Ibm: first fault-tolerant quantum computing system operational in 2029

IBM's approach is different from that followed by Google and AWS, and also from that of Microsoft. Google mainly uses surface code, which is very robust but requires a lot of physical resources, in particular many qubits to encode a single logical qubit. Willow is a quantum chip presented by Google in December 2024 that performed a test calculation in less than five minutes that a classical supercomputer would take billions of years to solve. However, it does not yet solve real practical problems (for now, it is only a theoretical test). The big step forward was reducing errors in calculations, a crucial obstacle in quantum computation.

AWS, with the Ocelot chip, has also addressed the problem of quantum error correction, promising a 90 per cent saving on overhead, but it is still in the experimental phase, focusing mainly on scalable architectures and reducing the physical complexity of circuits.

Finally, Microsoft is pursuing a very high-potential path, focusing on the physical discovery of new types of qubits in order to make quantum computing truly feasible. Microsoft's approach with the 'Majorana' project stands out in the quantum computing landscape because it focuses on a type of topological qubit, the so-called Majorana qubits, which are theoretically considered to be more stable and less error-prone than conventional qubits. Scientifically it is the most ambitious project, but technologically it lags behind the others.

The difference between IBM Quantum Starling and the others concerns both the objective of the projects and the type of architecture and technological advancement. "Ours is a hardware-aware approach," explained Chow, "BB codes can achieve error correction performance comparable or better than surface code using far fewer physical qubits.

On June 10, 2025, IBM unveiled its path to build IBM Quantum Starling, depicted in

The key milestone, the IBM researcher confirms, is set for 2029, but it will not be a closed process: the technological steps will be out in the open with a clear goal, a fault-tolerant quantum computer capable of performing 100 million logical operations on 200 logical qubits. To understand the impact, one only has to think that such a system could tackle problems that are unsolvable with current supercomputers, for instance in the simulation of complex chemical reactions or the design of new materials. The plan is based on a new roadmap published by IBM and three new processors. Quantum Loon (2025) will be used to test the components of the architecture, in particular the 'c-couplers', connectors that enable long-distance interaction between qubits on a single chip. Quantum Kookaburra (2026) will be the first modular processor capable of combining quantum memory and computational logic, with the goal of scalable fault tolerance. Quantum Cockatoo (2027), on the other hand, will connect two Kookaburra modules via 'L-coupler', a solution designed to avoid monolithic chips and create a distributed architecture. The culmination of this will be Starling in 2029.

But the real leap in scale will come beyond 2033, when IBM plans to realise architectures with 1 billion logic gates operating on 2000 logic qubits. In practice, this means making so-called quantum utility computing operational: quantum machines that are scalable, stable and can be integrated with classical HPC systems.

Underpinning this vision are innovations such as modular processors (including Starling and Blue Jay) and new low-overhead error correction techniques, such as those based on LDPC codes, which can achieve results comparable to surface codes with one-tenth of the physical qubits. This is no longer just a theoretical leap, but an engineering journey.