NVIDIA Quantum
Accelerating the future of scientific discovery.
Overview
What Is Accelerated Quantum Supercomputing?
Quantum processing units (QPUs) implement a new model of computing with potential impact across industries. Turning QPUs into useful quantum computers means integrating them with state-of-the-art AI supercomputers. NVIDIA is working with QPU developers to build accelerated quantum supercomputers.
Accelerating Quantum Computing Research
NVIDIA Accelerated Quantum Computing Research Center
The NVIDIA Accelerated Quantum Computing Research Center (NVAQC) is where qubits go to become accelerated quantum supercomputers. By integrating partner quantum hardware with a state-of-the-art NVIDIA GB200 NVL72 system, the NVAQC brings developers and academics the breakthroughs needed to scale quantum computing hardware and applications.
Solutions
Accelerating Quantum Computing
NVIDIA NVQLink
NVQLink is an open architecture for integrating all quantum processor modalities with AI supercomputing. NVQLink brings low-latency, high-throughput, GPU-accelerated QEC performance to all quantum hardware. NVQLink provides the missing puzzle piece for QPU builders, quantum controller vendors, and supercomputing centers to work together to scale quantum computers to hybrid, utility-scale applications.
NVIDIA CUDA-QX
Quantum researchers and developers across a wide range of domains, from quantum error correction to hybrid solvers, use GPU programming to accelerate their applications. This demands highly optimized, domain-specific libraries. NVIDIA CUDA-QX, built on top of NVIDIA CUDA-Q™, is a collection of libraries and tools for accelerating research and development toward useful accelerated quantum supercomputing.
Belief Propagation-Ordered Statistics Decoding (BP-OSD) is one of the most promising approaches for scalable quantum error correction. CUDA-Q QEC accelerates BP-OSD with state-of-the-art latency and throughput, offering a 29–35x speedup over industry standard decoders for a single shot, as well as an additional speedup of up to 42x for high-throughput use cases where many syndromes need to be decoded at once.
NVIDIA CUDA-Q
Quantum-accelerated applications won't run exclusively on a quantum resource but will be hybrid (quantum and classical) in nature. To transition from algorithm development by quantum physicists to application development by domain scientists, a development platform is needed that delivers high performance, interoperates with today's applications and programming paradigms, and is familiar and approachable to domain scientists.
With a unified programming model, NVIDIA CUDA-Q is a first-of-its-kind platform for hybrid quantum-classical computers, enabling integration and programming of QPUs, quantum emulation, GPUs, and CPUs in one system. CUDA-Q is built for performance, is open source, and provides high-level language to develop and run hybrid quantum-classical applications.
NVIDIA cuQuantum
NVIDIA cuQuantum is a set of low-level libraries for accelerating quantum circuit simulation. cuQuantum is primarily used by developers building circuit simulation frameworks and accelerates Cirq, Qiskit, PennyLane, and more.
cuQuantum offers state vector (cuStateVec) and tensor network (cuTensorNet) circuit simulation algorithms with multi-GPU acceleration.
The cuQuantum Appliance is a Docker container consisting of leading community frameworks accelerated by cuQuantum and optimized for the NVIDIA platform.
The NVIDIA cuQuantum Appliance is available in the NVIDIA NGC™ catalog.
NVIDIA cuPQC
To ensure the security and authenticity of the world’s sensitive data, it is now critically important that organizations migrate to algorithms that can withstand a quantum computing attack. This new quantum-safe encryption is known as post-quantum cryptography (PQC).
NVIDIA cuPQC provides secure and accelerated implementations of the leading PQC algorithms, advancing data security against quantum computer threats and enabling cryptographic research.
Resources
NVIDIA CUDA-Q Academic
To grow a quantum-ready workforce, NVIDIA is teaming up with academic institutions to develop educational resources through the CUDA-Q software development platform. CUDA-Q Academic offers training for both researchers and college students through self-paced, online modules, complete with interactive coding exercises and videos. Students gain the skills needed to work with the accelerated quantum supercomputers that will run useful applications.
NVIDIA Quantum Ecosystem
NVIDIA Quantum is enabling the entire quantum ecosystem—and some of the most important research happening today. From quantum computing startups to some of the largest companies in the world, academic labs and supercomputing centers to Fortune 500 companies, we’re proud to help our partners develop and leverage quantum.
Sign up for the latest news, updates, and more.
FAQs About Quantum Computing
Quantum computing is a paradigm-shifting technology that leverages the laws of quantum physics to solve extremely difficult problems—problems so hard that their solution on conventional supercomputers would require an impossible amount of time or resources.
The centerpiece of quantum computing is the quantum bit, or qubit. Classical bits can only exist in a state of 0 or 1, while qubits can exist in a superposition of both states simultaneously. This means N qubits in superposition hold information about an exponential (2ᴺ) number of binary configurations at once. Quantum computers additionally exploit the quantum properties of entanglement—where qubits become correlated so that the state of one instantly reflects the state of another—and interference, which allows certain computational paths to be amplified while others are canceled out. Together, these properties enable a fundamentally new model of computation.
Classical computers process information using bits that are strictly 0 or 1. Every operation is deterministic and sequential at the bit level. Quantum computers use qubits, which can represent a superposition of 0 and 1 simultaneously, allowing them to explore many possible solutions in parallel.
However, quantum computers are not simply “faster” classical computers. They are specialized accelerators suited to specific classes of problems—such as simulating quantum chemistry, optimizing complex systems, and breaking or building cryptographic protocols—where classical machines struggle even at supercomputer scale. For most everyday computing tasks, classical hardware remains more efficient. This is why quantum computers are expected to function as accelerators alongside classical supercomputers, not as replacements.
Quantum computing is expected to impact a wide range of industries, including defense, energy, logistics, engineering, medicine, finance, and retail. Key application areas include:
- Drug Discovery and Materials Science: Simulation of complex chemical and molecular systems to guide the development of new medicines, batteries, solar cells, and consumer products
- Financial Services: Faster estimation of risk metrics and financial predictions, enabling more profitable trading and more precise risk management
- Optimization: Solving complex logistical challenges, from global supply chains to genetic analysis
- Cryptography and Cybersecurity: Both offensive and defensive capabilities that fundamentally reshape cyber defense and national security
- AI and Machine Learning: Novel procedures that draw deeper insights from data and improve the reliability of predictions for decision-makers
Not all industries will benefit equally or at the same time. Chemical and materials science applications are expected to see early advantages, while others will require larger, later-generation quantum hardware. The most transformative quantum applications are likely still to be discovered.
Quantum computing devices exist today that can prepare, manipulate, and read out hundreds—and increasingly thousands—of qubits. However, these remain largely experimental. Current quantum computers are not yet able to compete with conventional supercomputers on meaningful, real-world tasks because their qubits are adversely affected by noise—environmental interference that corrupts computations.
The goalposts for practical relevance are captured by the concept of utility-scale quantum computing: the point at which a quantum computer can not only outperform classical alternatives but do so cost-effectively across economic, scientific, societal, or environmental dimensions. Achieving this requires continued advances in quantum hardware, quantum error correction, and algorithm development.
Quantum systems are extremely fragile. Even the slightest environmental disturbance—a stray electromagnetic pulse, vibration, or temperature fluctuation—can introduce errors that corrupt a computation. Quantum error correction (QEC) is the set of techniques used to protect quantum information against this noise.
QEC works by encoding many noisy physical qubits into a smaller number of cleaner logical qubits. Error-correcting codes monitor qubits repeatedly to detect where errors have occurred, enabling a classical computer to decode and correct them in real time. For QEC to work at all, the noise level in quantum hardware must be below a critical “threshold” error rate.
Fault-tolerant quantum computers (FTQCs)—which actively implement QEC—are considered the “holy grail” of the field, because only they are capable of running the large, complex algorithms needed for truly useful applications. Building FTQCs requires tight integration between quantum processors and high-performance classical supercomputers.
NVIDIA’s CUDA-Q QEC library accelerates one of the most promising QEC decoding algorithms (Belief Propagation-Ordered Statistics Decoding, or BP-OSD) with a 29–35× speedup over industry-standard decoders for a single shot, with additional speedups of up to 42× for high-throughput use cases.
Hybrid quantum-classical computing is the model in which quantum processors (QPUs) and classical computers work together—each doing what it does best. Quantum hardware accelerates specific computationally hard subroutines, while classical supercomputers handle compilation, calibration, control, error correction, post-processing, and other supporting tasks.
This pairing is not a temporary workaround but an enduring architectural reality. Most useful quantum applications will be inherently hybrid, requiring both classical and quantum hardware as part of a larger workflow. For certain tasks, low-latency communication between the two is critical, meaning QPUs and classical processors must be physically colocated.
This reciprocal relationship—where quantum computers enable supercomputers to solve new classes of problems, and supercomputers enable quantum computers to operate—is known as quantum-accelerated supercomputing.
NVIDIA is building the full-stack infrastructure needed to accelerate the development of useful, accelerated quantum supercomputers. Rather than building QPUs directly, NVIDIA focuses on the classical supercomputing layer that quantum hardware depends on—providing the software platforms, libraries, interconnects, and research infrastructure that the entire quantum ecosystem needs.
NVIDIA’s quantum computing portfolio includes:
- CUDA-Q: An open source, hybrid quantum-classical programming platform
- CUDA-QX: Domain-specific libraries for quantum error correction, hybrid solvers, and more
- cuQuantum: GPU-accelerated quantum circuit simulation libraries
- NVQLink: A high-speed, open interconnect for integrating QPUs with AI supercomputers
- NVIDIA Ising: An AI model family and training framework for automating quantum calibration and error-correction decoding
- cuPQC: Accelerated post-quantum cryptography implementations
- NVIDIA Quantum Cloud: Cloud-based access to NVIDIA’s quantum computing platform
- NVAQC: The NVIDIA Accelerated Quantum Computing Research Center
NVIDIA is also partnering with leading academic institutions worldwide through the CUDA-Q Academic program to train the next generation of quantum-literate scientists and engineers.
NVIDIA CUDA-Q is the world’s first unified programming platform for hybrid quantum-classical computers. It enables developers to integrate and program QPUs, quantum emulators, GPUs, and CPUs within a single system and a single programming model.
Key features of CUDA-Q include:
- Unified Hybrid Programming: Developers write a single program that distributes work across quantum and classical hardware automatically.
- QPU-Agnostic: CUDA-Q can compile and run code on a wide range of quantum hardware from different vendors.
- Open Source: The platform is open source and freely available.
- High-Level Language Support: CUDA-Q supports familiar languages like Python and C++, lowering the barrier for domain scientists who are not quantum physicists.
CUDA-Q is the foundation on which NVIDIA builds its entire quantum software stack, from error-correction libraries to the CUDA-Q Academic curriculum.
CUDA-Q Academic is NVIDIA’s initiative to grow a quantum-ready global workforce by partnering with universities and research institutions to develop educational resources built on the CUDA-Q platform. It offers self-paced, online learning modules complete with interactive coding exercises and videos, designed for both researchers and undergraduate/graduate students. Learners gain the practical skills needed to work with the accelerated quantum supercomputers that will power future real-world applications.
The curriculum is freely available via the CUDA-Q Academic GitHub repository, and NVIDIA partners with institutions directly to integrate the content into courses and research programs.
Current academic partner institutions include:
Arizona State University · Carnegie Mellon University · Chung Yuan Christian University Quantum Information Center · Dartmouth Engineering · ETH Zurich Quantum Center · Fordham University · IIT Madras · KAUST · Mälardalen University · Northwestern University · Pittsburgh Supercomputing Center · Polytechnic University of Valencia · Princeton University · Purdue University · Robert Morris University · Technion – Israel Institute of Technology · TUM Chair for Design Automation · UC Davis · University of Cambridge Physical Computation Laboratory · University of Chicago · University of Florida · University of Illinois Urbana-Champaign · University of Nevada, Reno · University of Pittsburgh · UPC Barcelona Tech
Researchers and educators interested in the program can sign up for more information.
NVIDIA NVQLink is an open, high-speed interconnect architecture designed to integrate any type of quantum processor (QPU) with AI supercomputers. Developed in collaboration with Argonne, Brookhaven, Fermilab, Lawrence Berkeley, Oak Ridge, and Sandia National Laboratories, among others, NVQLink provides:
- Low-latency, high-throughput communication between QPUs and GPUs—essential for real-time quantum error correction and hybrid workloads.
- Universality: NVQLink is designed to work with all quantum processor modalities (superconducting, trapped ion, neutral atom, photonic, and others).
- Open Architecture: QPU builders, quantum controller vendors, and supercomputing centers can all interoperate through NVQLink.
Leading supercomputing centers worldwide are already adopting NVQLink to integrate quantum processors into their facilities, enabling large-scale quantum-classical workflows.
The NVIDIA Accelerated Quantum Computing Research Center (NVAQC) is a dedicated research facility—located in Boston—focused on shortening the timeline to useful quantum computing. The NVAQC integrates partner quantum hardware with NVIDIA’s state-of-the-art GB200 NVL72 Grace Blackwell Superchips via the NVIDIA DGX™ Quantum architecture.
The NVAQC addresses four of the most pressing challenges in quantum computing:
- Scaling Quantum Hardware: Developing the AI supercomputing infrastructure needed to control and operate large-scale QPUs
- Correcting Quantum Errors: Using GPU-accelerated decoding to make real-time quantum error correction feasible
- Simulating New QPU Designs: Running comprehensive hardware simulations to rapidly iterate on better qubit designs
- Developing Hybrid Algorithms: Colocating quantum and classical hardware so researchers can build and test truly hybrid applications
Current NVAQC partners include Quantinuum, QuEra, Quantum Machines, and EQuS (Engineering Quantum Systems at MIT).
Post-quantum cryptography (PQC) refers to encryption algorithms designed to withstand attacks from future quantum computers. Today’s widely used encryption standards (such as RSA and elliptic-curve cryptography) are potentially vulnerable to quantum algorithms like Shor’s algorithm, which could break them far more quickly than any classical computer.
Organizations must begin migrating to quantum-safe algorithms now—before fault-tolerant quantum computers arrive—to protect sensitive data that has long-term confidentiality requirements (a strategy sometimes called “harvest now, decrypt later”).
NVIDIA’s cuPQC library provides secure, GPU-accelerated implementations of the leading PQC algorithms endorsed by NIST, advancing data security against the quantum threat and enabling cryptographic research at scale.
NVIDIA Ising is a model family, training framework, and cookbook for building and deploying AI specifically for quantum computing. It is designed so that quantum computing experts can harness state-of-the-art AI technology without needing machine learning domain expertise—removing a significant barrier for QPU developers and operators who are deep in quantum physics but not necessarily in deep learning.
Ising addresses two of the hardest near-term engineering challenges on the road to fault-tolerant quantum computers:
Quantum Calibration (Ising Calibration)
Quantum processors require continuous, precise calibration—a labor-intensive process of tuning hardware to minimize noise and maintain qubit fidelity. Ising Calibration 1 is a vision-language model (VLM) that can be deployed inside an agentic workflow to automate quantum processor bring-up and recalibration. The agent evaluates experimental results from the QPU and adjusts parameters accordingly—dramatically reducing the manual effort required to keep a quantum computer operating at peak performance.
On the QCalEval benchmark (NVIDIA’s published evaluation suite for quantum calibration models), Ising Calibration 1 outperforms leading general-purpose frontier models: It scores 3.27% better than Gemini, 9.68% better than Claude Opus, and 14.5% better than GPT.
Quantum Error-Correction Decoding (Ising Decoding)
Real-time error correction is one of the most computationally demanding requirements for fault-tolerant quantum computers. Every time the QEC code cycle runs, a classical decoder must process the resulting syndrome data and identify errors—fast enough to keep pace with the QPU. Ising Decoder SurfaceCode 1 provides two variants tuned for different operational needs:
- Fast Variant: 2.5× lower latency and 1.1× higher accuracy than PyMatching (the industry-standard decoder) for surface code distance d =13 at error rate p = 0.003
- Accurate Variant: 2.3× lower latency and 1.5× higher accuracy than PyMatching under the same conditions
Critically, Ising Decoding also includes a training framework so QPU builders can fine-tune decoder models to the specific noise characteristics of their own hardware, rather than relying on generic models.
Getting Started
NVIDIA Ising models are available via NVIDIA NIM™ APIs (for cloud inference), Hugging Face (for download and fine-tuning), and GitHub (for training code and tutorials). The framework integrates with NVIDIA CUDA-Q QEC for real-time decoding workflows and uses cuQuantum cuStabilizer alongside PyTorch for model training.
NVIDIA offers multiple on-ramps for developers, researchers, and students:
- CUDA-Q: Available as open source software with documentation, tutorials, and a growing community
- CUDA-QX: Domain-specific libraries for QEC, hybrid solvers, and more
- cuQuantum: GPU-accelerated circuit simulation SDK for developers building quantum simulation frameworks, available via the NGC catalog
- NVIDIA Quantum Cloud: Early access to cloud-based quantum computing APIs for running CUDA-Q projects on NVIDIA GPU systems
- CUDA-Q Academic: Self-paced training for students and researchers; see FAQ above for full details
- Resources Hub: Blogs, GTC session recordings, whitepapers, and tutorials available at the NVIDIA Quantum Computing Resource Center