… In the long run, the AI model builders will not rent huge amounts of capacity on the cloud. They will make the clouds sell them infrastructure on the cheap that they can amortize, thereby lowering the cost of tokens. …
… Microsoft says that it will use the Maia 200 compute engines to serve inference tokens for the OpenAI GPT-5.2 large language model, driving the Microsoft Foundry AI platforms as well as Office 365 copilots. …
… We can’t assume that this commitment from OpenAI to buy capacity on the Azure cloud to train its models and serve up tokens stretches all the way put to 2032, when the exclusive licensing of the source code to the GPT class of models created by OpenAI runs out. …