[Rocco DiPippo]

Some new consumer computers can already run Large Language Models (LLMs) on them. I predict that within a few years, personal computers with massive GPUs and massive amounts of RAM will be readily available to the public, making most AI data centers moot.

Because of this inevitability, I think that the vast majority of AI data centers currently being proposed won't even be built or will be abandoned before completion.

[dayglored]

Agreed.

Within a couple years most ground based AI will be running on people’s laptops for personal use.

A few years after that AI will be running in orbit for commercial, enterprise, and governmental use.

Big land based datacenters are a temporary aberration. The established ones like AWS, Azure, etc. will remain but this crop of proposed new ones is likely to fizzle.

[Candor7]

I agree 100 percent.

But I wonder how I will be cooling my new super AI lap top?

[dennisw]

Large Language Models (LLMs) ..... You have a $3000 one at home full 128gb ram. You are a fool n clown, the only thing you will use for is making AI images of your cat chasing cockroaches.

[Lockbox]

Some new consumer computers can already run Large Language Models (LLMs) on them. I predict that within a few years, personal computers with massive GPUs and massive amounts of RAM will be readily available to the public, making most AI data centers moot.

This appears to be a plan to move the load to a network of personal computers which lowers the load on data centers. Free outsourcing in other words. Thinking of Moore's Law, data processing will double every two years for a while. Where to get all the increased capacity?