IBM addresses a new approach to the correction of quantum errors

IBM has presented a new quantum computer architecture that says it will reduce the amount of qubits required for errors correction. The advance will support its goal of building a large -scale quantum computer and failure tolerant, called Starling, which will be available for customers by 2029.

Due to the lack of inherent reliability of the qubits (the quantum equivalent of the bits) of which quantum computers are built, the correction of errors will be crucial to build reliable devices on a large scale. Errors correction approaches spread each information unit in many physical qubits to create “logical qubits.” This provides redundancy against errors in individual physical qubits.

One of the most popular approaches is known as a surface code, which requires approximately 1,000 physical qubits to form a logical qbit. This was the approach in which IBM initially focused, but the company finally realized that creating hardware to support it was an “impossible dream of engineering,” Jay GambettaIBM Quantum vice president said at a press conference.

Around 2019, the company began investigating alternatives. In paper Posted in Nature Last year, IBM researchers described a new error correction scheme called Low Quantum Density Parity Verification Codes (QLDPC) that would require approximately one tenth of the number of qubits needed by surface codes. Now, the company has presented a new quantum protection architecture That can realize this new approach.

“We have deciphered the code to the correction of quantum errors and it is our plan to build the first quantum tolerant to large -scale failures,” said Gambetta, who is also a member of IBM. “We are sure that it is now a matter of engineering to build these machines, instead of science.”

IBM presents a new quantum roadmap

IBM will take the first step to realize this architecture at the end of this year with a processor called Loon. This chip will have couplers that can connect distant qubits in the same chip, which is key to implementing QLDPC codes. These “non -local” interactions are those that make the approach more efficient than the surface code, which is based solely on the qubits that communicate with their neighbors.

According A roadmap Run together with the details of the new architecture, the company plans to build a follow -up processor called Kookaburra in 2026 that will have a logical processing unit and a quantum memory. This will be the first demonstration of the type of base module of which subsequent systems will be built. The following year, IBM plans to join two of these modules to create a device called Cockatooo.

The road map does not detail how many modules will be used to create Starling, IBM’s planned commercial offer, but the computer will have 200 logical qubits and will be able to execute 100 million quantum operations. Said exactly how many physical qubits they will be required, he said Matthias SteffenIBM member, which leads the quantum processor technology team. But it is likely that the new architecture requires the order of several hundred physical qubits to create 10 logical qubits, he added.

IBM plans to build Starling by 2028, before making it available in the cloud the following year. It will be hosted in a new quantum data center in Poughkeepsie, NY, and will lay the basis for the final system on the current IBM road map, a machine removed logical logic codenada Blue Jay.

IBM’s new architecture is a significant advance about its previous technology, he says Mark HormvathGartner analyst vice president, who was informed before the announcement. The increase in the connectivity of the new chip makes it substantially more powerful and is supported by significant advances in 3D manufacturing. And if it helps IBM to reach 200 logical qubits, that would take quantum computers to the scope of solving practical problems, says Hurth.

However, Horvath adds that IBM’s modular approach is deposited to get there could be challenging. “That is a very complicated task,” he says. “I think it will eventually work. It’s fair, it’s much further than people think they are.”

One of the most remaining obstacles is to improve door fidelities throughout the device. To successfully implement this new architecture, error rates must decrease by an order of magnitude, Steffen of IBM admitted, although the company is confident that this can be achieved. One of the main roads will be to improve the coherence times of the underlying qubits, which refers to how long their quantum status can maintain. “We have evidence that this is really one of the main bottlenecks to improve door errors,” says Steffen.

In isolated test devices, IBM has managed to take the average coherence times to 2 milliseconds, but translating that to larger chips is not simple. Steffen said the company recently advanced with its Heron chips, going around 150 to 250 microseconds.

Significant engineering challenges also remain in support of infrastructure, said Steffen, including connectors that unite different parts of the system and amplifiers. But a great advantage of the new architecture is that it requires much less components due to the reduced number of physical qubits. “This is one of the reasons why we are so excited about these QLDPC codes, because it also reduces all non -substance processors,” he says.

This story was updated on June 10, 2025 to correct some details of the current IBM roadmap.