Bridging the HPC Software Gap for Practical Quantum Computing - EE Times

In the global competition to build a functional quantum computer, a quiet consensus is forming among high-performance computing (HPC) centers: Hardware is advancing rapidly, but the software required to run it is falling behind.A report released today by French quantum computing startup Alice & Bob and analyst firm Hyperion Research indicates that the primary bottleneck in integrating quantum systems with modern supercomputers is no longer the qubits themselves. Instead, it is the underlying software infrastructure, compilers, and schedulers that translate scientific problems into machine instructions.“The quantum industry and the hardware have been progressing quite significantly,” Juliette Peyronnet, general manager U.S. for Alice & Bob, told EE Times in an interview. “There have been tremendous investments and signals from the market. Adoption is accelerating. But one gap that is identified here is in regard to the software stack. Hardware and quantum computers are not going to work in isolation; they are going to be hybrid resources integrated for the first adopters in high-performance computing centers.”The report, HPC-Quantum-AI: Shaping the Next Compute Era, draws on interviews with 15 directors and engineers at major supercomputing facilities, including Oak Ridge National Laboratory, Lawrence Berkeley National Laboratory, and Japan’s RIKEN. It warns that without standardized software interfaces, centers risk wasting months on custom integration plumbing.“HPC end users everywhere are looking for guidance and insights for what it will take to integrate the computational power of quantum into the current and projected advanced computing ecosystem,” said Bob Sorensen, SVP of research and chief analyst for Quantum Computing at Hyperion Research. “This report covers the key issues required to successfully navigate that process by QC vendors and end users alike.”At the center of this integration is how to correct the errors that make quantum computers unstable. Standard quantum processors are highly sensitive to noise, requiring complex error-correction schemes that demand up to a thousand fragile physical qubits to protect a single logical qubit, the basic unit of calculation.Alice & Bob is pursuing cat qubits, a superconducting quantum technology designed to ease that scaling bottleneck. Combined with LDPC error-correction codes, its architecture is claimed to reduce physical hardware requirements by up to 200× compared to competing approaches.“With Alice & Bob, the main change is that we only have one type of error to correct,” Peyronnet told EE Times. “Quantum computers performing operations have two types of errors that can come and perturb the good functioning of your machine: bit-flip errors and phase-flip errors. Those are computationally expensive and demanding to correct. In our platform, the bit-flip time is corrected through the design of the device itself. You collapse on one dimension the error-correction scheme that you are trying to implement, meaning that you need way less physical qubits for one logical qubit.”This design focuses error-correction efforts entirely on phase-flips. Recent experimental results from the company’s labs showed that its cat qubits could remain stable against bit-flips for up to an hour, far exceeding standard superconducting qubits, which typically experience bit-flips in milliseconds.Even with a reduced footprint, quantum computers cannot operate independently. Correcting phase-flips requires a classical computer to monitor physical qubit measurements. In superconducting systems, this decoding process generates a torrent of data to process within a sub-10-microsecond window to prevent the processor from stalling.“They’re going to have to leverage classical computing to perform some of their key operations internally within a quantum node,” Peyronnet said. “But they’ll also have to interact externally with a bunch of classical resources for performing useful applications. And all of this software plumbing for the fault-tolerant regime—the machine that will actually deliver value—is not there yet.”To bridge this gap, Alice & Bob is working with Nvidia, an investor in the startup. Nvidia is promoting its CUDA-Q programming model as a platform to run heterogeneous backends. However, standard software platforms still require substantial customization to accommodate different hardware modalities.“We work closely with Nvidia, and we work with CUDA-Q, but there is still a lot of development that needs to happen to make sure that the specificities of cat qubits are taken into account,” Peyronnet argued. “We are co-working with Nvidia to make sure things are cat-qubit compatible. A product is not just something you can plug into a quantum computer, and somehow all of your problems are solved. There’s just a lot of collaboration that still needs to happen.”While governments and venture capitalists have poured billions into quantum computing, commercial returns remain a long-term prospect. Major financial institutions such as JPMorgan Chase and HSBC are researching hybrid systems for portfolio optimization, where even a minor improvement yields significant financial value. Yet, the broader commercial timeline is measured in years.“Definitely, there is no return on investment today for anyone,” Peyronnet recognized. “This is still a technology that needs to mature in order to deliver value for enterprises. The first machines that will be useful won’t be for the industry. They will be useful for a niche community of researchers specialized in some problems where we can see overlap with their interest in what quantum computing can achieve.”Juliette Peyronnet noted that the industry is in an early adopter phase that is expected to persist for several years.“We are the first adopter, innovator type of persona,” Peyronnet said. “And we are going to stay here at least until the beginning of 2030, and then we’re going to progress toward making this technology more accessible to other use cases. We’re maybe 20 to 30 years behind where artificial intelligence is today. It’s still very early. So yes, there is going to be ROI, but it’s going to take time.”Rather than waiting for hardware to mature in isolation, the report urges supercomputing centers to engage in co-design, in which domain scientists, software developers, and hardware specialists collaborate on specific applications before physical processors are delivered. Alice & Bob is currently using a $3.9 million award from the U.S. Department of Energy’s ARPA-E program to co-design a chemistry workflow with GE Vernova, Los Alamos National Laboratory, and the University of Michigan, targeting rare-earth-free magnet simulations.For HPC centers, the long-term lesson is that readiness cannot be outsourced or delayed. The supercomputing facilities that build in-house software and operational expertise today will be the ones to deploy the first useful quantum supercomputers in the coming decade.How BSC Contributes to Europe’s Hybrid Quantum StrategyNvidia Bets on the Classical Side of Quantum ComputingHow Spain Built a Quantum Ecosystem Without Calling It One RELATED TOPICS: CAT QUBITS, HIGH PERFORMANCE COMPUTING (HPC), QUANTUM ADVANTAGE, QUANTUM COMPUTING, QUBITS, SUPERCOMPUTING COMPANIES: ALICE & BOB, HYPERION RESEARCH, IBM, NVIDIA Pablo is a seasoned engineer with 30+ years of experience. For over 10 years, he's been a contributing editor for EE Times (now editor of the Supply Chain section). He also wrote for EPSNews, InformationWeek, EBN, LightReading, Network Computing, and IEEE Xplore. His coverage spans Supply Chain, Semiconductors, Networks, IoT, Security, and Smart Cities. He holds an MEng, Electrical and Electronics Engineering from The Ohio State University. Follow Pablo on LinkedIn You must Register or Login to post a comment.This site uses Akismet to reduce spam. Learn how your comment data is processed.