Revenue growth has been dramatic. OpenAI generated $3.7 billion in revenue in 2024 before jumping to $13.07 billion in 2025. By the end of that year, monthly...
In theory costs could come down with each new hardware generation if the we dont keep pushing models the to max extent of what the hardware can do while pushing size.
E.g Claude Opus today, only trained in a similar size and manner as today, will be cheaper to run on whatever the next GPU that comes out with higher speeds and processing capabilities, unless of course NVidia raises the cost substantially. Given the current situation I think nvidia might do that which would hamper this lowering of costs, but it should possible, if not slower.
E.g 10 years from now it will be cheaper to run a opus similar model. But 10 years from now everyone will want the mythos of today, then. That wont be cheaper.
This has been stated since ChatGPT was released and has not happened. The video cards released specifically for LLM usage do not benchmark particularly better than the previous generation. And it’s still unbelievably expensive to run these cards and maintain the facility and, again, you only get like 3 or 5 years out of them! That’s a crazy investment lol
But the new GPUs absolutely can run the GPTs from back then better. We just dont want that anymore, we want the better bigger models that continue to be as or more expensive as what it was back then.
When you replace the cards in 5 years it’ll run it even better. We just wont want that then.
Edit: and gains dont have to be huge, even 5-10% between generations, but take that to 10 years like I said and it can be substantial.
In theory costs could come down with each new hardware generation if the we dont keep pushing models the to max extent of what the hardware can do while pushing size.
E.g Claude Opus today, only trained in a similar size and manner as today, will be cheaper to run on whatever the next GPU that comes out with higher speeds and processing capabilities, unless of course NVidia raises the cost substantially. Given the current situation I think nvidia might do that which would hamper this lowering of costs, but it should possible, if not slower.
E.g 10 years from now it will be cheaper to run a opus similar model. But 10 years from now everyone will want the mythos of today, then. That wont be cheaper.
This has been stated since ChatGPT was released and has not happened. The video cards released specifically for LLM usage do not benchmark particularly better than the previous generation. And it’s still unbelievably expensive to run these cards and maintain the facility and, again, you only get like 3 or 5 years out of them! That’s a crazy investment lol
But the new GPUs absolutely can run the GPTs from back then better. We just dont want that anymore, we want the better bigger models that continue to be as or more expensive as what it was back then.
When you replace the cards in 5 years it’ll run it even better. We just wont want that then.
Edit: and gains dont have to be huge, even 5-10% between generations, but take that to 10 years like I said and it can be substantial.