Comment by dev_l1x_be

I think the third company (likely Google) is going to make LLMs financially feasible with:

- dedicated hardware (https://cloud.google.com/tpu)

- optimized models (https://research.google/blog/turboquant-redefining-ai-effici...)