IBM, Fermilab, and NVIDIA Plan a Modular Quantum Computing Co-Design Platform

Related stories

IBM announced plans to collaborate with NVIDIA and Fermi National Accelerator Laboratory (Fermilab) on a modular quantum computing co-design platform to advance the integration of quantum and classical computing and explore how these technologies scale together in modular architectures for fault-tolerant quantum computers. The collaboration spans developing new scientific applications, integrating supercomputing and quantum resources, and designing critical components of future quantum computers. Under the plan, the partners intend to co-locate NVIDIA DGX B200 systems with IBM quantum computers, including systems powered by IBM Quantum Nighthawk. IBM Quantum Nighthawk combines a square lattice with a high-speed qubit reset architecture and is accessible through IBM Quantum Platform, where new users can start with 10 minutes of quantum execution time each month on the free plan.

Why Modular Quantum Computing Matters for Scalable Systems

According to MIT Lincoln Laboratory, modular quantum computing is a technique for reaching quantum computation at scale by sharing computational workloads across multiple processing nodes. MIT Lincoln Laboratory researchers note that superconducting qubits generally support only nearest-neighbor or local interactions, making quantum interconnects necessary to transfer fragile quantum information between distant processing nodes. Unlike classical computers that route signals through wires, quantum interconnects must transmit quantum states such as entangled photons through microwave waveguides without destroying quantum coherence. To test how multiple quantum processors can work together, IBM, NVIDIA, and Fermilab’s Superconducting Quantum Materials and Systems (SQMS) Center will explore innovations for executing large sample counts and scaling classical compute alongside linked quantum processors.

Co-Design Platform Hardware and Architecture

The co-design platform pairs classical accelerated computing nodes with IBM’s newest superconducting quantum processor hardware to evaluate how linked quantum processors operate together.

IBM Quantum Nighthawk Processor Specifications

IBM says IBM Quantum Nighthawk is the most powerful quantum processor IBM has built, running circuits of up to 7,500 two-qubit operations on 120 qubits at speeds up to 100,000 samples per second. On the IBM Quantum Nighthawk r2 technical blog, IBM details that Nighthawk r2 combines 120 programmable qubits, 218 dedicated couplers, and 120 independent qubit reset elements for a total of 458 physical quantum elements. In Nighthawk’s square-lattice architecture, most qubits connect to four nearest neighbors, compared to two or three in previous IBM architectures.

Active Dissipative Qubit Reset

Nighthawk r2 replaces conditional reset with a dissipative reset gadget that links each programmable qubit through a tunable coupler to a cold environment, pulling a qubit’s effective T1 from a median of about 200 microseconds down to roughly 25 nanoseconds and reducing idle time between circuit runs to 1 microsecond. IBM reports that Nighthawk r2 delivers 25x higher circuit throughput than IBM Quantum Heron, which runs at roughly 4,000 circuits per second, while lowering initialization error by around 25x.

Integrating Quantum Processors into Supercomputing Workflows

Running hybrid quantum-classical workloads requires workload schedulers that can allocate quantum processors alongside classical compute nodes. Slurm is a widely used scheduler for many of the world’s largest supercomputers that decides which jobs run, when they run, and which CPUs and GPUs they use, and previously had no native concept of a quantum processor.

Quantum Resource Management Interface (QRMI)

To close that gap, IBM and NVIDIA, along with Pasqal, worked with the open-source community to integrate quantum computers into supercomputing workflows using the Quantum Resource Management Interface (QRMI). QRMI is an open-source, vendor-agnostic layer that directly integrates with Slurm Hierarchical Resources (HRES) scheduling capabilities. According to IBM’s QRMI documentation, QRMI is written in Rust and exposes Python and C APIs for integration into computational environments.

Multi-Node QPU Resource Scheduling

Unlike GPUs that operate within individual computing nodes, quantum processing units (QPUs) can be shared across multiple nodes, and HRES lets administrators manage that shared resource directly and set limits for each part of the system. In IBM’s launch announcement, the company notes that this integration lets administrators govern shared quantum resources within Slurm’s core resource model.

How High-Energy and Nuclear Physicists Use the Co-Design Platform

Fermilab’s announcement outlines how physics research teams will apply the co-design platform to scientific workloads and hardware testing.

High-Energy and Nuclear Physics Simulation Teams

Fermilab and IBM will research what is needed to execute complex calculations in high-energy and nuclear physics (HENP) that are beyond classical systems alone, including lattice gauge theories, neutrino interactions, and nuclear simulations. “Some of the most fundamental problems in science remain out of reach for classical computers alone. Scaling quantum and classical together will open new frontiers in high-energy and nuclear physics and beyond,” said Jerry Chow, IBM Fellow and Chief Technology Officer of Quantum-Centric Computing at IBM.

HEPCloud Distributed Research Communities

Fermilab will lead HENP application engagement and coordinate software workflows and researcher access across national laboratories, universities, and international institutions through its HEPCloud infrastructure. In a later phase, Fermilab’s SQMS Center intends to integrate its own ultracoherent superconducting radio-frequency (SRF) cavity hardware into the platform so researchers can investigate cavities as quantum memories, entanglement buffers, and resources for bosonic error correction.

Quantum-Classical Simulation Benchmarks and Chemistry Results

New research from IBM, RIKEN, and NVIDIA demonstrates how quantum-classical workflows execute molecular simulations across combined quantum and GPU hardware. Building on sample-based quantum diagonalization (SQD), researchers studied Fe4S4, an iron-sulfur cluster found inside the proteins of living cells whose interacting electrons make it difficult to simulate. The IBM, RIKEN, and NVIDIA team began with an intentionally imperfect but realistic guess of the molecule’s state, running circuits of up to 4,000 two-qubit operations on 72 qubits and collecting 5,000,000 samples.

The workflow integrated quantum computers and NVIDIA GPUs, including those inside RIKEN’s JHPC-quantum GPU supercomputer ROQUO, achieving what the partners describe as a 79-fold improvement in accuracy over earlier runs of the same experiment. According to the IBM, RIKEN, and NVIDIA paper, the Fe4S4 results came within 4.3 millihartree of the accuracy achieved using density-matrix renormalization group (DMRG), a classical simulation method. “By combining quantum computing with HPC, we achieved accuracy approaching that of DMRG, one of the leading classical methods, for a strongly correlated iron–sulfur cluster,” said Seiji Yunoki, Team Principal of the Computational Materials Science Research Team at RIKEN. The IBM and RIKEN research is supported by Japan’s New Energy and Industrial Technology Development Organization (NEDO) through its Research and Development of Quantum-Supercomputers Hybrid Platform for Exploration of Uncharted Computable Capabilities project, while IBM says a separate collaboration with Cleveland Clinic was named a finalist for the 2026 ACM Gordon Bell Prize.

IBM, Fermilab, and NVIDIA Partner Roles

The three organizations bring distinct hardware, software, and research infrastructure to the co-design effort. IBM provides global hybrid cloud, artificial intelligence (AI), and consulting services to clients in more than 175 countries, and governments and corporate entities in financial services, telecommunications, and healthcare use IBM’s hybrid cloud platform and Red Hat OpenShift. The IBM Quantum Network includes more than 325 total network members, more than 65 commercial partners and startups, and more than 95 completed prototypes.

Separately, NVIDIA released the TCBP-JOBS decoding algorithm in its CUDA-Q QEC 0.9.0 library, which NVIDIA says delivers up to a 13,000x speedup for quantum error-correction research and cuts design computation time from 12 weeks to 45 minutes. “Realizing quantum computing’s potential requires quantum processors and accelerated computing to evolve together as accelerated quantum supercomputers,” said Krysta Svore, vice president of applied research for quantum computing at NVIDIA. Meanwhile, Fermilab is managed by Fermi Forward Discovery Group for the U.S. Department of Energy (DOE) Office of Science, and its SQMS Center is one of five DOE National Quantum Information Science Research Centers, bringing together more than 300 experts from 43 partner institutions across national laboratories, academia, and industry. “Our ambition is to help lead the development of quantum computers built for discovery: bringing high-energy and nuclear physicists together to co-design the applications, algorithms and modular architectures that will ultimately make fault-tolerant quantum computing useful for science,” said Anna Grassellino, chief technology officer and director of the SQMS Center at Fermilab.

Questions about modular quantum computing co-design platform

How does Slurm schedule quantum processors alongside CPUs and GPUs?

Slurm can allocate, reserve, and account for QPU time with the same policies high-performance computing (HPC) users use for CPUs and GPUs, allowing HPC centers to manage a quantum-classical workflow as a normal Slurm job without a separate queue, login, or operating model. Previously, supercomputing sites had to attach QPUs through plugins outside Slurm’s core resource model. Because QRMI is vendor-agnostic, the same integration approach can support quantum processors from different providers.

What operating systems and compilers does QRMI require?

QRMI supports AlmaLinux 9, Amazon Linux 2023, CentOS Stream 9 and 10, RedHat Enterprise Linux 8, 9, and 10, Rocky Linux 8 and 9, SuSE 15, Ubuntu 22.04 and 24.04, and MacOS Sequoia 15.1 or above. Compiling QRMI requires Rust compiler 1.91 or above, a C11-compatible compiler such as GCC or Clang, make or cmake, openssl, zlib, and Python 3.11, 3.12, or 3.13 for the Python API.

How do Nighthawk r2 and SQMS hardware support error correction and scaling?

Nighthawk r2 supports mid-circuit measurement and dynamic circuits, allowing repeated use of auxiliary qubits during quantum error detection and error correction protocols. At Fermilab, SQMS engineers are also developing a large-scale, high-cooling-power milli-kelvin dilution refrigerator for future quantum computing data centers.

Also Read: GIGABYTE expands AI portfolio with GIGABYTE AI TOP AERO

Future Phases and Roadmap Milestones

In the initial phase of the Fermilab, IBM, and NVIDIA platform, researchers will model and emulate communication between quantum modules to study how link errors, latency, and feedback affect performance, guiding SQMS and IBM in developing physical quantum interconnects. The co-design platform aligns with the DOE’s Quantum Genesis initiative, which aims to deploy a fault-tolerant, scientifically relevant quantum computing capability by 2028. Looking further ahead, IBM’s quantum roadmap plans for Nighthawk to scale to 10,000 operations on hundreds of qubits in 2027 and reach 100,000,000 operations on 200 qubits by 2029 via IBM Quantum Starling, complementing IBM’s separate co-design platform in Poughkeepsie, New York with Oak Ridge National Laboratory (ORNL) and AMD that pairs Nighthawk r2 processors with ORNL’s Umbra HPC system built on AMD Instinct MI355X GPUs and 5th Gen AMD EPYC CPUs.

Subscribe

- Never miss a story with notifications


    Latest stories