AWS, NVIDIA Offer Deep Dive Into Their Partnership to Develop Hybrid Quantum Computing

Insider Brief

• AWS and NVIDIA have integrated the CUDA-Q quantum development platform into Amazon Braket, streamlining hybrid quantum-classical computing workflows for researchers.
• NVIDIA GPUs accelerate quantum circuit simulations, delivering up to 350x speed-ups over CPUs, enabling faster algorithm development and testing.
• Researchers can seamlessly transition from simulations to quantum hardware, using Braket’s pay-as-you-go model to access systems from IonQ, Rigetti and IQM.

AWS and NVIDIA are teaming up to address one of the biggest challenges in quantum computing: integrating classical computing into the quantum stack, according to an AWS Quantum Technologies blog post. This partnership brings NVIDIA’s open-source CUDA-Q quantum development platform to Amazon Braket, enabling researchers to design, simulate and execute hybrid quantum-classical algorithms more efficiently.

The Role of Hybrid Computing

Hybrid computing — where classical and quantum systems work together — is actually a facet of all quantum computing applications. Classical computers handle tasks like algorithm testing and error correction, while quantum computers tackle problems beyond classical reach. As quantum processors improve, the demand for classical computing power grows exponentially, especially for tasks like error mitigation and pre-processing.

The collaboration between AWS and NVIDIA is designed to ease this transition by providing researchers with seamless access to NVIDIA’s CUDA-Q platform directly within Amazon Braket. This integration allows users to test their programs using powerful GPUs, then execute the same programs on quantum hardware without extensive modifications.

Faster Simulations, Easier Workflows

Incorporating CUDA-Q into Amazon Braket brings significant performance gains, according to the team. One example is the speed-up offered by NVIDIA GPUs for quantum circuit simulations. Tests showed that simulating a 21-qubit algorithm on a GPU was 350 times faster than on a CPU. This kind of acceleration enables researchers to explore larger-scale quantum systems more efficiently.

The team writes that the integration of CUDA-Q with Amazon Braket is designed to help researchers focus on building algorithms rather than managing infrastructure. Ultimately, the team believes this could be a critical step forward in making hybrid quantum computing accessible.

Amazon Braket users can now leverage this integration to streamline their research. CUDA-Q programs can run on all Braket-supported quantum hardware, including systems from IonQ, Rigetti and IQM, by changing just one line of code.

Preparing for the Quantum Future

Quantum error correction — arguably the most important technique required for scalable quantum computing — relies heavily on classical co-processing. As quantum systems evolve to handle more qubits and deeper circuits, researchers will require not only faster hardware but also more sophisticated tools to manage the interplay between quantum and classical systems.

AWS and NVIDIA’s collaboration lays the groundwork for such advancements. The partnership aims to develop a hybrid quantum-computing infrastructure that blends classical computing with quantum systems, offering flexibility and scalability for cutting-edge workloads.

The team emphasized the importance of forward-looking infrastructure, writing: As quantum computing technologies mature, state-of-the-art workloads will have diverse and increasingly demanding requirements for the associated classical compute resources: from ultra-low latency co-processing for quantum error correction (QEC) decoding and feedback and supercomputing-scale classical pre- and post-processing to quantum hardware control and calibration, and AI-enabled quantum simulations. To address these challenges, AWS is working with NVIDIA to evaluate the latency and compute requirements of future workloads, as well as develop a quantum computing stack that gives customers the performance and flexibility needed to get the most out of emerging quantum computing technologies.”

Benefits for Researchers

The integration allows researchers to:

• Accelerate Simulations: NVIDIA GPUs significantly reduce simulation times, enabling faster testing of quantum algorithms.
• Simplify Hybrid Workflows: With CUDA-Q on Braket, researchers can seamlessly transition from simulations to quantum hardware execution.
• Focus on Research: Managed access to GPUs and quantum devices removes the need for researchers to handle complex infrastructure.

Additionally, AWS provides a pay-as-you-go model, ensuring cost efficiency for researchers working on demanding computational problems.

A Glimpse into the Results

Performance tests conducted by AWS showed the CUDA-Q platform outperforming other open-source simulators for a 29-qubit algorithm, running significantly faster on a single GPU instance. The flexibility of Braket Hybrid Jobs also allows researchers to scale workloads across multiple GPUs, achieving even greater speed-ups for certain tasks.

For example, parallelizing a 30-qubit circuit across eight GPUs led to a 6.5x speedup in runtime, while distributing 128 different circuits yielded speedups of up to 17x for parameterized circuits. These gains are critical as researchers push the boundaries of quantum computing.

Future Directions

As the collaboration between AWS and NVIDIA deepens, their focus will include evaluating the latency and computational requirements for emerging quantum workloads. The ultimate goal is to create an integrated quantum-classical computing ecosystem that supports researchers at every stage, from algorithm design to execution on cutting-edge quantum hardware.

The researchers describe this integration as just the beginning: “This release is a first step to a broader collaboration between AWS and NVIDIA to explore the quantum stack as quantum and classical become inextricably linked over time.”

For researchers interested in exploring these tools, AWS provides detailed example notebooks to get started. You can check out the example notebooks in the team’s GitHub repo.