Microsoft has hit a new major milestone in the path to quantum computing at scale by more closely integrating quantum computer hardware from Quantinuum with its high-performance computers on the Azure cloud. This, the tech giant says, will allow developers to utilise the best processor for the job including switching seamlessly from QPU, CPU and GPU tasks. Microsoft says this is an essential development in the path to scalable, fault-tolerant quantum computing.
The breakthrough will see Microsoft open its integrated hybrid feature to Azure Quantum users, which includes developers and scientists. While quantum computing might not have the power or scale necessary today, the company says this puts it firmly on the path to quantum advantage.
The seamless integration of classical and quantum compute power is a first for the cloud sector, Microsoft claims, allowing researchers to start developing hybrid applications that run on today’s hardware and could run when we reach quantum advantage.
Microsoft is making the 20 qubit H1-1 quantum computer produced by Quantinuum available as part of the hybrid rollout. The machines recently hit a landmark five-digit quantum volume benchmark, used to determine the potential usability of the machine as a whole.
“The ability to develop hybrid quantum applications with a mix of classical and quantum code together will empower today’s quantum innovators to create a new class of algorithms,” wrote Krysta Svore, distinguished engineer and VP of advanced quantum development at Microsoft in a blog post on the announcement.
“For example, now developers can build algorithms with adaptive phase estimation that can take advantage of performing classical computation, and iterate and adapt while physical qubits are coherent,” she said. Adding that students can learn algorithms leveraging high-level programming constructs and scientists can explore ways to advance quantum error correction on real hardware. “Taken together, a new generation of quantum algorithms and protocols that could only be described in scientific papers can now run elegantly on quantum hardware in the cloud. A major milestone on the journey to scaled quantum computing has been achieved,” she said.
Quantum and AI
Many in the quantum computing sector share the view that the cloud will be essential to deliver quantum computing at scale. While some very large companies and governments may own machines, the technology is growing at such a pace that delivery over the cloud is the most viable solution and allows for easier hybrid integration.
“A fundamental part of our plan to reach scale is to integrate our quantum machine alongside supercomputing classical machines in the cloud,” wrote Svore. “A driving force of this design is the reality that the power of the cloud is required to run a fault-tolerant quantum machine. Achieving fault tolerance requires advanced error correction techniques, which basically means making logical qubits from physical qubits. While our unique topological qubit design will greatly enhance our machine’s fault tolerance, advanced software and tremendous compute power will still be required to keep the machine stable.”
The achieve fault tolerance Microsoft says the quantum machine will have to integrate with peta-scale classical computers on a network that can handle ten to 100 terabits per second bandwidth between the two processors. Then at every logical clock cycle of the quantum computer the back and forth will be required to keep it alive and yielding a reliable output. “Fault tolerance in quantum computing at scale means that a machine has to be able to perform a quintillion operations while making at most one error,” she said.
The value of the cloud is that it can achieve this at scale, and at a reduced cost. Microsoft gave the example of a project involving ETH Zurich and the Pacific Northwest National Laboratory that saw a ten-times cost reduction for the simulation of a catalytic chemical reaction over non-cloud solutions.
Microsoft says the future of computing is the integration of quantum, HPC and AI at scale. It will allow for new classes of cloud applications including sorting massive data sets using AI, using HPC to narrow options and then quantum at scale to improve the accuracy of the model.