This is the core of my belief that data center construction is a huge bubble.
AI is not a bubble, IMO, though we may see a retrench and some companies with sky-high valuations will crash to more reasonable ones. But data center demand is probably a bubble, and the main driver will be reduction in the actual amount of power and data center space required to serve escalating demand.
I think hardware and model improvements will pace or maybe outrun demand and then when demand starts to saturate will keep going and leave a lot of orphaned data centers.
I disagree. I own over 1TB of vram at home. I can tell you that it's not a bubble. From my builds, I would rather have cloud, cloud is easier. From running small models like Qwen3.8-27B to large models like Qwen3.8-2.4T. I can tell you that small models will never be enough or match up. Everyone will want the smartest model, not just a good enough model.
> Everyone will want the smartest model, not just a good enough model.
Not so sure about this. There’s always a potential threshold. After all, we don’t all use the most powerful computers, the latest phones, the highest resolution cameras, the fastest or best cars.
I am already not interested in cloud LLMs and I don’t even use the best (on paper) model that I can run locally. I prefer a model that people insisted (here) was “dead on arrival” but appears to work better for me.
I think the difference is that AI as an edge. That edge will turn into more money, better quality of life, etc. Of course, with serious skills, you might be able to use use a not so smart model to keep up with folks with smart models. People are lazy tho, and will prefer for AI to do all the work if it means they do none.
I don’t know. But if you assume that the AI companies have an enterprise subscription product available to any given market sector then given their desperate need for money they will sell it to anyone and everyone. So it becomes a widely used standard feature. Having access to it doesn’t give you an edge; it puts you on the same surface as everyone else.
It’s like being an algorithmic betting exchange gambler; being the first with access to some new stream of information may give you a very temporary advantage but once everyone has access to it, the market prices it in.
If you want an advantage you have to seek it out elsewhere.
It might be in the harness or tooling, but if your cloud LLM can write it quickly for you, it’s the same for your competitors; their cloud LLM can write it for them.
The idea that using a cloud LLM is an edge — an advantage — doesn’t stand up well to scrutiny.
api · · focus · HN ↗
AI is not a bubble, IMO, though we may see a retrench and some companies with sky-high valuations will crash to more reasonable ones. But data center demand is probably a bubble, and the main driver will be reduction in the actual amount of power and data center space required to serve escalating demand.
I think hardware and model improvements will pace or maybe outrun demand and then when demand starts to saturate will keep going and leave a lot of orphaned data centers.
segmondy · · focus · HN ↗
dofm · · focus · HN ↗
Not so sure about this. There’s always a potential threshold. After all, we don’t all use the most powerful computers, the latest phones, the highest resolution cameras, the fastest or best cars.
I am already not interested in cloud LLMs and I don’t even use the best (on paper) model that I can run locally. I prefer a model that people insisted (here) was “dead on arrival” but appears to work better for me.
segmondy · · focus · HN ↗
dofm · · focus · HN ↗
The edge is somewhere else.
fittingopposite · · focus · HN ↗
dofm · · focus · HN ↗
It’s like being an algorithmic betting exchange gambler; being the first with access to some new stream of information may give you a very temporary advantage but once everyone has access to it, the market prices it in.
If you want an advantage you have to seek it out elsewhere.
It might be in the harness or tooling, but if your cloud LLM can write it quickly for you, it’s the same for your competitors; their cloud LLM can write it for them.
The idea that using a cloud LLM is an edge — an advantage — doesn’t stand up well to scrutiny.
The only edges left will be human.