Google released earnings yesterday and had a big cloud beatHere’s my question: if every hyper scaler is now using their data centers for cloud compute (MSFT,AMZN,GOOGL,META)At some point supply > demand and the margins of cloud go down, no? When it was just AWS and azure I could understand the margins, but if we keep building hundreds of billions of data centers, and AI turns out to not be intensive, wouldn’t the cloud compute business get eviscerated via margin compression?Doesn’t this literally kill MSFT and MSFT?
>>62513631*kill MSFT and AMZN
>>62513631No. MSFT’s shitty CEO kills MSFT. As for AMZN, they’ll be fine.
>>62513631Datacentres are a bad businessHigh energy costsWeather viunerablehigh insuranceand the network compute, GPS, CPUs and RAM going into them depreciates to zero in 4-5 yearsSoftware WAS a great business, nearly 100% margin over staff costs, low distribution costs, basically legal immunity on product faults etcBeing a datacentre landlord selling it as a service is a shit business. Go look up what happened to the last generation of datacentre companies like Sunguard.Never ever invest in anything with a jeet CEO
>>62513657Amazons retail business was built on selling cheap crap from China and is effectively over.Beyond that they are a datacentre landlord as well, as is meta (spyware and ads) and google (with an ad agency and spyware business bolted on).Shit business models. Advertising and marketing is a vastly overvalued industry as well.
>>62513631They could just all mine Bitcoin
>>62513631All I know is that a new generation of chips—much less power-hungry—is coming to market soon. Chinese chips, neuromorphic chips, etc.
>>62513960Then your full of shit and a bad business is a bad business. ASML makes the stuff tha makes the chips. China is locked out and going nowhere and staying firmly 10-15 years behind and is late to a bubble that is bursting not becuase of tech but because of shiller cape and retarded jeet ceo business models and duplicitous earnings and financing.
Tesla robotaxi shit Vs financial realityhttps://www.youtube.com/watch?v=xGi6j2VrL0o
To know when compute demand runs out you need to look at the end point application of real world AI> 1 human brain (ideally AGI) worth of compute per flock camerais a realistic goal. Since it's a matter of safety, there's no price too high either. We have a long way to go so even the drooling retards ITT who will remain doubters well into 2035 will still have time to make money on it into 2040 and beyond
