Nvidia and Amazon announced an expanded partnership Wednesday that includes deploying an additional 2 million Nvidia GPUs, including Blackwell Ultra, Rubin, and Rubin Ultra chips, across Amazon Web Services data centers in 2027 and 2028. The deal follows an agreement five months earlier to deploy over 1 million GPUs, with Nvidia stating that demand had exceeded initial expectations. Nvidia CFO Colette Kress said the company would also send Vera CPUs to AWS, part of a broader integration extending to networking hardware, open models, and robotics platforms including Omniverse, Cosmos, Isaac, and Jetson, which AWS will use to power its warehouse robot fleet. The expansion comes even as Amazon continues developing its own competing chips, including Trainium and Graviton processors, which the company said recently crossed a $25 billion annualized revenue run rate. Nvidia separately reported second-quarter sales of $96.2 billion, with data center revenue reaching $89 billion, and said it had committed $279 billion toward securing future manufacturing and supply capacity.
In other company news, The Information reported that Nvidia had agreed to acquire Hugging Face for $12.9 billion, though Business Insider reported the talks remained unsigned and could still fall apart. Neither company responded to requests for comment. The acquisition would strengthen Nvidia’s position in open source AI as rivals including OpenAI, Google, and Anthropic develop their own chips to reduce dependence on Nvidia hardware. Hugging Face CEO Clem Delangue has publicly aligned with Nvidia’s open source advocacy this year, including signing a letter urging the US government to support open-weight models. The reported price marks a substantial jump from Hugging Face’s $4.5 billion valuation in 2023, and comes after the company previously turned down a $500 million Nvidia investment last year over concerns about ceding influence to a single dominant investor.
