It is possible to prove, mathematically, that DRM cannot work when an attacker has physical possession of the machine. It is simply impossible - unless the CPU can execute encrypted code where the decryption key is a galactic secret and cannot be found using an electron microscope.

When the code executes on a remote server, however, you have a completely different kettle of fish.