[jdiff]

I think there's a lot of spare GPU compute. I don't think it'd be made available for this, or if it was, I believe it'd most likely be from a company that won't survive the AI bust. But I think there's a lot of spare compute sitting idle out there.

If there weren't, I think token prices would be higher than they are, downtime would be higher than it is, and there would be far fewer free tiers or more restrictions on free tiers.

[Zababa]

There's also an alternative which is that token prices are already high, downtimes do exist (semi frequent on Claude) or are managed by serving degraded versions/quants (speculation that has never really be proven afaik) or reducing thinking time ("juice" values from open AI), and that free tiers are not used much at all or are a loss leader.

[YmiYugy]

I don’t think there is a ton of spare capacity. But even if there was the machines are designed for AI not gaming, e.g. low CPU clocks, the GPUs don’t have RT cores or video encoders, etc. Stadia had custom hardware, GeForce Now uses workstation GPUs, xCloud has custom servers. Gaming also isn’t a good use for spare capacity. Demand is very spiky and people don’t have a lot of time flexibility.

[jvanderbot]

There's a pretty fair staging argument of AI for work, Gaming for play, that would keep north american data centers with north american customers running 2/3 of the day.