IBM unveils Nighthawk r3, eyes 1,000+ qubits by 2027

IBM on Wednesday announced Nighthawk r3, the next processor in its Nighthawk line. IBM says the new chip will run 10,000 quantum operations across 520 programmable qubits — 2,034 physical qubits in total — accurately enough to tackle larger, more complex problems. IBM plans to connect two Nighthawk r3 chips into a single system of more than 1,000 qubits in 2027.

The new chip builds on Nighthawk r2, introduced in September and available today. Nighthawk r2 handles more than 7,500 quantum operations across 120 qubits, delivering up to 100,000 samples per second. Nighthawk r3 adds 10 links between chips, a feature IBM calls l-couplers, meant to join processors into bigger systems down the line.

More than 34 billion circuits have already run on Nighthawk r2 across partners such as E.ON, Algorithmiq and Q-CTRL, who report speed gains as high as 19x against what IBM’s older hardware could do. E.ON reported a 15-times speedup in quantum machine learning applied to energy-grid optimization, while IBM said a 12-times speedup let it and Oak Ridge National Laboratory simulate real magnetic materials with results consistent with neutron-scattering experiments. An extended experiment with Algorithmiq ran eight times faster than the original version on IBM’s earlier Heron processor, producing circuits IBM estimates would need about 240 times more memory for a classical computer to simulate at comparable accuracy. IBM also says it has demonstrated the pieces needed to link processors together: connecting Nighthawk r1 chips with four l-couplers reached 99.3% fidelity transferring a quantum state between them, short of its 99.9% target, and the company separately linked two cryogenic modules into one system — the system reached its operating temperature in less than five days, cooling to below 15 millikelvin.

IBM Quantum Platform now combines quantum processors with classical high-performance computing in a single managed environment. Working with RIKEN, IBM says it ran the same sample-based quantum diagonalization workflow about five times faster, while handling roughly 100 times more computations than in its original study, and Cleveland Clinic and RIKEN used the platform to process 608 million samples for modeling biological structures 33 times faster than without the automation. “Useful quantum computing hinges on whether results can be delivered with accuracy and speed, both are fundamental,” Jay Gambetta said. He is IBM Research’s director and holds the title of IBM Fellow.

Why it matters

IBM’s bet is that linking many modest quantum chips together, rather than building one giant processor, is the practical path to machines large enough to tackle problems classical supercomputers can’t — and the speedups partners are already reporting on live workloads like grid optimization and materials simulation are the evidence IBM is pointing to.