Amazon Braket Notebook Instances Now Natively Support NVIDIA CUDA-Q, Revolutionizing Quantum Development on AWS
AWS stated that Amazon Braket Notebook settings now natively support CUDA-Q, NVIDIA’s open-source hybrid quantum-classical computing architecture, in a key quantum computing platform update. Making NVIDIA’s technology more accessible simplifies hybrid quantum-classical system developers’ and scholars’ workflows.
You can also read What Is Amazon Braket? How Does It Work And Advantages
Customers can now run CUDA-Q programs directly in Amazon Braket Jupyter notebook instances without the need for extra initial setup with the integration. Upgrading the notebook instances’ base operating system to Amazon Linux 2023, which offers more security, better performance, and higher compatibility—all essential for contemporary quantum development workflows—enables native support.
You can also read Top Swiss Startups 2025: Quantum QFX, Y-Quantum And Zuriq
Development of a Smooth Hybrid Workflow
Hybrid quantum-classical algorithms can now be easily developed and tested by quantum researchers and developers with this native capability. The platform makes it simple for users to create, model, and execute these hybrid algorithms in the Braket-managed notebooks.
The native CUDA-Q support offers the following important features:
- Development of Hybrid Workflows: Users can combine GPU-accelerated calculations with quantum simulations.
- GPU Acceleration: By making NVIDIA GPUs accessible through Amazon Braket, developers may create algorithms and simulate more quickly.
- Access to Quantum Hardware: The platform enables users to easily move their work from simulation to the actual quantum hardware supported by Braket. This contains Quantum Processing Units (QPUs) that are available in a single managed environment and are offered by IonQ, Rigetti, and IQM.
Amazon Braket’s standing as a comprehensive platform for quantum computing research and development is strengthened by this latest release. To utilize CUDA-Q within the managed notebook environment, it streamlines processes that previously required developers to control local deployment or route execution via Hybrid Jobs.
You can also read Amazon Braket Program Sets 24x Quantum Workload Speed
Accessing the New Environment and Pre-Installed Packages
Customers have access to the most recent Braket kernel when they establish a new Amazon Braket notebook instance (NBI). You may see the new conda_braket tile when you open the notebook. By choosing this tile under “Notebook” or “Console,” the new conda_braket kernel is used to launch a Python console session or Jupyter notebook.
The conda_braket kernel is now the default kernel linked to the “CUDA-Q and Braket” tile, which opens example notebooks.
The most recent compatible packages for the four top quantum development frameworks—Braket, CUDA-Q, PennyLane, and Qiskit—are pre-installed in the Braket notebook environment. By using a particular command, developers can confirm the precise installed packages.
Several important package versions are confirmed to be present in the output, including:
• amazon-braket-algorithm-library==1.6.2
• amazon-braket-default-simulator==1.31.4
• amazon-braket-pennylane-plugin==1.33.5
• amazon-braket-schemas==1.26.1
• amazon-braket-sdk==1.102.6
• cudaq==0.12.0.post1
• cudaq-qec==0.4.0.post1
• cudaq-solvers==0.4.0
• PennyLane==0.42.3
• PennyLane_Lightning==0.42.0
• qiskit==1.4.4
• qiskit-aer==0.17.2
• qiskit-algorithms==0.4.0
• qiskit-ionq==0.6.1
• qiskit_braket_provider==0.6.0
Examples of quantum applications using all four libraries—Braket, CUDA-Q, PennyLane, and Qiskit—are pre-installed on Amazon Braket notebook instances. Clicking on the “CUDA-Q and Braket” tile on the Launcher page allows users to view CUDA-Q examples. This opens the example notebook 0_hello_cudaq_jobs.ipynb and displays the nvidia_cuda_q/ directory with further examples.
You can also read Amazon Braket SDK Architecture And Components Explained
Implementation and Example Suggestions
AWS advises utilizing tiny, CPU-based instance types like ml.t3.medium for typical laptop instances. However, Amazon Braket Hybrid Jobs can still be used to run CUDA-Q programs that need intensive GPU support. GPU-powered instances, such ml.p3.8xlarge, are advised for these demanding tasks. Among the GPU-accelerated instances that are available are:
ml.p3.2xlarge(featuring 1 NVIDIA V100 GPU).ml.p3.8xlarge(featuring 4 V100 GPUs).ml.p3.16xlarge(featuring 8 V100 GPUs).
A customer’s Amazon Elastic Compute Cloud (EC2) service quota should be greater than or equal to the number of instances they want to run when using GPU-powered instances concurrently through Hybrid Jobs.
Background: The Previous Docker-Based Setup
Before this native integration, developers needed to set up a Jupyter kernel running a CUDA-Q Docker container in order to run CUDA-Q apps interactively in Braket notebooks. Utilising the open-source NVIDIA CUDA-Q platform and Amazon Braket Hybrid Jobs required this approach, which allowed users to run CUDA-Q’s simulators on Amazon Braket-supported quantum hardware backends and powerful NVIDIA GPUs.
Docker is pre-installed on Braket Notebook instances, enabling development in a controlled environment using Docker images found on the NVIDIA NGC Container Registry.
To configure a custom kernel in the prior setup, a few particular procedures were needed:
- Dockerfile creation: A stable release A Dockerfile needs a CUDA-Q image like nvcr.io/nvidia/quantum/cuda-quantum:cu12-0.9.1. This file provided instructions for installing ipython, ipykernel, and amazon-braket-sdk.
- Image Building: Next, the image was constructed, which could take three to five minutes.
- Kernel Configuration: For the new kernel (such as docker_cudaq), a kernel.json file was produced that contained command-line parameters for starting the kernel with the freshly constructed Docker image.
- GPU Access: Users required to add “–gpus=all” to the kernel in order to access GPU instances (such as ml.p3.2xlarge, ml.p3.8xlarge, or ml.p3.16xlarge) in the CUDA-Q program.configuration in JSON.
The preceding configuration can mount a directory like /home/ec2-user/amazon-braket-examples/examples into the CUDA-Q container. Developers did not need to alter the Jupyter kernel specification if they required more Python modules or updated CUDA-Q versions; they only needed to make changes to the Dockerfile and rebuild the Docker image.
This intricate setup is now mostly avoided for standard development in the controlled environment with native support.
You can also read IonQ and UChicago Announce Strategic Partnership for Quantum
