NVidia is the one selling shovels to the golddiggers, a business that always was a profitable one. If AI goes bust, they can go back to selling GPUs to gamers.
Microsoft makes profit because they get paid by OpenAI for their Azure platform.
Facebook makes profit because of ads? Don’t know why they were even mentioned here, their AI experiments seem mainly invested in making better bots for their platforms.
No AI Company is profitable in any way, and barring some technological breakthrough, they never will be. Currently it’s all subsidized with huge dollar amounts by venture capitalists. The problem is that inference is expensive (You need a specialized datacenter for Nvidias new hardware, the hardware itself is expensive, the training of models costs a lot and you can never stop training new ones just to keep up with the joneses, you need lots of power to run all of this, and the GPUs used have to be replaced on the regular).
At the moment it all gets propped up by the largest financing rounds ever, and that is without even counting the Venture Capital that gets funneled to OpenAI/Anthropic through AI startups.
Especially the subscription model will disappear very soon. Users with 200$ Subscriptions blow through tokens costing a multitude of what they pay. API pricing per token will be the default, and those prices will make every single AI startup weep as soon as the subsidies disappear, and then close up shop, which will be the beginning of the end for OpenAI/Anthropic.
NVidia is the one selling shovels to the golddiggers, a business that always was a profitable one. If AI goes bust, they can go back to selling GPUs to gamers.
Microsoft makes profit because they get paid by OpenAI for their Azure platform.
Facebook makes profit because of ads? Don’t know why they were even mentioned here, their AI experiments seem mainly invested in making better bots for their platforms.
No AI Company is profitable in any way, and barring some technological breakthrough, they never will be. Currently it’s all subsidized with huge dollar amounts by venture capitalists. The problem is that inference is expensive (You need a specialized datacenter for Nvidias new hardware, the hardware itself is expensive, the training of models costs a lot and you can never stop training new ones just to keep up with the joneses, you need lots of power to run all of this, and the GPUs used have to be replaced on the regular).
At the moment it all gets propped up by the largest financing rounds ever, and that is without even counting the Venture Capital that gets funneled to OpenAI/Anthropic through AI startups.
Especially the subscription model will disappear very soon. Users with 200$ Subscriptions blow through tokens costing a multitude of what they pay. API pricing per token will be the default, and those prices will make every single AI startup weep as soon as the subsidies disappear, and then close up shop, which will be the beginning of the end for OpenAI/Anthropic.