What Led to Amazon's Decision to Shut Down AWS Panorama?

From my experience, I found Panorama was mainly for prototyping computer vision projects. Unfortunately, it stepped on the toes of other Kinesis video solutions and some IoT products. The hardware was pretty pricey compared to a basic NUC setup, and the only real advantage was easier code deployment via their management console. For production, it didn’t really hold up, as you could run small ML models directly on third-party cameras just as effectively. Moreover, AWS is shifting focus towards generative AI, which scales better, requires powerful GPUs, and has a simpler learning curve.

Many of Amazon's recent shutdowns target "built solutions," which can be tough for AWS to maintain due to their less customizable nature. I suspect that AWS Panorama faced similar issues, as custom hardware solutions like that tend to have high maintenance costs versus the revenue they generate. It seems like customers were either too small or had demanding needs that didn't justify the investment compared to offering building blocks for others to develop their own solutions using services like Kinesis, Media services, or S3, which could replicate much of Panorama's functionality.

Honestly, I'm not surprised by the shutdown. AWS Panorama had its share of significant flaws. The service wasn't as easy and secure as it claimed to be; it was actually one of the worst experiences I had. It relied on two modes—CPU inference and GPU—but the GPU mode lacked the necessary security. The documentation was poor and outdated, leading to a lot of frustration. We ultimately moved to a Jetson Nvidia Developer Kit, which was a much better investment in terms of time and money. Oh, and they used a now-obsolete ML framework, which didn't help either.