IBM Signs $240 Million Deal with Together AI

IBM has announced a multi-year, $240 million agreement with Together AI, which specializes in cloud infrastructure for AI, to deploy AI infrastructure on IBM Cloud to meet high demand for AI computing capacity. The project centers on a computing cluster based on NVIDIA HGX B300 systems, equipped with Blackwell processors and Spectrum-X networking, with the first phase including about 2,000 Blackwell 300 GPUs installed in the United States. This infrastructure will be primarily dedicated to inference—the execution of already-trained models to produce responses, generate content, or power AI agents in production. Inference has become the true driver of computing power demand, pushing cloud providers and chipmakers to invest heavily. For Together AI, the deal offers a chance to accelerate its services around open-source models; founded in 2022, the company specializes in their deployment at scale, with models like DeepSeek, MiniMax, or Kimi, and claims over 400,000 billion tokens processed monthly, having recently raised $800 million at an $8.3 billion valuation. For IBM, the partnership aims to strengthen its cloud and enterprise AI offerings, competing with AWS, Microsoft Azure, and Google Cloud. The cluster is expected to be operational in the first quarter of 2027, with a significant portion of its capacity possibly reserved months before launch.