They were only using 5% capacity of their datacenter GPUs. They could have trained their deep learning models close to 20X more, etc. He mainly just said it wa

They were only using 5% capacity of their datacenter GPUs. They could have trained their deep learning models close to 20X more, etc. He mainly just said it was a signal of their inefficiency as an organization and was a final straw, not a sole reason. He detailed many other major misallocations of resources throughout his time there.