Just days after unveiling a quantum computer in Chicago powered by its 120-qubit Nighthawk r2 processor, IBM on Wednesday outlined Nighthawk r3, a planned successor with 520 programmable qubits that it aims to connect with a second chip in a 1,040-qubit quantum computer in 2027.

The expansion would more than quadruple the number of programmable qubits per processor. IBM is designing r3 to run circuits with 10,000 quantum operations, compared with more than 7,500 demonstrated on r2. Each r3 chip would contain 2,034 physical qubits in all.

Making the chips work together will require 10 L-couplers per r3 processor, IBM-developed superconducting links that transfer quantum information between chips. In October, IBM used four L-couplers to connect two Nighthawk r1 processors, reporting 99% average state-transfer fidelity. The company aims to improve that figure to 99.9%.

IBM is also preparing the cooling infrastructure needed for linked processors. In August, it tested two custom cryogenic modules joined into a shared cooling environment. They reached below 15 millikelvin, roughly −459.6°F, less than 0.03°F above absolute zero. Each module’s larger enclosure provides up to 12 times as much wiring space as IBM’s other quantum systems, making room for the connections needed to control and link more processors.

For researchers, last week’s Chicago installation is a working predecessor to the proposed 2027 machine. IBM unveiled the Nighthawk r2-powered Quantum System Two at the University of Chicago’s Hyde Park Labs on Nov. 6, offering hardware for research into drug discovery, power grid optimization and high-energy physics. The r2 processor delivers up to 100,000 measurement samples per second, 25 times the throughput of IBM’s earlier Heron processor, according to the company.

IBM says partners including E.ON, Algorithmiq and Q-CTRL have completed more than 34 billion quantum circuit executions on r2. E.ON reported a 15-fold speedup over Heron for an electric grid machine learning workload, while an experiment with Algorithmiq executed quantum circuits eight times faster than on Heron.

IBM is also adding classical high performance computing resources to its Quantum Platform. Researchers can now connect quantum workloads to their own computers or supercomputers, or use classical computing resources available directly through the Platform. IBM says Cleveland Clinic and RIKEN used the expanded platform to automatically process 608 million quantum measurement samples, completing the work 33 times faster than manual processing.

Separately, on Nov. 5, IBM, Oak Ridge National Laboratory and AMD announced a research platform at IBM’s Poughkeepsie facility pairing Nighthawk r2 with ORNL’s Umbra HPC system. IBM says the co-located hardware reduces delays between quantum measurements and the classical calculations used to identify errors and guide further operations.

That integration of quantum and classical computing will be demonstrated at SC26 in Chicago next week. IBM and University of Illinois Urbana-Champaign researchers, working with the National Center for Supercomputing Applications, plan to study electronic conductivity in quantum materials using IBM Quantum System Two and NCSA’s GPU-powered DeltaAI supercomputer.