Subscribe
0%
Illustration for: IBM's Nighthawk r3 targets 10,000 operations on 520 qubits, with a 1,040-qubit linked system in 2027. What the numbers mean

IBM's Nighthawk r3 targets 10,000 operations on 520 qubits, with a 1,040-qubit linked system in 2027. What the numbers mean

IBM previewed its next quantum processor, which is designed to run deeper circuits on four times as many programmable qubits as today's Nighthawk r2, and plans to link two of them in 2027. Most figures are IBM design targets or self-reported speedups against its own older chips.

Listen to this article0:00 / 4:15

IBM on November 11 previewed Quantum Nighthawk r3, a processor it says is designed to execute 10,000 quantum operations across 520 programmable qubits, built from 2,034 physical qubits in total. IBM also said it plans to join two r3 chips into a single 1,040-qubit system in 2027.

Why it matters: the useful size of a quantum computation depends on two things at once, how many qubits you can program and how many operations (gates) you can run before noise swamps the answer. IBM is pushing both numbers up on the same chip family and starting to link chips together, which is the route its roadmap relies on to reach larger, error-corrected machines. But r3 isn't available yet, and the headline figures are targets.

r2 today vs r3 next

Nighthawk r2 Nighthawk r3
Status Deployed in September, per Constellation Research Announced; IBM gave no availability date
Programmable qubits 120 520 (2,034 physical)
Operations per circuit More than 7,500 Designed for 10,000
Chip-to-chip links — 10 "l-couplers"
Speed Up to 100,000 samples per second Not stated

According to IBM's announcement, r3 keeps r2's square-lattice layout and fast operations and adds the l-couplers that will be used to connect processors.

What IBM says r2 has already done

IBM says clients and partners including E.ON, Algorithmiq and Q-CTRL have run 34 billion circuits on Nighthawk r2, and reports speedups of up to 19x over earlier IBM processors. Examples it gives:

  • E.ON: a 15x speedup in quantum machine learning for energy-grid optimization.
  • IBM and Oak Ridge National Laboratory: a 12x speedup on a magnetic-materials simulation, compared with IBM's older Heron chip, per The Quantum Insider.
  • Algorithmiq: an experiment running 8x faster than its original Heron version. IBM estimates the new circuits would need about 240x more memory for leading classical tensor-network methods to match.
  • RIKEN: an IBM–RIKEN sample-based quantum diagonalization workflow ran about 5x faster while doing 100x more computation than the original study.
  • Cleveland Clinic and RIKEN: 608 million samples processed 33x faster than without workflow automation.

Linking chips: progress and a gap

IBM says it connected Nighthawk r1 processors with four l-couplers and achieved 99.3% fidelity in transferring quantum states between them, short of its 99.9% target. It also connected two cryogenic modules into one system for the first time; IBM says they reached operating temperature in under five days and cooled below 15 millikelvin.

Separately, IBM, AMD and Oak Ridge's Quantum Science Center are pairing Nighthawk processors with an ORNL system built on AMD Instinct MI355X GPUs and 5th Gen EPYC CPUs as a co-design platform for quantum error correction, Constellation Research reports.

IBM Research director Jay Gambetta framed the pitch around accuracy and speed, saying useful quantum computing depends on both.

How to read the claims

  • Targets, not results. The 10,000-operation figure for r3 and the 2027 1,040-qubit system are design goals.
  • Self-reported benchmarks. Nearly every speedup comes from IBM or its partners, many are "up to" figures, and most compare against IBM's own Heron chip rather than other vendors' hardware or the best classical methods.
  • Interconnect isn't there yet. 99.3% state-transfer fidelity versus a 99.9% goal is the number to watch, because linking chips is how IBM plans to scale.
  • No advantage claim here. The announcement reports scale and speed. IBM points to the community-run Quantum Advantage Tracker for openly published circuits, but The Quantum Insider notes those July experiments ran on Heron, not Nighthawk.

For researchers and companies already on IBM's cloud, r2 is what you can use now; r3 is the chip to plan around. Constellation Research notes IBM is building its stack toward the Starling system it targets for 2029, while rivals such as Quantinuum and IonQ scale their own approaches.

Sources

  1. IBM Newsroom — IBM Quantum Nighthawk Demonstrates New Levels of Speed and Quality, with Plans to Scale Beyond 1,000 Qubits in 2027 newsroom.ibm.com
  2. The Quantum Insider — IBM Plans 1,040-Qubit Quantum System with Next-Generation Nighthawk Processors thequantuminsider.com
  3. Constellation Research — IBM outlines Quantum Nighthawk r3 processor, quantum roadmap constellationr.com

#IBM#quantum computing#Nighthawk#qubits#Oak Ridge#AMD#RIKEN