Google Colaboratory, universally shortened to Colab, is a hosted notebook environment that Google opened to the public around 2017. Its premise is best captured by its own FAQ: “Colab is a hosted Jupyter Notebook service that requires no setup to use and provides free of charge access to computing resources, including GPUs and TPUs” (https://research.google.com/colaboratory/faq.html). For a field where getting a working environment was often the hardest first step, “no setup” was the entire pitch.
Before Colab, running a neural network meant assembling a stack on your own machine: a compatible Python, the right versions of numerical libraries, a CUDA toolkit matched to your GPU driver, and often hours of failed installs before a single line of model code ran. Worse, many learners and researchers had no GPU at all. Colab dissolved that barrier. A student with only a web browser and a Google account could open a notebook, select a GPU or TPU runtime, and start training, with the heavy hardware living in Google’s data centers rather than on a laptop.
Technically, Colab is built directly on the open-source Jupyter project. The FAQ notes that notebooks are stored in Google Drive and can be loaded from GitHub, and that they “can be shared just as you would with Google Docs or Sheets,” bringing the collaborative editing model of Google’s office suite to executable code (https://research.google.com/colaboratory/faq.html). The Jupyter document format underneath meant a Colab notebook was portable: it could be downloaded and run in a local Jupyter installation, and vice versa.
The free tier came with deliberate limits. The FAQ is candid that “resources in Colab are prioritized for interactive use cases” and are not guaranteed, with the service steering away from long-running background jobs or attempts to bypass the notebook interface (https://research.google.com/colaboratory/faq.html). Sessions could be reclaimed, idle runtimes disconnected, and a paid Colab Pro tier later offered more reliable and more powerful hardware. The economics were honest: free GPUs at planetary scale are a loss leader, rationed accordingly.
Colab’s significance to software history is that it made the development environment for machine learning a public utility rather than a privilege of the well-equipped. A generation of practitioners wrote their first model in a Colab cell, and a generation of tutorials, courses, and research artifacts shipped as one-click “Open in Colab” links. It also propagated the notebook as the default surface for AI work, carrying both the accessibility and the hidden-state hazards of that format to a far wider audience than the desktop Jupyter ever reached.