Magic State Distillation Overhead

Short answer: Magic states are special resource states that fault-tolerant quantum computers consume to run non-Clifford gates such as the T gate. Preparing them cleanly, called distillation, is often called a bottleneck for scalable computing. Papers from 2019 and 2025 report two kinds of progress: protocols whose overhead stays constant as the target error shrinks, and circuit designs that make distillation much cheaper in space and time. These are theory results for codes and circuits.

Why magic states are needed

A 2019 paper in Quantum explains the setting. Quantum error correction is expected to be essential for a large-scale machine, with two-dimensional topological codes such as surface codes the most promising. With these, preparing logical Pauli eigenstates and measuring logical Pauli products, for example by lattice surgery, are fault-tolerant, which is enough for logical Clifford gates. Logical non-Clifford gates such as T cannot be executed directly. Instead they use a resource state that is consumed to run the gate, and for T gates that state is a magic state. A 2025 Nature Physics paper adds that most schemes for realistic quantum computing need magic states for universality, and since preparation may be noisy, distillation is needed to improve their accuracy and suppress errors.

The overhead problem

The 2025 paper defines the cost as the overhead, the ratio of noisy input magic states to cleaner outputs. It is known to scale as O(log^gamma(1/epsilon)) as the output error rate epsilon goes to zero, where gamma is a constant, and it says reducing gamma is highly desirable to remove the bottleneck. A 2019 Physical Review Letters paper recalls that Hastings and Haah showed in 2018 a distillation routine with sublogarithmic overhead, achieving gamma of about 0.6779 and refuting a conjecture that gamma is at least 1. They asked whether gamma can be made arbitrarily close to 0, and the 2019 authors answer yes for routines using qudits of prime dimension.

Constant overhead in 2025

The 2025 paper resolves the question for qubits, demonstrating protocols with the optimal exponent gamma equal to 0, that is, magic state distillation with constant overhead. The authors show this is achievable for the most important magic states such as the T state and the CCZ state. They build the first asymptotically good quantum codes with transversal non-Clifford gates using algebraic geometry constructions, and give an efficient decoder with linear decoding radius.

Making distillation cheaper in practice

The 2019 Quantum paper argues distillation need not dominate the resource cost of fault-tolerant machines. By writing distillation circuits in a form that separates qubits capable of error detection from those that are not, most logical qubits used in distillation can be encoded at a very low code distance. This reduces the space-time cost and the number of qubits. In extreme cases, the authors say, distilling a magic state can cost less than performing a logical Clifford gate on full-distance logical qubits.

What this does not tell you

Constant overhead is an asymptotic statement about code families as the target error goes to zero, and does not give the actual number of qubits for a given machine. The cheaper circuits are design results, not an operating factory.

FAQ

What is a magic state?

A resource state consumed to perform a non-Clifford gate such as T.

Why distill them?

Noisy preparation needs improving, and distillation suppresses errors.

What is the overhead?

The ratio of noisy input magic states to clean outputs.

What did the 2025 paper show?

Protocols with constant overhead, gamma equal to 0, for qubit magic states including T and CCZ.

Is it solved in practice?

No. These are asymptotic and design results, and the 2019 Quantum paper only shows costs can be much lower than assumed.