This article reveals OpenAI's significant spending on inference through Microsoft Azure and details the complexities of their revenue-sharing agreement. The reported inference costs and revenues differ from previously stated figures, suggesting that OpenAI's financial situation may be more complicated than understood. The analysis challenges the accuracy of OpenAI's claimed revenues.

OpenAI has partnered with Cerebras to deploy 750 megawatts of wafer-scale AI systems, marking the largest high-speed AI inference initiative. This collaboration aims to enhance AI performance and accessibility, delivering responses up to 15 times faster than traditional GPU systems.

OpenAI has adopted a new data type called MXFP4, which significantly reduces inference costs by up to 75% by making models smaller and faster. This micro-scaling block floating-point format allows for greater efficiency in running large language models (LLMs) on less hardware, potentially transforming how AI models are deployed across various platforms. OpenAI's move emphasizes the efficacy of MXFP4, effectively setting a new standard in model quantization for the industry.