Selling Gen AI dreams with zero GPUs
Saw @Suhail tweeting today about another massive GPU crisis and hoarding compute bhai.. Literally 10 mins later I was sitting in a pitch where leadership promised a full Generative AI transformation roadmap to a client by next quarter.
Every single client wants one right now. But with physical infrastructure drying up globally... how are any of these tech and consulting firms planning to actually deliver? My org is signing these massive enterprise AI deals but ngl we don't even have the compute to back it up. We are just selling vaporware at this point.
Does Deloitte have their own ai infra?
Won't they just use bedrock or openrouter or whatever?
Bhai exactly. Most of these massive transformations are just going to be API wrappers around Bedrock or Azure. But even getting provisioned throughput there is getting tough tbh.
