not to mention that it depends on CUDA, whose legal terms can be found here (archived copy).

i'd rather spare myself the need to download a 1.4 GB archive to verify CUDA's contents, so may the next EULA snippets suffice to give everyone an idea of its "proprietariness":

1.2.1.2. Usage Rights: Developer may install and use multiple copies of the Materials on a shared computer or concurrently on different computers, and make multiple back-up copies of the Materials, solely for Licensee's use within Licensee's Enterprise. "Enterprise" shall mean individual use by Licensee or any legal entity (such as a corporation or university) and the subsidiaries it owns by more than 50 percent.

1.2.1.3. Further Redistribution Rights: ... Licensee may authorize Customers to further redistribute the Redistributable SOFTWARE that such Customers receive as part of the Licensee Application, solely in binary form ...

1.2.1.6. Limitations: No Reverse Engineering ... No Separation of Components ... No Rental ... No Modifications

python-tensorflow-cuda, tensorflow and tensorflow-cuda will be blacklisted as a result of this