[Chyzwar]

All SOTA model providers are losing money. When users run Opus, they are essentially renting a GPU cluster worth half a million dollars for a $100/$200 subscription. If they want to IPO, they need to show a projection toward profit. For that reason, they want to discourage power users and attract normies.

[energy123]

> All SOTA model providers are losing money.

Source? I only read one article on this topic and they approximated gross margins at 50%.

> When users run Opus, they are essentially renting a GPU cluster worth half a million dollars for a $100/$200 subscription.

They use a large batch size, you're sharing the GPU with many other people.

[Chyzwar]

Gross margin calculated on API token pricing with discounted training and hardware deprecation.

I am no so sure about batch sizes, chatgpt napkin calculation for 5T model show 10-300 sessions.