← Back to feed

NVIDIA’s bet on quantum computing does not involve building qubits. It involves using AI to make everyone else’s qubits work. Ising, released April 14 and widely available as of this week, is the first family of open AI models designed specifically to solve the two hardest engineering problems in quantum hardware: calibration and error correction.

Both problems have historically required manual tuning by specialists. Ising automates them.

What Ising Actually Is

The family has two distinct components with different architectures targeting different problems.

Ising Calibration is a vision-language model. Quantum processors generate measurement data from physical hardware — optical readouts, qubit state diagrams — that currently require human experts to interpret and act on. Ising Calibration reads those measurements and autonomously tunes the processor, reducing calibration time from hours to minutes in early deployments.

Ising Decoding is a 3D convolutional neural network built for real-time quantum error correction. Quantum errors are probabilistic and must be corrected faster than they compound. Ising Decoding performs that correction in real time with 2.5x the throughput and 3x the accuracy of pyMatching, the current open-source standard for quantum error correction.

Both models are available on GitHub, Hugging Face, and NVIDIA’s build.nvidia.com, and integrate with the CUDA-Q hybrid quantum-classical software platform and the NVQLink QPU-GPU hardware interconnect.

Day-One Adoption Is Unusually Broad

Ising Calibration is already deployed at Atom Computing, Academia Sinica, EeroQ, Conductor Quantum, Fermi National Accelerator Laboratory, Cornell University, UC San Diego, UC Santa Barbara, the University of Chicago, the University of Southern California, and Yonsei University.

That list spans national labs, Ivy League institutions, and commercial quantum hardware companies across multiple qubit modalities — including superconducting, trapped ion, and neutral atom architectures. Shipping a model family with that breadth of day-one adoption is a meaningful signal about where NVIDIA positioned this technology before the public launch.

The Infrastructure Logic

NVIDIA’s framing is deliberate: Ising is not a quantum chip play, it is an infrastructure play. The company’s competitive position in AI came from owning the software and hardware stack that made GPU clusters useful. The Ising strategy follows the same logic applied to a different layer — rather than competing with IonQ, Quantinuum, or IBM at the hardware level, NVIDIA is building the tooling that makes all of those platforms easier to operate and scale.

That positioning drove an immediate market reaction. Following the announcement, QUBT gained 29.85% over three consecutive trading sessions. IONQ surged 50.13%. Both moves reflect investor recognition that NVIDIA entering quantum infrastructure at the software level accelerates the entire sector’s timeline.

Key Numbers

• Ising Decoding speed: 2.5x faster than pyMatching
• Ising Decoding accuracy: 3x better than pyMatching
• Ising Calibration architecture: vision-language model (VLM)
• Ising Decoding architecture: 3D convolutional neural network
• License: open source
• Availability: GitHub, Hugging Face, build.nvidia.com
• Integrations: NVIDIA CUDA-Q, NVQLink QPU-GPU interconnect
• Market impact: QUBT +30%, IONQ +50% (3-day window post-announcement)