Nvidia CEO Jensen Huang said the company plans to sell twice as many chips next year as it does this year, making the announcement at an AI infrastructure summit.

Huang's chip sales outlook

Speaking at the AI Infrastructure Summit, Huang stated that Nvidia's chip volumes are on track to double in the coming year relative to the current year. The remarks highlight the scale of demand the company is experiencing for its semiconductor products amid accelerating AI infrastructure buildout globally.

NVIDIA Corporation (NASDAQ: NVDA) expanded its open-source CUDA-Q platform on Sept. 14, 2026, with the addition of CUDA-Q Logical. This new orchestration layer enables researchers to design and test fault-tolerant quantum computing applications programmatically.

The platform addresses the complexity of codesigning algorithms, error-correction codes, and hardware architectures for logical qubits. By allowing easy switching between system components, the tool aims to shorten timelines for useful quantum-GPU supercomputing.

Performance Gains at Fermilab

Fermi National Accelerator Laboratory (Fermilab) used CUDA-Q Logical to accelerate fault-tolerant architecture development. The lab reduced the timeline from five months to three weeks, achieving a 7x speedup.

Anna Grassellino, chief technology officer at Fermilab, noted that the team explored resource combinations in three weeks instead of the typical five months required for building specialized infrastructure.

Hardware Efficiency Improvements

Iceberg Quantum modeled its fault-tolerant architecture for Diraq’s qubits using the platform. The modeling showed that 1,000 logical qubits can be created with just 150,000 physical qubits.

This requirement is roughly 10x fewer physical qubits than Diraq’s previous estimates. The reduction in hardware needs helps lower the barrier for practical, fault-tolerant quantum computing.

New Benchmarking Standard

Sandia National Laboratories introduced QUOPS, a benchmark for measuring the readiness of fault-tolerant quantum hardware. QUOPS provides a hardware-agnostic method to track progress toward utility-scale applications.

Sandia shared early results in a preprint ahead of IEEE Quantum Week, reporting initial benchmarks for QPUs from Google, IBM, and Quantinuum. A reference implementation of QUOPS is now available in NVIDIA CUDA-Q.

Ecosystem Adoption

Several organizations are integrating their technologies with NVIDIA’s quantum-GPU supercomputing platform:

• Anyon Computing developed a new quantum control system using NVIDIA NVQLink.
• Quandela developed a QPU-GPU architecture.
• BlueQubit launched the Quantum Flywheel grant program to provide researcher access to NVIDIA accelerated computing.
• IonQ announced progress on DQAOA-GPT, a quantum generative AI framework.

What the Numbers Show

The divergence between Iceberg Quantum’s new estimate (150,000 physical qubits for 1,000 logical qubits) and Diraq’s prior estimates highlights a significant efficiency gain. This 10x reduction in physical qubit requirements suggests that software orchestration tools like CUDA-Q Logical can materially alter hardware roadmaps by optimizing resource allocation before fabrication.