IBM to build first large-scale error-correcting quantum computer by 2029

The American corporation IBM announced plans to launch the world's first large — scale quantum computer with error correction-a system called IBM Quantum Starling. The project is expected to be completed by 2029. The computer will be located in the company's new data center in Poughkeepsie, New York.

According to IBM, the performance of Starling will be about 20,000 times higher than that of existing quantum systems. To describe its state, it would require a computational space exceeding the capabilities of 10 ^ 48 modern supercomputers.

Starling will be the foundation for IBM's Blue Jay platform, which can handle up to 1 billion operations using 2,000 logical qubits. Each logical qubit will be formed from several physical ones — this architecture will significantly reduce the error rate, which is especially important for the practical application of quantum computing in pharmaceuticals, materials science and other high-tech industries.

The central element of the system will be the use of new error correction codes-qLDPC. This technology reduces the required number of physical qubits by 90% compared to classical approaches. Two IBM research studies confirm that these codes provide system scalability and the ability to process data in real time.

The company also presented a roadmap for the phased development of a quantum architecture:

IBM Quantum Loon, a platform for testing long — distance connections on a chip, is scheduled to be released in 2025.;

In 2026, the IBM Quantum Kookaburra modular processor will appear, combining quantum memory and computing;

In 2027, IBM Quantum Cockatoo is scheduled to launch-a bundle of modules for building a distributed architecture without the need to increase the size of individual chips.

By 2029, IBM expects to reach the level of 100 million operations with 200 logical qubits. This, according to the developers, will be a step towards the practical use of quantum technologies in problems that cannot be solved using classical computing systems.