AWS Panorama, Amazon's edge computer vision appliance and service, reached end of support on 31 May 2026. AWS says Panorama devices and applications no longer function after that date because they depend on the Panorama cloud service. Teams that used it must move their camera AI to current NVIDIA Jetson hardware or edge servers with an open media pipeline such as DeepStream, plus their own device management and security.
Updated October 2026
If your safety alerts, line counts or traffic analytics ran on a Panorama appliance, they are now dark. Every week without them means missed incidents, manual counting and audit gaps. The good news: the cameras, models and business logic you built are still valuable. What changes is the runtime underneath them, and 2026 edge hardware is far more capable than the 2021 appliance.
Who this is for: IT, OT and computer-vision leads in manufacturing, food processing, logistics, airports, retail and smart cities who ran or evaluated AWS Panorama and need a supported path for edge video AI.
What changed in 2026
- AWS's Panorama end-of-support page confirms: no new customers from 20 May 2025, and from 31 May 2026 the console, resources, devices and applications no longer work. AWS says it has no plans to extend.
- AWS's own migration advice is to replace the appliance with "a similar off-the-shelf device based on the current generation Nvidia Jetson platform" or an edge server, and to rewrite apps to remove Panorama-specific API calls.
- For the new media pipeline, AWS points to GStreamer, NVIDIA DeepStream (version 8.0 adds Jetson AGX Thor and Blackwell support) or the NVIDIA Triton inference server.
- Device management is now your job: AWS lists IoT Greengrass, Systems Manager and EKS/ECS Anywhere as building blocks.
What AWS Panorama was (history)
AWS announced general availability of the Panorama Appliance in 2021. It added computer vision to existing RTSP/ONVIF IP cameras: a small on-site box ran models locally, while the Panorama cloud service handled deployment and management. Typical uses were safety alerts at industrial sites, defect detection, queue and traffic monitoring and product counting. The idea was right; the dependency on one managed service is what made the shutdown so disruptive.
Panorama vs today's alternatives
| Option | Hardware | Pipeline | Management | Best for |
|---|---|---|---|---|
| AWS Panorama (ended) | Panorama Appliance | Panorama SDK | Panorama console | No longer available |
| Minimal-change port | Jetson device | PyAV / OpenCV app class | Greengrass or Systems Manager | Few cameras, fast recovery |
| Re-architect on DeepStream / Triton | Jetson Orin/Thor or GPU edge server | GStreamer + TensorRT/Triton | Containers on EKS/ECS Anywhere or K8s | Many streams, more models, more sensors |
| Edge AI platform (e.g. NexaStack) | Your choice of edge hardware | Managed inference + vision agents | Central deploy, monitor, govern | Multi-site fleets, private data, no lock-in to one service |
Reference architecture after Panorama
RTSP / ONVIF
replaces appliance
DeepStream / GStreamer
TensorRT / Triton models
RTSP credentials
deploy, update, rollback
MES, SCADA, tickets
health, accuracy, drift
Figure: Post-Panorama edge vision architecture, following AWS's migration guidance. Existing cameras stream to a Jetson device or GPU edge server running a DeepStream or GStreamer pipeline with TensorRT or Triton inference. A secrets store protects camera credentials, a fleet manager handles deployment and updates, events flow to plant systems, and monitoring tracks device health and model accuracy.
AWS also recommends following the Well-Architected Security Pillar and using hardware security features such as Greengrass V2 hardware security integration.
Panorama devices offline? Book a free Panorama migration assessment with NexaStack. We'll review your cameras, models and sites and give you a hardware and cut-over plan.
Migration plan: step by step
1. Inventory what Panorama was doing
List every appliance, camera stream, model, frame rate and downstream integration (alerts, dashboards, MES). Note which outputs the business still depends on.
2. Choose hardware by stream count
Size by cameras, resolution, frame rate and models per stream. Jetson devices suit a handful of cameras; GPU edge servers suit dozens. See GPU parallel processing for vision.
3. Port or re-architect the app
Remove Panorama API calls. A quick port with PyAV/OpenCV restores service fast; a DeepStream or Triton design handles more streams and sensors. Our guide to real-time video streaming covers pipeline choices.
4. Rebuild management and security
Store RTSP credentials securely, sign and roll out containers centrally, and keep a rollback path. Patterns in MLOps for edge AI apply directly.
5. Validate before cut-over
Run the new stack in parallel on recorded footage, compare detections and alert timing, then switch site by site.
How to start in 30 days
- Week 1: inventory appliances, streams and models; rank sites by business impact.
- Week 2: stand up one Jetson or edge server at the top site; port the main model.
- Week 3: connect alerts and dashboards; add credential storage and remote updates.
- Week 4: parallel-run, sign off accuracy and latency, then roll out the template to remaining sites.
Where it's heading: 2026 → 2030
| Period | Edge computer vision |
|---|---|
| 2021 | Single-vendor appliances like Panorama bring CV to existing cameras. |
| 2026 | Panorama ends; teams move to open pipelines on Jetson and GPU edge servers with their own fleet management. |
| 2028 (expected) | Vision-language models run on edge devices; operators query cameras in plain language. |
| 2030 (our view) | Cameras become sensors for governed physical-AI agents that act, not just alert. |
Where NexaStack fits
NexaStack's Edge AI solution is built for exactly this gap: it deploys, manages and updates AI models across many edge nodes through central orchestration, processes data locally for low latency, keeps sensitive video inside enterprise boundaries and keeps operating with limited or no cloud connectivity. Combined with Vision AI's real-time video intelligence, you get a supported home for your camera AI without depending on a single cloud service.
FAQ
Is AWS Panorama discontinued?
Yes. AWS ended support on 31 May 2026. After that date the console, applications and devices no longer work.
Can I keep using my Panorama Appliance offline?
No. AWS states the device and applications depend on the Panorama cloud service and will not function once it is discontinued.
What replaces AWS Panorama?
AWS recommends partner solutions or building your own with current NVIDIA Jetson devices or edge servers, a pipeline such as DeepStream or GStreamer, and management through Greengrass, Systems Manager or EKS/ECS Anywhere.
How much does migrating cost, and will it disrupt operations?
Costs depend mainly on stream count and hardware. You keep your cameras and usually your models. Parallel-running one site first keeps disruption low and gives a firm number for the rest.
What are the latest edge computer vision trends in 2026?
Jetson Thor-class edge compute, DeepStream 8.0 multi-camera tracking, vision-language models at the edge and platform-based fleet management.
Get your camera AI back online. Start a 30-day Panorama replacement pilot with NexaStack on your top site.
