---
title: Google Runs a 2024 Quantum Theory on Real Hardware as Edinburgh Moves Quantum Simulation Onto Ordinary Computers
description: Google cut magic state errors 40-fold on its Willow chip, and Edinburgh put photonic quantum simulation on ordinary computers at about 5% of the power.
author: Darie Nani (Editor-in-Chief)
date: 2026-10-19T12:28:48.316Z
updated: 2026-10-19T12:28:48.328Z
canonical: https://www.sovereignmagazine.com/article/google-quantum-magic-state-edinburgh-algorithm
image: https://cdn.nanimediahouse.com/google-willow-quantum-chip-461704.webp
categories: Science & Tech
content_type: Analysis
region: Global
publication: Sovereign Magazine
schema_type: Article
---

On October 14 the University of Edinburgh said its scientists had written an algorithm that lets off-the-shelf computers run quantum simulations that until now needed some of the world's most powerful supercomputers. A day later Google's quantum team said it had taken a technique that existed only in a 2024 research paper and made it work on its own Willow chip, cutting errors about 40-fold in one of the most expensive jobs a quantum computer has to do.

[EPB, the city power utility in Chattanooga, began selling commercial access to an IonQ quantum computer in early October](https://www.sovereignmagazine.com/article/epb-first-us-utility-commercial-quantum-computing). Google wrote that if its idea worked on hardware, "it could substantially reduce the size of the quantum computer needed for practical tasks," and Edinburgh says cheaper simulation helps inform the design of improved devices.

## Google Got Magic State Cultivation Working on Its Willow Chip

To run the full range of useful programs, a quantum computer needs a steady supply of specially prepared inputs called magic states, and none of them can be reused. Some estimates put the cost of making them at up to 95% of a quantum computer's entire capacity, and researchers have seen the bottleneck coming since 2004, Google's Emma Rosenfeld and Craig Gidney wrote in a [company blog post](https://blog.google/innovation-and-ai/models-and-research/quantum-computing/magic-state-cultivation-quantum-computing/) on October 15.

In 2024 Gidney and fellow Google researchers Noah Shutty and Cody Jones developed a method they called magic state cultivation, starting from an idea already in the research literature. The machine injects a noisy magic state into a small region of the chip, runs a series of checks and discards every attempt that comes out flawed.

On Willow, a processor with 105 superconducting qubits, the team cut errors about 40-fold compared with magic states prepared without cultivation. The T-states, one kind of magic state, reached a record logical fidelity of 99.99%, and the team moved them from a small error-correcting code into a larger one. The work appeared in the journal Science. Rosenfeld and Gidney wrote that the experiment "takes a key theoretical idea and turns it into an experimentally demonstrated reality."

At one stage Rosenfeld's measurements looked physically impossible, and the cause turned out to be ordinary lab noise from drifting electrical equipment and temperature changes, which was interfering with the measurement technique.

## Ordinary Computers Matched Some Photonic Quantum Devices in Edinburgh's Tests

In the Edinburgh team's tests, classical computers running the new algorithm could match and even outperform some current photonic quantum devices at Gaussian boson sampling, a simulation designed to show what quantum devices powered by light can do that ordinary computers cannot. The university says the algorithm could make it cheaper and easier to run the experiments that show the limits of current quantum computers and guide the design of better ones, work that until now needed computing power available at only a handful of facilities worldwide.

A team led by the university and EPCC, the UK's first National Supercomputing Centre, working with Polytechnique Montreal, wrote the algorithm, which runs the simulation using around 5% of the computing power a supercomputer needed, the university said in [its announcement](https://www.ed.ac.uk/news/software-advance-could-aid-quantum-progress).

"Our work has contributed significant improvements in memory use, running time and energy costs in the simulation of photonic quantum computing devices, making mainstream what was before only accessible to a handful of facilities in the world," said Raul Garcia-Patron Sanchez of Edinburgh's School of Informatics.

The team is now working to scale the method up. The university says that could let scientists simulate photonic quantum computers in unprecedented detail.

## Google Still Has to Show Cultivation Works at a Larger Scale

Rosenfeld and Gidney wrote that Google still needs to demonstrate magic state cultivation at a larger scale before it is realistic for practical quantum computing, and that plenty of technical challenges remain. The Quantum Insider's Matt Swayne noted that the 40-fold figure applies to preparing the states, not to the performance of the processor as a whole, and that the quality figures count only the attempts the system kept, so the share it throws away will affect what the technique eventually costs to run.

In a separate test of how well the states held up in storage, errors came out about seven times higher than simulations predicted, The Quantum Insider reported. Swayne called the result an important but incremental advance toward practical quantum computing.

## FAQ

**Q: What is a T-state?**
T-states let an error-corrected quantum computer carry out operations that are difficult to perform directly inside an error-correcting code, according to The Quantum Insider.

**Q: What are distance-3 and distance-5 codes?**
Code distance measures how many physical qubit errors must combine to cause an undetected logical error: at least three for a distance-3 code and at least five for a distance-5 code. Google moved cultivated states from a distance-3 color code to a distance-5 surface code in an experiment that used 54 qubits, according to The Quantum Insider.

**Q: Why did Google use reinforcement learning?**
The team used reinforcement learning, a form of machine learning that adjusts controls based on feedback, as a final calibration step to cut the rate at which the circuit detected errors. Training ran for about four hours before data collection, The Quantum Insider reported.

**Q: Who funded and published the Edinburgh research?**
The work was published in the journal Science Advances and supported by the UK's Engineering and Physical Sciences Research Council and the Natural Sciences and Engineering Research Council of Canada.