IBM maps out Nighthawk r3: 520 programmable qubits and 10,000 operations, with a plan to link two chips past 1,000 qubits in 2027

The backstory: why a quantum computer has to be both fast and accurate

A quantum computer's basic unit is the qubit. Unlike a regular bit, which is either 0 or 1, a qubit can occupy a more complex state, which is useful for certain kinds of problems. Qubits are also fragile. The longer a calculation runs and the more operations it uses, the more errors tend to pile up. So the field is chasing scale and quality, and speed as well.

Speed matters because a quantum computer often answers by sampling: the same circuit has to be run many times before the result is clear. Demanding applications may need millions to billions of samples. If the machine is too slow, a calculation that works in principle can still take too long, or cost too much, to be useful.

What IBM announced

On November 11, IBM laid out design targets for its next processor, Nighthawk r3: 10,000 accurate quantum operations across 520 programmable qubits, with 2,034 physical qubits on the chip. Programmable qubits are the ones used to write and run algorithms; physical qubits include the broader set of hardware units that support the computation, and there are usually more of them. The r3 design also calls for 10 chip-to-chip links, called l-couplers, to connect processors into larger systems.

This needs to be stated plainly: r3 is a processor that "will be built." The release did not give an availability date. IBM also said it plans, in 2027, to link two r3 processors into a 1,040-qubit system. That is a roadmap, not a machine already delivered.

What is in use now is Nighthawk r2, introduced in September. IBM says r2 can deliver up to 100,000 samples per second and support computations with more than 7,500 quantum operations on 120 qubits. Clients and partners including the energy company E.ON, Algorithmiq, and Q-CTRL have, according to IBM, run more than 34 billion circuits on r2 and reported performance gains of up to 19 times over previous IBM quantum processors.

Named examples are also IBM or partner figures. E.ON reported a 15 times speedup on r2 for quantum machine learning used in energy-grid optimization, compared with earlier work on IBM's Heron processor. IBM said it moved a magnetic-materials simulation, first run this year with Oak Ridge National Laboratory, onto r2 and saw a 12 times speedup, with results still consistent with neutron-scattering experiments. An Algorithmiq experiment on a heterogeneous quantum material ran about 8 times faster on r2 than on Heron, and IBM estimated that matching that accuracy with leading classical tensor-network methods would take about 240 times more memory.

How far the wiring has actually been tested

Linking two large chips into one machine first requires a working bridge. IBM said it has already connected Nighthawk r1 processors this year with four l-couplers, reaching 99.3% fidelity when transferring a quantum state, short of its 99.9% target. Fidelity, here, is how closely the state still matches the original after it is sent from one chip to the other. IBM also said it connected two cryogenic modules into a single system for the first time, reached operating temperature in under five days, and cooled below 15 millikelvin — a fraction of a degree above absolute zero, the kind of environment quantum chips typically need.

Jay Gambetta, director of IBM Research, said useful quantum computing depends on delivering results with both accuracy and speed, and that Nighthawk is already showing what that looks like. As the architecture scales and is integrated more tightly with classical computing, he said, the capability will rise again. That is the company's own reading of its path.

What it means for you

In the near term, this does not turn a phone or a home computer into a quantum machine. It is specialized equipment for research groups and for teams in areas such as energy and materials, and r3 is still a plan.

If IBM later builds r3 to the design it has published, and can reliably join two of the chips in 2027, quantum computers would move closer to handling larger problems in a practical amount of time. For now it is better to keep two facts separate: partners are already reporting speedups on r2, and the 520 programmable qubits and the 1,000-plus-qubit system are still targets that have to be met.

Original source published: IBM Newsroom, Nov 11, 2026, 20:00 Beijing time (UTC+8)