Reducing the costs of building quantum computers has implications that extend well beyond the realm of physics. It accelerates the timeline for practical applications such as drug design, the creation of innovative materials, and even the potential to undermine the encryption that secures bank accounts and Bitcoin wallets.
Quantum computing is powered by qubits, which are quantum bits capable of representing a mixture of 0 and 1 simultaneously. This feature allows them to evaluate multiple possibilities concurrently. However, qubits are extremely delicate; minor fluctuations in temperature or electrical noise can disrupt a qubit’s state and compromise computations. This vulnerability underscores the industry’s focus on error correction, which involves distributing individual pieces of information across numerous qubits to identify and rectify errors.
While current error correction methods effectively manage simple tasks, complex operations require a different approach that isn’t safeguarded in the same manner. To execute these operations, the machine must first generate what are known as magic states, functioning like single-use fuel cells: they are consumed immediately when used.
The production of pristine magic states has historically been prohibitively costly. Reports indicate that this can consume up to 95% of a quantum computer’s overall capacity, as noted by Google, leaving little bandwidth for actual computations.
In 2004, physicists Sergei Bravyi and Alexei Kitaev proposed a conventional solution: magic state distillation. This method condenses many imperfect magic states into a handful of high-quality ones. Although effective, it necessitates extensive qubit “factories” dedicated solely to this purpose.
A 2024 proposal from Google researchers introduced a more economical approach, termed magic state cultivation. Unlike a factory, this technique functions like a filter, applying a rough magic state to a small area of the chip. The process involves running a series of checks, discarding any flawed attempts, and continuing the refinement of those that pass. Their research even suggested that further distillation might be unnecessary in practical scenarios.
“Utilizing Google Quantum AI’s Willow processor, our protocol achieved a remarkable 40-fold error reduction compared to uncultivated states. Additionally, we attained T-states with an unprecedented logical fidelity of 99.99%,” Google reported.
Despite this success, only around 8% of attempts passed all checks, meaning the chip discards more than nine out of ten initiatives to maintain just one clean state.
Cultivation plays a role in Gidney’s 2025 estimate, predicting that RSA-2048 encryption could potentially be compromised with under a million noisy qubits—a significant decrease from the 20 million he forecasted in 2019. Prior to this experiment, magic state cultivation had only been explored theoretically or through simulations.
In March, Google further refined its findings with a Quantum AI whitepaper co-authored by Ethereum Foundation researcher Justin Drake and Stanford cryptographer Dan Boneh. The document estimates that breaking Bitcoin’s elliptic curve cryptography would require fewer than 500,000 physical qubits. This follows the company’s Quantum Echoes experiment conducted last October, which marked the first verified quantum performance enhancement using actual hardware.
However, this does not equate to immediate threats to digital wallets. The experiment utilized a maximum of 54 out of Willow’s 105 qubits, and the paper notes that factoring an RSA-2048 key would necessitate magic states with error rates around one in ten billion—approximately a million times cleaner than what Google recently accomplished.
According to Google, the next phase will focus on scaling up the cultivation process. The team anticipates that reducing Willow’s fundamental hardware noise by half would lower magic state errors to about two in 100,000, as error rates diminish with the cube of the chip’s noise level.