@nprofile1q... @nprofile1q...
Try is the operative word. They have inference costs, which scale with users. They also have training costs, which are independent of the number of users.
The more users you have, the more easily you can amortise the training costs (assuming you’re making some profit per user).
As you increase the price, you decrease the number of users (assuming that this isn’t an inelastic market, and new technology outside defence or healthcare rarely is). This makes it possible to reach operational profitability: the inference costs are less than the price. But it also means you need to a,or rise the training costs across fewer customers. This means you need to put up the prices more, and so on.
If you don’t train new models, you have no moat, so you’ll lose market share quickly.
Try is the operative word. They have inference costs, which scale with users. They also have training costs, which are independent of the number of users.
The more users you have, the more easily you can amortise the training costs (assuming you’re making some profit per user).
As you increase the price, you decrease the number of users (assuming that this isn’t an inelastic market, and new technology outside defence or healthcare rarely is). This makes it possible to reach operational profitability: the inference costs are less than the price. But it also means you need to a,or rise the training costs across fewer customers. This means you need to put up the prices more, and so on.
If you don’t train new models, you have no moat, so you’ll lose market share quickly.
1
