Machine Learning(ML) is extremely computationally intense and makes heavy use of Parallelization. Thus, it is strongly advised to make use of a GPU (or a cluster of them) when training or running ML models.
This can be done locally if you have a sufficiently equipped machine. If your machine does not have a GPU or if a single one is not enough for your needs, then cloud computing is advisable.
Multiple cloud providers exist:
Google Colaboratory, or Colab for short, is a cloud computing environment that is free of charge. Shared access to Tesla K80's GPU's is more than enough for most basic ML tasks, as well as all of the fast.ai lecture code.
The development environment of Colab is a Jupyter Notebook.
Similar to Google Colab, Kaggle offers a cloud computing environment with free access to K80's GPU's. Development environment is also Jupyter Notebooks.
Paid cloud providers have varying features, but generally they offer higher performance and stability, at a price. Suitable for production needs.