My experience has been all GPU-intensive workflows have been pushed to the cloud. It works a lot better for CI/CD purposes as well, and most of the larger datasets are too practically large for your laptop, it ends up being prohibitively slow to download datasets from databases to your own laptop and then train on your local machine.
I could be biased since most of my network is in the startup scene in SV, where hardware cost is generally the LAST thing most companies worry about. I haven’t seen a non-mac software company that’s not a 5000+ dinosaur person company.
My experience has been all GPU-intensive workflows have been pushed to the cloud. It works a lot better for CI/CD purposes as well, and most of the larger datasets are too practically large for your laptop, it ends up being prohibitively slow to download datasets from databases to your own laptop and then train on your local machine.
I could be biased since most of my network is in the startup scene in SV, where hardware cost is generally the LAST thing most companies worry about. I haven’t seen a non-mac software company that’s not a 5000+ dinosaur person company.