This seems somewhat misleading. Lots of products take a lot of investment in them for many years before they reach profitability. The Boeing 787 Dreamliner, for example, was in development for 7 years and another three years after that before it was profitable. The Falcon 9 rocket took 13 years to develop and now it’s the most profitable satellite launcher around. The Dyson bag-free cyclonic vacuum cleaner took 15 years to develop.
Most of this AI stuff has only been in heavy development since ChatGPT burst upon the scene in 2023. It’s not unreasonable to see the industry still heavily into the investment and development side of things.
Building a jet doesn’t require over a trillion dollars of capex, and selling jets is profitable. There’s solid evidence that inference isn’t profitable, and the AI labs need inference to be extremely profitable if they’re going to meet their absolutely ludicrous contractual performance targets. Oracle is expecting hundreds of billions of dollars from OpenAI by like 2030. That shit is not happening.
Whether inference is profitable or not is not a global yes/no question. It depends heavily on the circumstances, what you’re using it for and what you’re charging for it. A lot of the money being invested in research right now is going into making inference cheaper, which would of course make it more profitable to sell at current price points. Or just run it yourself, local models are getting quite capable these days.
I wouldn’t bet on any specific company being the ones to survive this, especially not first-movers like OpenAI. More likely they’ll spend their money blazing the trail and the ones to profit from it will be the ones who followed along behind. When a company goes bankrupt it doesn’t poof out of existence, its assets get sold off at pennies on the dollar and then whoever bought those assets gets a chance at running them without the overhead of the previous company’s debt.
That’s true but the pricing doesn’t match the expense by a long shot. The question is - what is going to give out first? Capital, or the all-you-can-eat pricing? (and yes the latter is closing as we speak)
This seems somewhat misleading. Lots of products take a lot of investment in them for many years before they reach profitability. The Boeing 787 Dreamliner, for example, was in development for 7 years and another three years after that before it was profitable. The Falcon 9 rocket took 13 years to develop and now it’s the most profitable satellite launcher around. The Dyson bag-free cyclonic vacuum cleaner took 15 years to develop.
Most of this AI stuff has only been in heavy development since ChatGPT burst upon the scene in 2023. It’s not unreasonable to see the industry still heavily into the investment and development side of things.
Building a jet doesn’t require over a trillion dollars of capex, and selling jets is profitable. There’s solid evidence that inference isn’t profitable, and the AI labs need inference to be extremely profitable if they’re going to meet their absolutely ludicrous contractual performance targets. Oracle is expecting hundreds of billions of dollars from OpenAI by like 2030. That shit is not happening.
Whether inference is profitable or not is not a global yes/no question. It depends heavily on the circumstances, what you’re using it for and what you’re charging for it. A lot of the money being invested in research right now is going into making inference cheaper, which would of course make it more profitable to sell at current price points. Or just run it yourself, local models are getting quite capable these days.
I wouldn’t bet on any specific company being the ones to survive this, especially not first-movers like OpenAI. More likely they’ll spend their money blazing the trail and the ones to profit from it will be the ones who followed along behind. When a company goes bankrupt it doesn’t poof out of existence, its assets get sold off at pennies on the dollar and then whoever bought those assets gets a chance at running them without the overhead of the previous company’s debt.
That’s true but the pricing doesn’t match the expense by a long shot. The question is - what is going to give out first? Capital, or the all-you-can-eat pricing? (and yes the latter is closing as we speak)