IBM’s quantum bet: Run on classical what you can, on quantum what you must
Network World ·

More qubits can expand what a quantum computer attempts, but they do not, by themselves, make it useful. IBM’s new Nighthawk r3 processor is intended to address that gap by increasing scale without giving up the speed and quality of its current generation. Nighthawk r3 is 4x larger than Nighthawk r2, speeds up processes up to 19x, and can execute 10,000 quantum operations across 520 qubits, IBM said. This allows it to tackle larger, more complex problems. But while it’s building a bigger quantum processor, IBM isn’t moving entire workloads onto it. Its strategy is hybrid: Leave conventional tasks to classical computers and isolate the work that calls for quantum. “This colocation of quantum and classical, and giving the right part of the problem to each one, is one of the most exciting and important research areas that we have right now, and in the near-term years to come,” said Oliver Dial , VP for Quantum Systems at IBM. Nighthawk r3 a ‘meaningful step’ “The power of a quantum processor is related to its scale, the number of qubits, its quality, how many operations it can run, and its speed,” Dial said. Scale, size, and quality support more complex computations, but speed determines whether they can be completed within reasonable timeframes at reasonable costs. IBM Quantum Nighthawk r2, which was introduced in September, represented what Dial called a “huge step forward in speed.” Its qubit reset technology reduces the time between circuit executions, allowing applications to more efficiently collect the millions or billions of samples needed for complex computations. Nighthawk r3 will build on r2’s performance, Dial said. It is designed to support 10,000 quantum operations on 520 programmable qubits, with 2034 physical qubits, and introduces 10 chip-to-chip interconnects, called l-couplers , designed to connect processors into larger systems. By 2027, IBM plans to link two Nighthawk r3’s together into a single system with more than 1,000 qubits. The exact speedup users can get from Nighthawk r2 or r3 depends on the workload; early partners are seeing increases of 9x up to 19x over previous IBM quantum processors, Dial said. But quantum isn’t optimal for everything, and it can be error-prone. To address these issues, IBM’s Quantum Platform is bringing high-performance (HPC) classical and quantum computing together into a single managed environment. Dial explained that industry-relevant workloads demand layers of quantum error correction and mitigation techniques alongside noise learning and post-processing. Users can now integrate quantum workflows directly with a range of classical compute, or with newly-available HPC and classical resources. They will get “precise control” over how they can detect, understand, and better correct errors within the combined architecture, giving them the ability to time operations better, reset qubits faster, and extract more information from quantum measurements. Gartner VP Analyst Mark Horvath called Nighthawk r3 a “meaningful step forward by IBM.” He pointed out that, only a few years ago, most physicists and analysts predicted we wouldn’t reach this level of achievement before 2045, if ever. The Nighthawk r3 has a strong chance of delivering meaningful results that will impact a large range of fields, and lays a solid foundation for building into the future, Horvath said. It is a “strong step towards a cryptographically relevant quantum computer.” Colocation an exciting and important research area Dial agreed that the next few years will be about making quantum business-relevant ; he sees a strong potential for optimization, more specifically “multi-objective optimization,” for instance, in logistics, when weighing the trade-off between risk, delivery time, or cost. In principle, anything that you can do on a classical computer, you can also do on a quantum computer, Dial said. “But if there is something that you can run on a classical computer, you should run it on a classical computer. It’s going to be cheaper, faster, more reliable,” he pointed out. Due to their underlying computational abilities, quantum computers are “exponentially faster” than classical computers for some functions, he noted. The goal is to extract the portions of algorithms that require quantum, then run everything else on a classical computer. The benefits are fewer errors and improved accuracy. Gartner’s Horvath sees the biggest immediate impact of the technology in financial services, chemistry, and overall logistics like delivery and scheduling. IBM’s advancements could also drive improvements in machine learning (ML), weather prediction, economic output modeling, and engineering. He compared quantum computers to specialized graphics cards, explaining that there are about 20 algorithms that (theoretically) run at advantage on a quantum computer, but right now most of the work is focused on 4 of that 20, “with the others being out of reach.” Quantum computers don’t have conventional operating systems, he noted, so they are paired up with classical systems that perform tasks like setting up the problem and communicating, while the quantum computer “only does the things it’s good at.” This colocation is important because fault tolerance is “fiendishly precise work,” he said. Quantum computers are sensitive to almost every kind of disturbance: vibration, temperature, ambient magnetic fields, even changes in gravity. They require a very controlled environment, and any disturbance could affect the machine. “Mastering quantum physics to turn it into a computing strata is a tough lift,” Horvath said. He likened it to an auditorium full of 3-year-olds: “Now imagine your task is to get them to sing the Star-Spangled Banner in multi-part harmony. It’s even more difficult than that.” Optimization problems, such as Traveling Salesman/Knapsack, scheduling, or logistics, run on quantum today, but don’t always demonstrate consistent advantage, Horvath said, adding that a processor with 520 programmable qubits could lead to improvements in areas like trading strategies, logistics, and chemistry. Additionally, quantum algorithms that scientists have never before been able to run would be within reach, leading to new solutions in math, quantum Monte Carlo simulations (random sampling), and improvements in AI. Dial acknowledged that quantum is a field still full of hype, with “big plans and big dreams.” But IBM has a quantum roadmap, he said, and has consistently stuck with it. “We have a lot of fundamental work to do between now and 2029,” he acknowledged. But “the same way that AI is revolutionizing things this year, quantum is coming.”