Resources
Your NVIDIA® Jetson Xavier™ Platform Needs a Migration Plan. CTai LABS Can Help You Build It.
Move your Edge ai application to NVIDIA Jetson Orin™ or Jetson Thor™ with a migration strategy that covers hardware, software, models, sensors, performance, and deployment.
By Ceri Nelmes
Tech Marketing Leader & Journalist | Building Brands at the Edge of What’s Next | Connect Tech & CTai Labs
Key Takeaways
- The July 1, 2026 last-buy date identified by Connect Tech for affected NVIDIA Jetson Xavier, Jetson™ TX2i, Jetson TX2 NX, and Jetson Nano platforms has passed.
- Migration involves more than selecting a replacement module. Operating systems, NVIDIA JetPack™, CUDA®, TensorRT™, camera interfaces, drivers, containers, models, application dependencies, power, thermal design, and mechanical integration may all be affected.
- CTai LABS can begin at any point, from an initial migration question or undocumented legacy system to an active port, failing build, performance problem, or production-validation program.
- Connect Tech has mapped supported carrier boards and systems to potential NVIDIA Jetson Orin and Jetson Thor migration targets.
- CTai LABS turns that hardware path into an application-level migration plan, then supports implementation, integration, benchmarking, validation, and deployment.
- The customer retains its application, models, data, system design, and intellectual property.
NVIDIA® Jetson Xavier™ platforms helped establish Edge ai across robotics, industrial automation, medical devices, vision systems, autonomous machines, and other embedded applications. As those platforms approach or reach lifecycle limits, organizations need to decide how to preserve working systems while moving toward a supported compute and software foundation.
CTai LABS helps engineering and product teams manage that transition from architecture through deployment. We assess the complete system, identify dependencies that could interrupt the migration, select the appropriate Connect Tech hardware path, port and optimize the software stack, and validate the resulting application on its target platform.
The objective is not simply to make the application run on new hardware. It is to establish a migration path that protects essential functionality, meets performance requirements, addresses deployment constraints, and creates a maintainable foundation for the next stage of the product.
Download the NVIDIA Jetson Migration Guide
Why NVIDIA Jetson Xavier Migration Requires More Than a Module Swap
A module change can affect nearly every layer of an Edge ai system.
The source and target platforms may use different versions of Ubuntu, Linux for Tegra, NVIDIA JetPack™, CUDA®, TensorRT™, cuDNN, NVIDIA DeepStream SDK, NVIDIA VPI, camera drivers, container images, and third-party libraries. Changes in compute architecture, power profiles, available interfaces, memory, storage, thermals, and physical dimensions can also affect the system.
NVIDIA Jetson AGX Xavier and Jetson Xavier NX are limited to the JetPack 5 branch. NVIDIA has announced that JetPack 5 is scheduled to reach end of life in Q3 2026 and recommends beginning migration, validation, and production-readiness planning as early as possible. NVIDIA JetPack 5 EOL notice
Current NVIDIA Jetson Orin platforms support newer software foundations. NVIDIA JetPack 6.2.1 introduced an Ubuntu 22.04-based root file system, Linux kernel 5.15, and updated NVIDIA ai libraries. NVIDIA JetPack 7.2.1 now provides an Ubuntu 24.04 and Linux kernel 6.8 foundation across supported Jetson Orin™ and Jetson Thor™ platforms. Exact module, feature, peripheral, and software compatibility must be confirmed against the release selected for the project. NVIDIA JetPack documentation and current JetPack releases.
These changes can create practical engineering questions:
- Will the existing model execute correctly under the target TensorRT version?
- Are custom CUDA kernels compatible with the new compute architecture and toolchain?
- Will the camera, serializer, sensor, or capture pipeline work with the target carrier and board support package?
- Can existing containers be rebuilt for the new operating-system and library baseline?
- Does the new platform preserve required latency, throughput, memory use, and power consumption?
- Will ROS or ROS 2 nodes, NVIDIA DeepStream pipelines, device drivers, and middleware behave consistently?
- Does the target system meet the original environmental, mechanical, power, security, and lifecycle requirements?
- Can the transition be staged without interrupting active deployments?
CTai LABS addresses these questions as one connected migration program.
Start at Any Point in the Migration
CTai LABS can begin wherever your program is today.
You may have an early-stage requirement and need help selecting the target platform. You may have a deployed Jetson Xavier system with limited documentation. Your team may already have purchased Jetson Orin hardware but be encountering build failures, driver issues, degraded inference performance, or peripheral incompatibilities. You may also need an independent validation program before releasing the updated system into production.
Useful starting materials can include:
- Current NVIDIA Jetson module and JetPack version
- Connect Tech carrier board or Edge system part number
- Source code, containers, build scripts, or software images
- ai models and representative datasets
- Camera, sensor, networking, and storage requirements
- Target latency, throughput, accuracy, power, and thermal limits
- Mechanical drawings, cable requirements, and enclosure constraints
- Known software defects or migration blockers
- Deployment environment and applicable validation requirements
- Production schedule and lifecycle expectations
If some of these inputs do not yet exist, CTai LABS can help define them as part of the engagement.
A Structured Path from Xavier to the Next Platform
1.
Establish the Existing System Baseline
Migration begins with an accurate record of what is running today.
CTai LABS inventories the compute module, Connect Tech hardware, operating system, JetPack and L4T versions, ai frameworks, model formats, libraries, drivers, interfaces, sensors, containers, application services, and deployment procedures.
We also identify undocumented dependencies that may only become visible during the move, such as pinned packages, custom kernel changes, proprietary camera drivers, hard-coded device paths, legacy Python environments, or application assumptions tied to the original platform.
2.
Define the Target Requirements
The most appropriate migration target depends on the application, not only the source module.
CTai LABS works with the customer to define required inference performance, sensor bandwidth, memory, storage, networking, power, thermal limits, environmental specifications, mechanical constraints, software lifecycle, and future expansion.
This separates requirements that must be preserved from capabilities that can be improved during migration.
3.
Select the Connect Tech Hardware Path
Connect Tech’s migration guide maps Jetson TX2i, TX2 NX, Nano, Jetson Xavier NX, and Jetson AGX Xavier platforms to supported Jetson Orin and Jetson Thor options. The guide includes carrier-board part numbers and considerations related to form factor, interfaces, power, software compatibility, and platform architecture. Connect Tech Jetson Migration Guide
Depending on the existing design and new requirements, the hardware path may include:
- A compatible module transition on an existing Connect Tech carrier
- A migration to a Connect Tech carrier board for NVIDIA Jetson Orin
- A rugged Connect Tech Edge system for industrial, transportation, aerospace, defense, or autonomous applications
- A Jetson Thor-ready Connect Tech platform for workloads requiring greater compute headroom or a new architecture
- A custom Connect Tech solution when standard platforms do not satisfy the required interfaces, mechanics, power, or environment
Hardware selection is completed alongside application analysis so that compute, I/O, camera connectivity, networking, storage, power, and thermal requirements remain aligned.
4.
Port the Software Stack
CTai LABS develops the target software baseline and moves the application in controlled layers.
The work can include:
- Board support package configuration
- Operating-system and JetPack migration
- CUDA and TensorRT updates
- Model conversion and engine regeneration
- NVIDIA DeepStream or GStreamer pipeline migration
- ROS or ROS 2 integration
- Camera and sensor bring-up
- Container rebuilding
- Python and C++ dependency updates
- Kernel-module and device-driver integration
- Boot, storage, networking, and update configuration
- Logging, monitoring, and diagnostic improvements
For some Jetson Xavier-to-Orin programs, a shared JetPack 5 baseline may support a staged approach: establish the application on the new hardware first, then move to a newer JetPack branch. Other systems require a more direct software transition. Jetson Thor programs should be treated as a broader architectural migration because the target platform introduces a newer software and compute foundation.
5.
Validate Models, Pipelines, and Interfaces
A successful build does not prove that the migrated system is ready for deployment.
CTai LABS tests the complete application on target hardware using representative models, data, sensors, and operating conditions. Validation can cover:
- Model output and functional equivalence
- End-to-end latency
- Inference throughput
- CPU, GPU, accelerator, and memory utilization
- Camera and sensor synchronization
- Dropped frames and buffer behaviour
- Startup, shutdown, and recovery
- Storage and network performance
- Power consumption and thermal behaviour
- Long-duration stability
- Fault handling and diagnostic visibility
Results are compared with the original system and the target requirements. Where the new platform exposes bottlenecks or opportunities, CTai LABS can optimize preprocessing, inference, memory movement, scheduling, containers, and application logic.
6.
Prepare the System for Deployment
The final phase turns the migrated application into a repeatable deployment baseline.
Deliverables may include:
- Target hardware and software architecture
- Migration dependency assessment
- Configured software image
- Ported application and ai pipeline
- Updated models or TensorRT engines
- Driver and peripheral configuration
- Benchmark and validation results
- Build and deployment instructions
- Configuration records
- Known limitations and risk register
- Production handoff documentation
- Recommended next-stage validation activities
The customer retains ownership of its application, models, data, device design, and intellectual property.
Choose Between Jetson Orin and Jetson Thor Based on the Workload
NVIDIA Jetson Orin is a practical migration target for many existing Jetson Xavier applications. It supports current Edge ai, robotics, vision, and autonomous-system workloads across multiple performance and power levels, with a broad Connect Tech ecosystem of carrier boards and rugged systems.
NVIDIA Jetson Thor may be appropriate when the program requires greater compute headroom, larger or more complex models, advanced robotics workloads, expanded sensor processing, or a longer architectural runway. It should not be treated as an automatic replacement solely because it is newer. Its software baseline, system requirements, power, thermal profile, interfaces, and application needs must be evaluated as part of the complete design.
CTai LABS helps determine which platform fits the actual workload and deployment constraints, then validates that decision on the intended Connect Tech hardware.
Common Migration Engagements
Recover and Migrate a Deployed Vision System
A deployed Jetson Xavier system may depend on a legacy camera stack, a pinned JetPack release, a custom TensorRT engine, and an application that is no longer reproducible from its original build instructions.
CTai LABS can reconstruct the software baseline, document its dependencies, select the Connect Tech Jetson Orin migration path, rebuild the application, bring up the camera pipeline, regenerate the inference engine, and validate output, latency, and stability.
Move an Autonomous Machine to a Supported Compute Platform
An autonomous system may combine cameras, LiDAR, GNSS, CAN, ROS or ROS 2, localization, perception, planning, and control. Migration risk accumulates across the interfaces between these components.
CTai LABS can map the architecture, preserve required interfaces, update the software foundation, integrate the target Connect Tech hardware, and test the sensor-to-decision pipeline under representative workloads.
Complete a Migration That Has Stalled
A team may already have selected a Jetson Orin or Jetson Thor target but encounter incompatible packages, camera failures, container issues, TensorRT conversion errors, or performance below expectations.
CTai LABS can join at that point, isolate the blockers, stabilize the target environment, complete the application port, and establish measurable acceptance criteria for deployment.
Do Not Wait for a Production Interruption
The last-buy date is only one milestone. The larger risk is continuing to depend on a platform and software baseline without a validated replacement.
Starting now creates room to identify dependencies, obtain target hardware, rebuild the software environment, validate models and peripherals, correct performance issues, and qualify the complete system before the migration becomes urgent.
Connect Tech provides the hardware migration map. CTai LABS provides the integration path that turns the selected platform into a working, validated Edge ai system.
Download the NVIDIA Jetson Migration Guide
Bring your current NVIDIA Jetson module, Connect Tech carrier or system part number, application requirements, and known migration concerns. Use the guide to review potential NVIDIA Jetson Orin and Jetson Thor targets, software-stack considerations, form-factor changes, I/O differences, and available Connect Tech migration paths.
Download the NVIDIA Jetson Migration Guide
Your Physical ai Integration Partner
CTai LABS brings together Edge ai software, NVIDIA accelerated computing, sensors, application integration, validation, and production-ready Connect Tech hardware. Whether your migration begins with an early question, a deployed Jetson Xavier product, an incomplete port, or a system already moving toward Jetson Orin or Jetson Thor, CTai LABS can meet the program at its current stage and help carry it through deployment.
ABOUT THE AUTHOR
Ceri Nelmes
Tech Marketing Leader & Journalist | Building Brands at the Edge of What’s Next |
Connect Tech & CTai Labs
Ceri Nelmes is Head of Marketing at Connect Tech and an experienced technology journalist and digital strategist. She covers the technologies and market shifts shaping embedded computing, Edge AI, Physical ai, robotics, autonomous systems, and the NVIDIA® ecosystem. Working with Connect Tech and CTai LABS subject-matter experts, she turns engineering developments into accurate, useful reporting for developers, technical buyers, business leaders, and the media.
Ready to Build Smarter?
Let’s create the intelligent Edge AI solution that moves your business forward.