Google Collaboratory, a tool that enables anyone to create and run arbitrary Python code using a web browser, is now offering a pay-per-use option. Since Google introduced premium Collab plans in 2020, Collab has undergone its first pricing adjustment. Users will now have the choice to purchase additional compute time in Collab with or without a paid subscription.
According to Google, the updated Collab tier won’t change the free tier, which is still available. The other significant change is that customers can now purchase “compute units,” which are units of computing power, starting at $9.99 for 100 and $49.99 for 500.
Now, paid customers have the freedom to burn up their compute quota, expressed in compute units, at any pace they see fit. As computer units are used up, a user has the option to decide whether to add more at a later time using pay-as-you-go. A user’s Colab use quota will return to our free tier limits once they have used up all of their compute units.
Google revealed that paid Colab users can now select between regular or “premium” GPUs in Colab, with the latter often having Nvidia V100 or A100 Tensor Core GPUs, in conjunction with the pay-as-you-go rollout. The business warns that acquiring a specific GPU chip type assignment isn’t guaranteed and depends on a variety of variables, including availability and a user’s paid balance with Colab (standard GPUs in Colab are typically Nvidia T4 Tensor Core GPUs).
Premium GPUs will obviously consume Colab compute units more quickly than regular GPUs.
The introduction of pay-as-you-go alternatives in Colab was hinted at some weeks ago when Google emailed Colab users to inform them that it was implementing the aforementioned compute units scheme for subscribers. The change was presented as a move toward greater openness that would give users “more control over how and when they utilize Colab.”
Some saw the action as a user-hostile attempt to increase prices or restrict Colab usage. However, a Google representative noted that restrictions have always been in place for all Colab usage paying plans, according to a statement to TechCrunch.
The representative explained in an email that “these adjustments are aimed to give users more visibility into… restrictions.” “Colab will continue to maintain its free tier, which includes fundamental GPU access.”
The sensitivity to price adjustments is an indication of how much Collab has expanded since it split off from a Google Research internal project in late 2017. Within the AI research community, the platform has established itself as the de facto digital breadboard for demos; it is typical for researchers to connect to Collab pages on or alongside the GitHub repositories holding their code.