The entire marginal cost to serve AI models is paid for by the API costs of all providers by nearly every estimation. The cost not currently recouped is entirely in the training and net-new infrastructure that they're building.
And the open source models are only months behind, so the big AI companies need to keep burning money on R&D with no end in sight. If OpenAI took a quarter off from model development, they might fall behind forever.
Are you sure? I thought tokens (or watts) were sold at such a loss that if current supply limits were reached they’d go broke
These companies are generally profitable for inference but it does not cover the cost of R&D (training).
If its profitable why are they banning people from using it in systems like claw?
1 reply →
The entire marginal cost to serve AI models is paid for by the API costs of all providers by nearly every estimation. The cost not currently recouped is entirely in the training and net-new infrastructure that they're building.
And the open source models are only months behind, so the big AI companies need to keep burning money on R&D with no end in sight. If OpenAI took a quarter off from model development, they might fall behind forever.
So why are they banning people from using it in systems like claw?