I know, but doesn't make sense to offer them in a couple of years. When we already have TensorFlow running in GPUs in amazon or other companies. Actually we mad

I know, but doesn't make sense to offer them in a couple of years. When we already have TensorFlow running in GPUs in amazon or other companies. Actually we made numbers in my company and Amazon GPUs are cheaper than Google Cloud with the current pricing. I am talking about more than 10 million inferences a day.