I don’t know about licensing issues but I know the issue is in part but that OpenCV’s cuda module does have some issues relating to the performance and stability. If you ever run the test suite when building OpenCV, you’ll see that there are quite a few things that are broken. It’s a lot to ask Nvidia to maintain something that’s not really their product, especially one that even working perfectly does not perform as well as solutions designed explicitly for Nvidia hardware, and Tegra specifically.
I would like to think they would take market share from Intel and others (OpenCV is an Intel project), but what might be just as likely to happen is people will use the plain old CPU based cv2 module and blame Nvidia when it’s slow. Code has to be rewritten to use the cuda module, and if you’re going to rewrite it, it might not make any sense to use OpenCV at all.