NVIDIA announced at GTC 2025 the release of NVIDIA Holoscan 3.0, the real-time AI sensor processing platform. This latest version provides dynamic flow control, empowering developers to design more robust, scalable, and efficient systems. With physical AI rapidly evolving, Holoscan 3.0 is built to adapt, making it easier than ever to tackle the challenges of today’s dynamic environments.
Whether you’re working with medical imaging, robotics, or industrial sensors, Holoscan 3.0 makes it easier to process AI workloads efficiently at the edge. This post explains the updates and features in the Holoscan 3.0 release, including:
- Application framework: NVIDIA Holoscan SDK
- Standardized solution for sensor IO: NVIDIA Holoscan Sensor Bridge
- Deployment: NVIDIA IGX Orin 500 support
NVIDIA Holoscan SDK??
NVIDIA Holoscan 3.0 introduces a major upgrade to the NVIDIA Holoscan SDK, with dynamic flow control. With this feature, you can now modify operator connections within a pipeline at runtime, enabling more flexible and adaptive workflows. Unlike traditional static pipelines with fixed connections, this feature supports dynamic routing, conditional branching, and iterative processing. Adjust data flow based on execution conditions, ensuring efficient handling of complex workflows.
Key use cases include conditional processing, adaptive workflow routing, iterative operations that evolve dynamically during execution, and robust error handling with recovery mechanisms. Say goodbye to manual control calls and complex code—Holoscan 3.0 does the heavy lifting for you with intuitive APIs.
In addition, the Holoscan SDK package is now available on Conda Forge for Python, making it easier than ever to install and manage dependencies, accelerating the development process.
Holoscan Sensor Bridge for streamlined sensor IO
Holoscan Sensor Bridge (HSB) simplifies one of the toughest parts of building and integrating any sensor: connecting non-PCIe or Ethernet sensors and integrating them into the NVIDIA Holoscan platform. HSB makes it easier to connect sensors and manage data streams effectively, saving time and reducing the complexity of the development process.
You can now create a standardized IO system over Ethernet with enhanced reliability, better detect errors, monitor and analyze data as it travels across networks, and align with safety compliance documentation. HSB helps accelerate the process of integrating sensor data into applications and ensures that data transfer remains secure and reliable, which is crucial for meeting regulatory standards.
Full support for NVIDIA IGX Orin 500
Holoscan 3.0 now fully supports NVIDIA IGX Orin 500, the hardware of choice when working on edge AI applications that require real-time processing, low-latency video, and high-performance AI inference. Whether building advanced surgical assistance, autonomous vehicles, or industrial automation systems, the IGX Orin 500 provides the computational power, form factor and energy efficiency, and scalability necessary for cutting-edge applications. By choosing the IGX Orin 500, you can create highly reliable, production-ready solutions that are ready for the demands of real-world deployments.
New reference application for end-to-end surgical video
As part of this release, NVIDIA is unveiling a new reference application designed to streamline the development of end-to-end surgical video solutions. This application provides a comprehensive developer toolset, including:
- High-speed data ingestion: Low-latency surgical video feed input for real-time processing.
- DeID: Ensures patient privacy by anonymizing identifiable data.
- Multi-AI inference: Supports simultaneous AI models for tool and organ segmentation, motion tracking, and 3D reconstruction to assist with precise navigation.
- Real-time visualization: Outputs processed video and insights, improving surgeon situational awareness.
This reference application accelerates sensor data integration and simplifies the development of edge AI applications for surgical use cases. Save time, ensure compliance, and leverage powerful features to enhance your solutions.
Get started with NVIDIA Holoscan 3.0
Physical AI is rapidly reshaping many industries. NVIDIA Holoscan 3.0 provides developers with a powerful platform for building real-time, AI-powered applications that seamlessly integrate with physical devices, enabling faster prototyping, simulation, and deployment of complex AI-driven systems in robotics, healthcare and beyond.
Ready to get started? Download NVIDIA Holoscan 3.0, read the NVIDIA Holoscan 3.0 Documentation, and check out the Holoscan Reference Applications.
