Both open AI and anthropic are making a profit on inference / api usage. So if you’re using Claude code and burning through millions of tokens they’re making money off that. The revenue from this is also growing rapidly.
What’s dragging down the bottom line is training, which is the big money sink. They have to do it though to make the next biggest model to compete with each other. They’re hoping that they’ll hit a plateau in capabilities, whether that be AGI or some government regulation, so they don’t have to train any more. At that point they can sit back and collect billions from the inference money printer.
I could be wrong, but I don’t think that’s what the question was about.
It was “if people are trying to use Anthropic or OpenAI’s services to build a profitable company, why wouldn’t Anthropic or OpenAI just build that company themselves”
And the answer there would be lack of subject matter experts. You need people who know the industry they’re building the product for if you want to have any fighting chance. AI can work as a force multiplier, trying to use it as a completely autonomous agent that gets near-zero human feedback is hopeless. I doubt any of the AI companies want to take on thousands of employees just to use their own AI to take risks on new industries when they could just sell the AI to people who take on the risk.
Both open AI and anthropic are making a profit on inference / api usage. So if you’re using Claude code and burning through millions of tokens they’re making money off that. The revenue from this is also growing rapidly.
What’s dragging down the bottom line is training, which is the big money sink. They have to do it though to make the next biggest model to compete with each other. They’re hoping that they’ll hit a plateau in capabilities, whether that be AGI or some government regulation, so they don’t have to train any more. At that point they can sit back and collect billions from the inference money printer.
I could be wrong, but I don’t think that’s what the question was about.
It was “if people are trying to use Anthropic or OpenAI’s services to build a profitable company, why wouldn’t Anthropic or OpenAI just build that company themselves”
And the answer there would be lack of subject matter experts. You need people who know the industry they’re building the product for if you want to have any fighting chance. AI can work as a force multiplier, trying to use it as a completely autonomous agent that gets near-zero human feedback is hopeless. I doubt any of the AI companies want to take on thousands of employees just to use their own AI to take risks on new industries when they could just sell the AI to people who take on the risk.