People familiar with inference costs think that the API prices are either already profitable enough. The major cost sinks are the subsidized subscriptions and RnD costs.
The subsidized subscriptions are similar to how Uber burned cash and jacked up prices after killing competition. The RnD costs remain difficult to offset unless these companies have sufficient scale
The RnD costs remain difficult to offset unless these companies have sufficient scale
Ya thats going to be deciding factor. Can you get enough users to offset it and stay alive before you run out of cash. If you slow down, and your competitors still have enough cash they’ll make a better product and you’ll start losing users.
Im not convinced something like Claude isnt profitable with enough users. I dont think people are spending more in compute than they pay.
Getting enough paying users though requires it to be better so more people will pay.
Obviously the free tier is at a loss, but I mean at a per paid user level.
People familiar with inference costs think that the API prices are either already profitable enough. The major cost sinks are the subsidized subscriptions and RnD costs.
The subsidized subscriptions are similar to how Uber burned cash and jacked up prices after killing competition. The RnD costs remain difficult to offset unless these companies have sufficient scale
Ya thats going to be deciding factor. Can you get enough users to offset it and stay alive before you run out of cash. If you slow down, and your competitors still have enough cash they’ll make a better product and you’ll start losing users.