Amazon and Nvidia announced an expanded partnership on August 20, 2024, adding 2 million Nvidia GPU chips to Amazon Web Services data centers beginning in 2027. The previous agreement made five months earlier committed more than 1 million GPUs to AWS for 2024. The new GPUs include Blackwell Ultra, Rubin and Rubin Ultra models, which are designed for large‑scale AI training and inference. The partnership also integrates Nvidia networking hardware, software, AI models, CPUs, data‑processing tools and robotics platforms across AWS and other hyperscalers. Amazon is developing its own AI chips, including Trainium accelerators for deep‑learning workloads and Graviton CPUs that compete with Intel and AMD. AWS is discussing the sale of Trainium chips to other companies, according to AI chief Peter DeSantis. Nvidia reported Q2 revenue of $96.2 billion, with data‑center sales of $89 billion, up 117 percent year over year. The company projects Q3 revenue of $108 billion, part of which will come from Rubin GPUs. Nvidia has committed $279 billion to secure supply and manufacturing capacity through 2028, up from $119 billion in the prior quarter. Jensen Huang said AI is now generating profitable tokens and that additional compute could increase profit for all participants, describing a new $200 billion total addressable market for Vera CPUs. Nvidia's physical AI stack, including Omniverse, Cosmos, Isaac and Jetson, will be adopted by Amazon to power its warehouse robots and enterprise services.