Diraq and QM Technologies Inc. have achieved a breakthrough in quantum computing by integrating silicon quantum processors with NVIDIA’s DGX Quantum, enabling real-time communication at 3.3 microseconds. This collaboration, leveraging AI and GPU acceleration, addresses critical scaling challenges for quantum systems.
Diraq and QM integrate NVIDIA DGX Quantum with silicon quantum processors.
Achieves 3.3-microsecond real-time communication with Grace-Hopper superchips.
Demonstrates real-time readout, AI calibration, and fast state initialization.
Results achieved within one week of system installation in Sydney.
Supports future real-time error correction and CUDA-Q integration.
Showcased at GTC Paris 2025 for hybrid quantum-classical advancements.
Diraq and QM deployed NVIDIA’s DGX Quantum architecture, combining NVIDIA Grace-Hopper superchips with QM’s OPX1000 controller, to achieve a 3.3-microsecond round-trip latency. “The importance of integrating classical computing power with future quantum computers can’t be understated,” said Andrew Dzurak, CEO and Founder of Diraq. Within one week, the team implemented three applications: real-time correlated readout, machine learning-driven calibration, and accelerated quantum state initialization, overcoming traditional quantum bottlenecks.
The integration addresses key quantum scaling issues. Real-time correlated readout, previously limited to post-processing, now leverages GPU acceleration for complex signal processing beyond FPGA capabilities. Automated calibration using machine learning reduced hours of manual tuning, while GPU-accelerated state initialization, based on Diraq’s Nature publication, ensures rapid quantum state preparation. “Diraq’s success in developing three distinct applications within a week proves that quantum-AI integration has moved beyond the hype stage,” said Itamar Sivan, CEO of QM.
The 3.3-microsecond latency enables real-time feedback loops critical for maintaining quantum coherence. “This collaboration exemplifies the future of computing, with GPU and QPU seamlessly integrated,” said Tim Costa, senior director of quantum and CUDA-X at NVIDIA. This low latency supports practical quantum computing by allowing commands and responses before quantum information degrades.
DGX Quantum’s integration with NVIDIA’s CUDA-Q platform positions it for future advancements, including real-time error correction. Compatible with Diraq’s CMOS-based silicon quantum dot technology, the system leverages existing semiconductor manufacturing for scalability. Diraq aims to scale qubit counts to millions, targeting applications in pharmaceuticals, finance, and energy. The results will be highlighted at GTC Paris 2025 by NVIDIA’s Elica Kyoseva.
Serving research institutions and quantum developers, Diraq and QM’s solution accelerates hybrid quantum-classical systems. With over 50 patents and $135M in funding, Diraq’s silicon-based approach, combined with QM’s control systems and NVIDIA’s AI infrastructure, sets a new standard for scalable, cost-effective quantum computing, driving innovation across industries.
Diraq is a global leader in building quantum processors using silicon 'quantum dot' technology, leveraging proprietary technology developed over 20 years of research. Diraq is a private company, founded in 2022 and headquartered in Sydney, Australia, with operations in Palo Alto, California, and Boston, Massachusetts. Our approach relies on the existing silicon manufacturing processes, known as CMOS, used by foundries to produce today's semiconductor components. By capitalizing on existing high-volume chip fabrication technology and semiconductor manufacturing capabilities, Diraq is accelerating the change that can transform computing as we know it. Diraq's platform architecture is purpose-built to drive the significant processing advances required to reduce cost and energy barriers, and to realize quantum computing's full societal and economic potential, forging a faster and cheaper road to market. Diraq's goal is to revolutionize quantum computing by driving qubit numbers on a single chip to the many millions, and ultimately billions needed for useful commercial applications.