WeWork had a massive cost base to offer their service. They had to rent/buy a lot of real estate to serve their customers. None of all the examples you list here except CoreWeave has a comparably high cost base.
All the mayor AI players have to rent/buy a lot of computing power to serve their customers. We're talking costs of 2-5x (or 10x? does anyone actually know?) their revenue. The question is, will they ever be able to pay their cost of revenue and operating expenses? Many people doubt. And we know that they do all sorts of funny book keeping shenanigans like SPVs and delayed deprecation.
What comes on top is, LLMs have no economy of scale. More customers scale linear with higher costs. Will people or companies be willing to pay 5x what they're paying now? Will future frontier models become more cost efficient or less? We simply don't know yet. At this point, it's a gamble.
All the mayor AI players have to rent/buy a lot of computing power to serve their customers. We're talking costs of 2-5x (or 10x? does anyone actually know?) their revenue. The question is, will they ever be able to pay their cost of revenue and operating expenses? Many people doubt. And we know that they do all sorts of funny book keeping shenanigans like SPVs and delayed deprecation.
What comes on top is, LLMs have no economy of scale. More customers scale linear with higher costs. Will people or companies be willing to pay 5x what they're paying now? Will future frontier models become more cost efficient or less? We simply don't know yet. At this point, it's a gamble.