So what?

Why on earth would you want an AI datacenter in space? Like what would you gain by doing that at an absurdly higher cost than you could build on them earth?

"Free" energy? lol.. just build nuclear powerplants or loads of solar, wind and batteries on earth. Its still going to be cheaper...

> How is anyone going to compete with that? There are a bunch of data-center-in-space startups

A better question is why would anyone even try?

> are mostly leasing their data centers from someone else

It's really not. Building your own datacentres is very expensive and more importantly takes a lot of time. They need compute now, so it makes perfect sense to rent it from failing AI companies like xAI which bought a lot of chips but don't have anything to do with them since their models are just not very good.

> But if SpaceX has the cheapest data center with the most capacity, they will be able to extract profits

Well.. that would be a first one, since no similar industry works that way. Compute is a commodity so unless your literally run out of space on earth to build datacenters or Nvidia/etc. stop selling to anyone but SpaceX that can't really happen, can it?

Even if their model is competitive or even surpasses e.g. Deepseek (which is far from given) how would that justify a huge valuation?

So yes, ridiculous things like that happen and markets are very often not rational at all (short and medium term at least).

Nortel, Sun, AOL, Cisco were all very innovative and rapidly growing companies. Until reality kicked in.

SpaceX IPO price already has many years of extremely high growth priced in. Comparing it to Facebook's or Google's IPOs is like apples to oranges.

Based on "sane"/traditional metrics that and much more is already priced in into the IPO valuation.

e.g. Google had a many times lower P/S ratio at their IPO and was actually profitable (and software companies usually have higher valuations than capital intensive ones like SpaceX anyway). SpaceX is already valued at more than Google was 10 years after its IPO while barely making a tiny fraction of its revenue.

Until companies start hiring 5x less engineers than they did before and well.. we are clearly moving towards that direction

I find anything below 50 tps or so entirely unusable...

Regardless its Apples to oranges anyway, inference is quite cheap for open weight models its just that Claude and OpenAI can charge very high margins compared to e.g. DeepSeek or various provider on OpenRouter since open models are a commodity.

The token cost for very large open models seems to indicate that. Of course DeepSeek is probably subsidizing their current pricing to a significant extent but then you have all other provides on OpenRouter and I can't imagine it would make sense for them to operate at a significant loss since they don't offer anything but inference.

Anthropic is just utilizing its position to maximize the prices as much as possible and see what the market can bear and of course they are not even close to offsetting their R&D and capital investments.

If LLMs become commoditized and open models become good enough for most use cases its hard to imagine that prices won't drop (and the AI bubble popping as a consequence of that...)

What? In what way would the change? They are already raising prices..