[eddd]

As a beginner in deep learning space, I am a bit baffled about the case "You need a lot of computational power". Good models learn fast, so if potential model looks promising on local machine, one can do training on gcloud for 100$ on high end machines. Where am I wrong in this line of thinking?

[nl]

No, you are absolutely right. And modern transfer learning improves this even more in many domains.