This is a list of cloud providers that offer a free tier for machine learning workflows. This is to lower the barrier of entry for machine learning and to prove you don't need a local rig to get started. If anything in this list is incorrect, reach out to me on X so I can make the fix. If there is a cloud computing platform that I've missed, also let me know.
If you'd like to support this resource, give it a start and follow me on X.
These are my favorite choices for getting started. They offer generous free tiers in a notebook-like structure to make it really easy to get up and going. They also don't rely on working with a single cloud provider.
- Google Colab: Offers free access to GPUs (usually NVIDIA T4 or P100) and TPUs, but with limited usage time and resources. Best for small projects and experimentation.
- Kaggle Notebooks: Provides 30 hours/week of GPU usage (NVIDIA Tesla P100 or T4) for free. It's a good option for data science and machine learning enthusiasts with access to Kaggle's datasets and community.
These are other options for more free compute. A lot of these are entry tiers to get you using their AI studio or cloud provider, but they're still worth checking out.
- Lightning AI: Offers one free studio with 22 GPU hours.
- Google Cloud Platform: Offers $300 in free credits to new users, which can be used for GPU instances. This is a good option for larger projects or more powerful GPUs, but the free credits may not last long depending on usage.
- Amazon SageMaker: Provides a free tier with limited access to various machine learning resources, including some GPU instances. It's a good starting point for exploring SageMaker's capabilities.
- Paperspace Gradient: Offers a free community tier with access to limited GPU resources for experimentation and learning.
- Vast.ai: This is a marketplace where you can rent underutilized GPUs from individuals at competitive prices. It's a good option for flexibility and potentially lower costs, but requires more setup and management.