IBM Targets 1,040-Qubit System in 2027 as Nighthawk r3 Design Unveiled

By Grace Kim, Chip Dispatch

IBM has unveiled the design of Nighthawk r3, a quantum processor designed to run 10,000 quantum operations accurately across 520 programmable qubits, drawn from 2,034 physical qubits. The company plans to link two r3 processors into a single 1,040-qubit system in 2027.

IBM announced the design on Wednesday. Each r3 chip keeps the square-lattice connectivity of the current generation and adds 10 chip-to-chip interconnects that IBM calls l-couplers.

How does IBM plan to scale beyond one chip?

The l-couplers are the scaling mechanism. Earlier this year IBM linked Nighthawk r1 processors with four l-couplers and reached 99.3% quantum state transfer fidelity, short of a 99.9% target. In the same experiment, IBM joined two cryogenic modules into one system for the first time. The combined system reached operating temperature in under five days and cooled below 15 millikelvin.

The 2027 plan depends on the couplers working across a pair of chips. IBM has not published fidelity targets for the 10-coupler r3 configuration.

What is the current chip delivering now?

Nighthawk r2, introduced in September, runs more than 7,500 operations on 120 qubits. It produces up to 100,000 samples per second using fast qubit reset. Clients and partners have executed 34 billion circuits on the chip. IBM reports speedups of up to 19x over its earlier processors.

Partner figures, released alongside the announcement, point to application-level gains:

• E.ON reported a 15x speedup in quantum machine learning for grid optimization.
• IBM and Oak Ridge National Laboratory reported a 12x speedup on a magnetic materials simulation.
• Algorithmiq said its quantum materials experiment ran 8x faster than the original Heron run, with circuits it estimates would need 240x more memory for leading classical tensor-network methods to match.

All three results were measured against IBM's Heron chip. These are company and partner figures, not independent benchmarks.

Jay Gambetta, director of IBM Research, framed the tradeoff directly. "Useful quantum computing hinges on whether results can be delivered with accuracy and speed, both are fundamental," he said.

Where does Nighthawk sit in IBM's roadmap?

The r3 disclosure follows two related moves. IBM installed its first Illinois quantum computer, a Nighthawk-powered System Two in Chicago. It also built a co-design platform with Oak Ridge and AMD that pairs Nighthawk with ORNL's Umbra system of AMD Instinct MI355X GPUs and EPYC CPUs to test error correction.

Wednesday's release gives no availability date for r3 as a product. It does not update IBM's fault-tolerance timeline either. That plan, set out in June 2025, targets Starling: a fault-tolerant machine with 200 logical qubits running 100 million operations, located in Poughkeepsie, New York, by 2029.

The gap between the two roadmaps is the commercial question. Nighthawk r3 and its 1,040-qubit dual-chip system serve near-term workloads, while Starling carries the fault-tolerance promise. If the l-couplers hit fidelity targets at scale, the 2027 system becomes the bridge between the two.