AI Cloud

The AI Cloud supports a range of scientific computing applications, including deep learning and machine learning. The cloud is based on two NVIDIA DGX-2 with V100 GPUs. It is based on a traditional HPC queueing system for resource management with modern cloud containers to hold your software. A good starting point for finding if your software can be obtained in an easy way is to go to the Nvidia GPU Cloud (NGC). If you need another piece of software that is not listed at NGC, please let us know. We might be able to build it for you.

Link to the new AI-Cloud documentation

Link to the old AI-Cloud pilot documentation (included for useful links to examples)  


Apply for access


The AI Cloud is available to all researchers at AAU for research purposes. Students may also get access upon their supervisor’s approval in a written email from the supervisor, stating that access for the students are necessary in order to conduct research at AAU (please be aware that this is a shared resource so please be mindful about your allocations).

Please fill out this email template in both cases (researcher or supervisor).

Access for students to GPU resources will be removed after each semester end immediately following February 1st and July 1st. This is our user policy on the AI Cloud platform. It is required that the supervisors resubmit their request every semester.

If the email template doesn't work, please write to with the follow pieces of information:

  1. Name and Email of your student if you are recommending your student
  2. Special requirements (special software not listed in the documentation, heavy HW requirements etc.)

Note: This information is just to make sure that you can fit in the system and that we have software for you.  We will contact you if we need to discuss your special requirements to find a solution.


We have some alternatives if you agree to any of the below statements:

  • you like to work from a local computer.
  • feel uncomfortable working with a HPC queuing system and/or command line tools.
  • dont need as much compute power.
  • dont need to do training, but only prediction.

The alternatives are to borrow:

If your interest in borrowing either of the above two alternatives, please also write to, and we make the necessary arrangements.