New details reveal that former NVIDIA CEO Jensen Huang used circular financing to fund AI startups and absorb excess TSMC wafers and memory rather than letting GPU prices drop
What makes you think there’s a flood of chips coming? The chip makers have been producing them at a constant rate. The electricity shortage is an issue, but that’s why the hyperscalers have been spending so much on generators and talking about restarting mothballed nuclear reactors.
If the AI bubble pops, the chips will mostly go unused, or at least lightly used, and the electricity demand won’t happen.
There is no limit to the amount of AI that can be used.
For something to be “used” it has to have more utility-value beyond its energy costs, and I’m not sure AI will ever have that. What we have right now is a waste of energy that produces slop code slower than human devs used to.
At least it is a better search engine but my impression is that enough people think the code is good enough for many usecases like creating prototypes or creating web pages.
AI will stay. There is no limit to the amount of AI that can be used.
Well, that’s debatable. Right now the advances and continued low costs are heavily subsidized by the speculation. As speculation slows, it remains to be seen how that impacts costs, demand and quality.
Shattering? Upholding the perception.
The next bubble is coming. Who is providing the electricity for the coming flood of chips?
Solar… there is enough power to go around.
What makes you think there’s a flood of chips coming? The chip makers have been producing them at a constant rate. The electricity shortage is an issue, but that’s why the hyperscalers have been spending so much on generators and talking about restarting mothballed nuclear reactors.
If the AI bubble pops, the chips will mostly go unused, or at least lightly used, and the electricity demand won’t happen.
Now that people know that there is excess silicon they can’t hide it anymore.
The AI company valuation may pop but AI will stay. There is no limit to the amount of AI that can be used.
For something to be “used” it has to have more utility-value beyond its energy costs, and I’m not sure AI will ever have that. What we have right now is a waste of energy that produces slop code slower than human devs used to.
At least it is a better search engine but my impression is that enough people think the code is good enough for many usecases like creating prototypes or creating web pages.
InB4 the Executive Admin Feds outlaw openweight models from China.
Well, that’s debatable. Right now the advances and continued low costs are heavily subsidized by the speculation. As speculation slows, it remains to be seen how that impacts costs, demand and quality.
There is certainly a limit. One we’ve already reached.
There isn’t excess silicon.