If this breakthrough can be successfully expanded, it will significantly reduce the hardware scale required to build fault-tolerant quantum systems, which is of great significance for the commercialization of quantum computing.

Written by: Yang Chen, Wall Street Journal

The Google Quantum AI team has made a key advance in the field of quantum computing, promising to greatly reduce the hardware costs required to build practical quantum systems.

The relevant research was published in the journal Science on Thursday. Google has validated a technique called "magic state cultivation" on its Willow processor, reducing the quantum error rate by about 40 times, with T state logic fidelity reaching 99.99%. This is the highest record in the field to date.

Magic state production has long been seen as a core bottleneck for the scalability of quantum computing—estimates suggest that as much as 95% of a quantum computer's computational power may need to be dedicated to generating magic states, leaving very few resources available for actual computation.

If this breakthrough can be successfully expanded, it will significantly compress the hardware scale required to build fault-tolerant quantum systems, which is of great importance for the path to commercialization of quantum computing.

Google also stated that the feasibility of this technology still needs to be verified on a larger scale, and there is still a considerable distance to practical quantum computing.

As of the time of writing, Google's stock price rose 0.24% on Thursday.

Magic State: The Resource Black Hole of Quantum Computing

Magic states are a key element for achieving "universal" quantum computing. For a quantum processor to perform complex operations that surpass classical computers, it must rely on magic states to generate the required set of quantum operations. However, once a magic state is used, it cannot be reused, making it a scarce resource.

The greater challenge lies in production costs. Some research estimates suggest that quantum computers may need to allocate up to 95% of total computational power for magic state production, leaving very few resources for actual computational tasks.

This bottleneck has been foreseen by researchers since 2004, and over the past two decades, academia has been continuously exploring ways to reduce magic state production costs, but progress has been slow.

Proposing a Magic State Cultivation Scheme, Significantly Improving Resource Efficiency

In 2024, Google Quantum AI researchers Craig Gidney, Noah Shutty, and Cody Jones proposed a magic state cultivation scheme based on existing literature.

The core idea is to design the magic state production process as a set of filtration mechanisms: injecting noisy quantum states into a small area on the chip, and then using a series of highly coordinated measurement operations to detect and eliminate defective attempts, thereby "cultivating" high purity, low error rate quantum states and concurrently encoding them into a larger usable form.

However, the transition from theory to hardware implementation was not smooth. Research scientist Emma Rosenfeld consistently obtained data in initial tests that "seemed to violate the laws of physics."

After months of troubleshooting, the team found that the root of the problem lay in a false assumption in the benchmark technology—noise from drifts in electrical equipment and temperature changes in the laboratory interfered with the validity of the benchmark tests.

To address this, the team not only added additional controls to suppress noise impacts, but also invented a new benchmarking technique using the cultivation process itself as a self-calibration tool.

Fidelity and Errors Both Set New Records

On the Willow processor, the aforementioned scheme ultimately achieved about a 40-fold reduction in error rates compared to uncultivated quantum states, with T state logic fidelity reaching 99.99%, resetting the records in the field.

The Google team also completed another significant demonstration: verifying that the cultivated magic state could be smoothly transferred between two different quantum error correction schemes—specifically, transplanting the quantum state from distance -3 color code (a newer error correction method suitable for cultivation operations) to distance -5 surface code (a more mature error correction method).

This "grafting" operation proved that the purified quantum states could stably survive and transfer within the hardware, paving the way for the integration of magic state cultivation technology into Google's existing quantum computing architecture.

Despite these advancements, Google clearly stated that the magic state cultivation technology still needs to be validated on a larger scale before it can be truly applied to practical quantum computing tasks, and many technical challenges still remain.