Researchers have introduced the concept of a thermodynamic computer—a new type of computing architecture that could potentially reduce the energy consumption of artificial intelligence systems dramatically. This was reported by Quantum Insider.
A team from Extropic and the Massachusetts Institute of Technology believes their approach could make certain AI tasks up to 10,000 times more energy-efficient compared to traditional computing.
Using Noise in Computations Instead of Fighting It
Modern processors, including GPUs used for training and running large language models, consume significant resources to perform precise deterministic computations. Physical noise and thermal fluctuations are typically seen as interference that must be minimized.
The authors propose an opposite approach. Instead of suppressing random thermal processes, they suggest using them as part of computations. This principle is called Thermodynamic Computing.
According to the researchers, many AI tasks—such as finding the most probable answer or optimal solution—are inherently probabilistic. Therefore, a computing system utilizing random physical processes could potentially perform these tasks much more efficiently than classical processors.
Addressing a Major AI Challenge
Interest in such architectures is driven by the growing energy consumption of the artificial intelligence industry. Major tech companies are investing billions of dollars in building data centers, and the demand for electricity to train and operate modern models continues to rise rapidly.
If the proposed architecture proves viable, it could reduce not only energy costs but also the operational expenses of AI infrastructure by decreasing the need for expensive computing clusters.
Practical Application Still a Long Way Off
Currently, this is a fundamental research project rather than a ready-to-use processor. The authors have presented the architecture and simulation results showing the advantages of the new approach for specific task classes. It may take years before commercial chips based on thermodynamic computing principles become available.
Nevertheless, the work reflects the industry’s growing interest in alternative computing architectures. As AI models continue to scale, there is increasing focus not only on their capabilities but also on the cost of computations. In this context, finding ways to drastically cut energy consumption is becoming a key research direction alongside the development of quantum and neuromorphic computers.
In May, Amazon implemented a new data center network architecture that accelerates data transfer and reduces energy consumption.
