A coupled-oscillator-based Ising chip for combinatorial optimization

Numerous practical problems—ranging from machine learning to bioinformatics—can be formulated as combinatorial optimization problems. However, the computational resources required to find an optimal solution to these problems using conventional von Neumann computers increases rapidly with problem size. Alternative solvers based on Ising machines, which directly leverage equilibrium characteristics of physical systems, offer a potential solution for such problems. Here we report a coupled-oscillator-based all-to-all-connected Ising chip that is manufactured in 65-nm complementary metal–oxide–semiconductor (CMOS) technology and operates at room temperature. The approach relies on logic-based coupling, which leads to low power consumption and a large number of all-to-all-connected spins. We show that the chip can solve representative combinatorial optimization problems in a more time- and energy-efficient manner than can optimized classical solvers in software and emerging quantum annealers. Due to its energy efficiency and all-to-all connectivity, our chip can also efficiently solve dense combinatorial optimization problems that cannot be effectively mapped or solved by quantum annealers.