What Is Quantum Computing?

Quantum processors control and manipulate components able to act as qubits (quantum bits). Unlike the classical bits realized by transistors in conventional processors—which can exist in one of two states, denoted 0 or 1—qubits allow information to be represented in ways offering much greater computation power. For example, qubits can exist in a superposition of both 0 and 1 states simultaneously during a quantum computation. For certain application areas, this can be leveraged to allow a kind of quantum parallelism within computations. Crucially, harnessing superposition in an ultimately useful way is nuanced and only known to be possible for certain classes of algorithms. 

A quantum processing unit (QPU)  is a device designed to isolate and manipulate qubits. Early in the history of conventional computing, the transistor was settled on as the most effective way to realize classical bits. However, today in quantum computing, there are still many candidates for how to physically build qubits. These different approaches to quantum computing include superconducting qubits, trapped ions, neutral atoms, and photonic systems. It is not yet clear whether one of these approaches will prevail for building useful quantum-GPU supercomputers, or whether multiple approaches will work in tandem.

Each approach faces its own technological challenges — superconducting qubits operate near absolute zero in dilution refrigerators; ion trap systems require an ultra-high vacuum. But all approaches also have challenges in common — qubits are inherently fragile, and environmental interactions can quickly cause computation-ending decoherence. Quantum error correction techniques to mitigate this kind of noise are critical for all kinds of qubits and present a central hardware engineering challenge in quantum computing.

Qubits will not replace bits but will work alongside them. All useful quantum applications are inherently hybrid: Conventional supercomputers orchestrate the operation of QPUs within larger hybrid workflows. More fundamentally, conventional supercomputers are also critical to even operate quantum operations, enabling task compilation, calibration, error correction, and post-processing. NVIDIA CUDA-Q™ is the unified programming model for hybrid quantum-classical computers, enabling developers to program QPUs, GPUs, and CPUs within a single system, seamlessly passing information between them and opening quantum development to both quantum researchers and other domain experts.

The inherent noise in quantum systems makes it impossible to run more than a handful of quantum operations on qubits before computations must halt. Historically, there have been two approaches to deal with this noise:

• NISQ (Noisy Intermediate-Scale Quantum) denotes an era of applications development aimed at finding uses for noisy quantum processors capable of returning useful results while running QPUs for only very short periods before noise destroys computations. The existence of useful NISQ applications remains an active area of research.
• Fault-Tolerant Quantum Computers (FTQCs) use quantum error correction (QEC) techniques to continually detect and correct errors occurring due to quantum noise—aiming to extend the computational lifetime of qubits indefinitely. QEC codes deal with noise by introducing a large overhead of qubit redundancy and complex classical algorithms (called decoders) to infer error locations. Through this process, many noisy physical qubits are distilled into a handful of reliable logical qubits. FTQC is widely considered to be the path to utility-scale quantum computing.

The decoding algorithms required to run QEC demand high-performance classical computing tightly coupled to the QPU. NVIDIA NVQLink™ is a reference architecture providing the low-latency, high-throughput connection needed, while NVIDIA CUDA-Q QEC provides out-of-the-box GPU-accelerated decoding algorithms, and NVIDIA Ising Decoding allows researchers to use AI to accelerate QEC.

Quantum processors will operate as accelerators inside data centers, working alongside the large-scale classical systems they depend on. This reciprocal relationship is called quantum-accelerated supercomputing. NVIDIA NVQLink is an open architecture that connects all QPU modalities to AI supercomputers with low latency and high throughput—providing the missing interconnect for QPU builders, quantum controller vendors, and supercomputing centers to scale quantum computers toward utility-scale applications.

The NVIDIA Accelerated Quantum Computing Research Center (NVAQC) integrates partner quantum hardware with the NVIDIA GB200 NVL72 system, giving developers and academics the infrastructure to drive the breakthroughs needed to scale quantum hardware and applications. Supercomputing centers worldwide—including ABCI-Q in Japan, Pawsey in Australia, and the Novo Nordisk Foundation Centre in Denmark—are already deploying QPUs alongside NVIDIA GPU systems.

Quantum programs and quantum software are typically written using familiar languages like Python or C++ within a quantum development framework. CUDA-Q allows developers to compose hybrid quantum-classical programs—specifying quantum gates and logic gates, high-level kernels, or variational circuits—and compile them to target a wide range of QPU devices and simulators. The NVIDIA cuQuantum SDK accelerates quantum circuit simulation on GPUs and supports, supporting frameworks including Cirq, Qiskit, and PennyLane.

AI and quantum computing are mutually reinforcing. In the near term, AI is already accelerating quantum hardware development and operation—a field called AI for Quantum. NVIDIA Ising is the world’s first family of open AI models specifically designed to accelerate the path to useful quantum computers, delivering breakthrough performance in quantum hardware calibration and error correction. Looking further ahead, quantum computing may, in turn, accelerate AI through quantum machine learning (QML), though those applications await later-generation hardware.

As quantum computers scale, algorithms like Shor's algorithm, developed by Peter Shor, will eventually be able to break widely used encryption standards like RSA and ECC. Post-quantum cryptography (PQC) refers to encryption algorithms designed to withstand quantum attacks. NVIDIA cuPQC provides secure, GPU-accelerated implementations of the leading NIST-standardized PQC algorithms, enabling organizations to migrate to quantum-safe security now—before large-scale quantum computers arrive.