers high stability and strong resistance to interference, making it ideal for use in high-performance computing, machine learning, AI, and low-power storage devices.
It also has applications in Internet of Things (IoT) devices, where energy efficiency is a crucial factor.
Despite the many advantages, carbon nanotube chips still lag behind traditional silicon chips in terms of integration density. For example, Nvidia's RTX 5090 GPU, which was announced in January 2025, contains a staggering 92 billion transistors, far exceeding the current capabilities of CNT technology.
However, the development of carbon-based chips is viewed as the next frontier in semiconductor technology, and China is currently leading the way in this race.
Peng Lianmao expressed his optimism about the future of carbon nanotube-based chips, stating that the ultimate goal is to make them mainstream within the next 10 to 15 years.
If successful, this technology could replace silicon-based chips in a wide range of applications, from supercomputers and data centres to smartphones and other electronic devices.
This transition would mark a significant shift in the semiconductor industry, paving the way for a new era of high-performance, low-power computing solutions.