# \#GCP Tensorflow question

**URL:** <https://community.cloudbank.org/t/gcp-tensorflow-question/48>\
**Category:** Uncategorized\
**Created:** [August 20, 2020, 1:15am UTC](https://community.cloudbank.org/t/gcp-tensorflow-question/48 "2020-08-20T01:15:34Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![rob](https://sea2.discourse-cdn.com/flex016/user_avatar/community.cloudbank.org/rob/32/5_2.png) [@rob](https://community.cloudbank.org/u/rob)\
**Post date:** [August 20, 2020, 1:15am UTC](https://community.cloudbank.org/t/gcp-tensorflow-question/48/1 "2020-08-20T01:15:34Z")

</div>

I read that TensorFlow on Google’s Tensor Processing Units (TPUs) are like 80 times faster than GPUs for CNN-type research computing. That sounds like GCP is by far the most cost-effective cloud for this. Is that right? Or are there equivalent alternatives on other clouds?