“Really, it’s QIR that carries the weight in this ecosystem,” says Sanders. “But QIR doesn’t have a marketing team, and since its implementation is stable and feature-complete, it doesn’t need one.”.

Figure 1. CUDA-Q Brings Dual-Rail Qubit Support to Quantum Circuits.

CUDA-Q is already supported by IonQ, QuEra, Quantinuum, Rigetti, and several other quantum hardware vendors. Nvidia states that the platform currently integrates with 75% of publicly accessible quantum processing units. Figure 1 shows CUDA-Q Brings Dual-Rail Qubit Support to Quantum Circuits.

At the end of 2024, Nvidia added support for AWS Braket. CUDA-Q is also accessible through Nvidia’s own Quantum Cloud platform, which merges GPUs and quantum processors with AI.

The other leading quantum computing platform is IBM’s Qiskit, which Quantum Circuits already supports. Qiskit works with IBM, IonQ, Rigetti, Alice & Bob, and Quantinuum, and is available via Amazon Braket, Microsoft Azure Quantum, and the IBM Quantum Platform.

“Qiskit is a strong platform and has been around far longer than CUDA-Q,” says Andrei Petrenko, head of product at Quantum Circuits. But unlike CUDA-Q, Qiskit is limited to Python rather than C++, making CUDA-Q more suitable for high-performance computing. “Qiskit doesn’t have the DNA to leverage AI with GPUs.”

Petrenko adds that Qiskit is mainly designed for standalone quantum programming and may incorporate classical CPU tasks, but not GPU-based workflows. This could change with IBM’s recent collaboration with AMD.

A New Spin on Quantum Hardware

Quantum Circuits’ dual-rail chip combines two distinct quantum computing techniques: superconducting resonators and transmon qubits. The qubit itself is a photon, controlled by a superconducting circuit. “It matches the reliability benchmarks of ions and neutral atoms while delivering the speed of a superconducting platform,” says Petrenko.

Another standout feature is real-time error awareness. “No other quantum computer tells you immediately if it encounters an error, but ours does,” he explains. This opens the possibility of correcting errors before scaling up, rather than addressing them after increasing qubit counts.

In the near term, this combination of high reliability and built-in error correction makes the platform a powerful tool for developing new algorithms. “It lets us explore new problems, and we’ve already leveraged it to demonstrate novel approaches in machine learning,” says Petrenko.

TechInsights’ Sanders confirms that this dual-rail method is distinct from most quantum computing approaches. By combining the strengths of different qubit types, it extends coherence times and integrates error correction more effectively.

“We’re in alpha access right now, working with select partners,” says Petrenko. “This allows us to understand the problem types our partners are tackling and demonstrate our unique capabilities.”

Although eight qubits is far fewer than industry leaders, scaling up more qubits typically increases errors rapidly. “The quantum computing industry can’t rely on a brute-force approach,” says Sanders. “It needs alternative qubit designs.” This creates a window of opportunity for companies experimenting with novel methods, like Quantum Circuits.

“The approach shows promise, but the industry still has significant work to do before realizing a fully general-purpose quantum computer,” Sanders adds.

Source: NETWORK WORLD

Cite this article:

Priyadharshini S (2025), Nvidia’s CUDA-Q Platform Now Supports Dual-Rail Qubits in Quantum Circuits, AnaTechMaz, pp.172