IBM, Quantinuum and Intel signal a new phase in the quantum computing race
Network World · · g4646571
Announcements from IBM and Quantinuum plus a recent acquisition by Intel show that the race to develop a quantum computer has shifted from focusing just on the qubits to integrating these computers with classical systems and software development platforms. To start with, IBM, AMD, and the DOE-funded Quantum Science Center at Oak Ridge National Laboratory are collaborating on a new research platform that pairs IBM’s 120-qubit Nighthawk quantum computer with a classical computer in order to allow researchers to better explore quantum error correction, the organizations announced. This is a research platform rather than a commercial product, says IDC analyst Heather West . “Quantum error correction requires measuring qubits, decoding the results, and applying corrections in real time, which depends on substantial classical compute close to the quantum hardware,” she says. “This platform gives researchers a place to test different error correction approaches on real hardware and understand what each requires in quantum and classical resources.” It’s part of a broader trend in the industry, West added. “Quantum computing is increasingly being developed as part of a larger hybrid environment alongside HPC and AI infrastructure,” West told Network World . “And we are seeing more collaboration among national laboratories, quantum hardware providers, and classical computing vendors.” “A quantum computer doesn’t work on its own,” says Jerry Chow, CTO of Quantum-Centric Supercomputing at IBM. “It needs to work together with other classical computing resources.” You can think of a quantum computer as similar to a graphics accelerator, Chow told Network World , where the GPU steps in when needed to handle graphics-related tasks or matrix operations while the CPU continues to handle most other common tasks. “CPUs are really good at arithmetic and they’ll keep doing that,” he says. “GPUs are really good at tensors — the new math that’s behind all the AI stuff. And quantum computers are really good at quantum circuits. So when your problem is best represented by quantum circuits, you can off-load that problem to the quantum computer.” In addition, some error correction techniques involve running error correction codes or other error mitigation strategies on classical computers. “This new co-design platform gives us the opportunity to explore that,” he says. The way it works is that Oak Ridge’s Umbra high-performance computing cluster is now located next to IBM’s Nighthawk quantum computer in IBM’s Poughkeepsie computing center. The Nighthawk chip sits inside IBM’s Quantum System Two housing, an enclosure about the size of a meat locker, he says, with a hexagonal central box and an electronics box around it, built for data center environments. The Oak Ridge National Laboratory’s Umbra is the size of a standard server rack, with AMD’s Instinct GPUs and EPYC CPUs. There are 16 GPUs total, he says, and the Umbra is linked to the Nighthawk with Ethernet. According to Chow, the Nighthawk has 120 programmable qubits, which is the number of qubits that can actually be used to run a quantum circuit. The machine actually has a total of 458 physical qubits, but the rest are used to help improve performance and accuracy. He confirmed that it’s a research platform. “We’re not offering packaged solutions by any means yet,” he says. In other news, Quantinuum has announced a roadmap for its Nexus development platform, and its goal is to make fault tolerance invisible by 2029, allowing developers with no knowledge of error correction to create quantum programs, where Nexus would pick the codes automatically. Quantinuum’s Nexus platform is designed to work with its Apollo fault-tolerant quantum computer, planned for 2029, but also works with the company’s current Helios quantum computer, both of which use the trapped ion architecture. It first launched in 2024 and now supports more than 200 organizations. “We’re building the first developer platform specifically for the fault-tolerant era so customers can start developing applications today and carry that investment forward,” Quantinuum CEO Rajeeb Hazra says in the announcement. Meanwhile, last week, Intel announced that it would buy quantum computing software company Classiq. The idea is to combine Classiq’s development platform with Intel’s spin-qubit manufacturing and existing software in order to transform Intel’s Quantum Computing Group “into a full-stack quantum computing business,” according to Intel CEO Lip-Bu Tan . The deal is reportedly valued at between $1.9 and $2.3 billion, making it the largest quantum-related acquisition to date.