Microsoft says it’s cracked the code on an important quantum computing problem

Microsoft says it’s figured out how to improve error rates in quantum computing, bringing quantum computing closer to a commercial state. 

The company worked in collaboration with quantum computing hardware maker Quantinuum to improve the performance of the qubit — the very basic unit of quantum computing. Qubits work by holding two different phases at once (instead of just a one and a zero, it’s both), but they aren’t very stable, making it easy for them to lose data. Researchers can now create several “logical qubits,” or qubits that are more stable while holding these different states.

Krysta Svore, vice president of advanced quantum development at Microsoft, told The Verge in an interview that because qubits are prone to errors, researchers needed to find a way to stabilize them.

“We need reliable quantum computing, and not just in theory; we need to demonstrate that it can work in practice,” Svore says. “I like to think of it as putting noise-cancelling headphones on the qubits.”

She says that these more reliable qubits help quantum computing “graduate” from level one, the more foundational level with qubits prone to mistakes and are usually referred to as noisy, to the next level, where scientists can run more calculations correctly and scale up the technology for more commercial use.

Other quantum computing experts welcomed Microsoft and Quantinuum’s advancement. Henry Yuen, associate professor of computer science at Columbia and a theoretical computer scientist, tells the Verge via email this may just be the beginning of more discoveries that make quantum computing easier.

“We’re far from the final destination, but the signposts are getting more frequent and are indicating that some major milestones are coming up soon,” Yuen says. “I’m sure there will be bigger and better demonstrations of quantum fault tolerance coming soon.”

Microsoft brought its qubit-virtualization system, which Svore says “abstracts” groups of physical qubits together, to Quantiuum’s quantum computer to create virtual logical qubits.

With it, users could create qubits with a longer fault tolerance, or time without encountering an error. The team created four reliable logical qubits from only 30 physical qubits. Previously, the scientific consensus was that hundreds of physical qubits were needed to make a couple of logical qubits that didn’t fail, and they would have taken decades to create.

The teams ran 14,000 calculations without losing the quantum state and found they improved the error rate by a factor of 800 over physical qubits. Svore says the system could detect and fix errors without destroying the logical qubit and keeping the string of calculations going. 

Microsoft is now figuring out how to bring this capability to Azure Quantum Elements, its platform for scientists to use AI, high-performance computing, and quantum computing to run scientific experiments. 

Yuen says that while he thinks the term “quantum virtualization” may be Microsoft’s branding for error-correcting code, its findings could be scalable for other quantum computing companies to try on their own.

Quantum computing has always seemed like far into the future innovation, despite the idea and experimentation being around for decades. Companies such as IBM, Microsoft, and Google have been trying to make quantum computing reliable, safe, cost-effective, and, more importantly, useful for years. 

Quantinuum chief product officer Ilyas Khan and senior director of offering management Jenni Strabley said in a blog post that they plan to continue improving the system to create more reliable logical qubits.

“In the short term – with a hybrid supercomputer powered by a hundred reliable logical qubits, we believe that organizations will be able to start to see scientific advantages and will be able to accelerate valuable progress toward some of the most important problems that mankind faces such as modeling the materials used in batteries and hydrogen fuel cells or accelerating the development of meaning-aware AI language models,” Quantinuum said in its post. 

Now, with Microsoft and Quantiuum’s work, it’s up to others to see if they can replicate the same thing.