Reversible Computing Escapes the Lab in 2025
Michael Frank, an academic researcher with over three decades of experience in computer engineering, believes the time has come to commercialize reversible computing. In July 2024, he transitioned from his role as a senior engineering scientist at Sandia National Laboratories to join Vaire Computing, a U.S. and U.K.-based startup.
Energy Efficiency Challenges
Frank emphasizes the urgency for unconventional approaches in computing due to the plateauing energy efficiency of conventional digital logic. According to an IEEE semiconducting industry road map report he helped edit, energy efficiency in the industry is set to stagnate, demanding novel solutions like reversible computing.
Reversible computing, a concept rooted in information theory, offers significant potential for energy savings. It promises up to a 4,000x energy-efficiency gain compared to traditional methods, according to Frank’s research at Sandia.
Vaire Computing’s Endeavors
Vaire’s initial prototype, scheduled for fabrication in the first quarter of 2025, aims to produce a chip that recovers energy in an arithmetic circuit. The company plans to introduce an energy-saving processor for AI inference by 2027. The full 4,000x improvement in energy efficiency is anticipated over the next 10 to 15 years.
Academic Foundations and Practical Challenges
The theoretical basis for reversible computing dates back to 1961 when IBM’s Rolf Landauer discovered the energy cost associated with erasing information. Charles Bennett, Landauer’s successor at IBM, addressed practicality issues by introducing the concept of “decomputing” to manage memory storage effectively.
Reversible computing has been explored since the 1990s, notably by students under MIT’s Thomas Knight. However, practical challenges remained, particularly in reducing power usage in external circuits—a problem Vaire is tackling.
Innovations in CMOS Technology
Reversible computing with CMOS technology involves innovative methods to recover and recycle signal energy. By embedding the circuit in a resonator, Vaire aims to reduce heat loss and enhance energy efficiency. The introduction of adiabatic transistor switching minimizes unnecessary heat generation, setting the stage for practical applications.
Future Prospects
Vaire is developing a reversible adder embedded in an LC resonator, with plans to expand into machine learning applications. The company is also exploring microelectromechanical systems (MEMS) resonators, offering higher quality factors and reduced energy loss.
Despite challenges in custom manufacturing and integration, the potential for significant energy efficiency gains drives Vaire’s efforts. Over the next decade, the company aims to achieve a 4,000x improvement in performance, contingent on advances in resonator technology.