IBM and Together AI Sign $240 Million Agreement for Open-Source AI Inference Cluster
IBM and Together AI have signed a $240 million agreement to deploy a large NVIDIA-powered open-source AI inference cluster on IBM Cloud, expected to be available in Q1 2027.
IBM and Together AI have signed a $240 million, multi-year agreement to build a large-scale open-source AI inference cluster on IBM Cloud. The system will use NVIDIA HGX B300 infrastructure and is expected to be available in the first quarter of 2027, according to IBM’s announcement of the agreement and planned deployment.
The arrangement is aimed at inference rather than the training process that produces a model. Together AI will run inference for open-source models on the cluster, giving those models access to a substantial pool of NVIDIA hardware through IBM’s cloud platform. The distinction matters: once a model exists, serving responses to users and applications at scale becomes an infrastructure problem of its own.
The hardware will be built around NVIDIA HGX B300 systems, with Spectrum-X networking connecting the cluster. Quartz independently reported the same $240 million agreement and the planned use of HGX B300 systems and Spectrum-X, with availability also expected in Q1 2027. The two announcements describe the same basic effort: IBM supplying the cloud infrastructure and Together AI operating inference workloads for open-source models.
That division of responsibility makes the deal more than a hardware order. IBM is committing its cloud to a dedicated inference deployment, while Together AI is bringing the model-serving work that turns computing capacity into an actual service. The agreement therefore links the infrastructure layer and the model-access layer in one project, rather than treating open-source AI as something that ends when a model is released.
The timing is still prospective. The cluster is expected to become available in the first quarter of 2027, so the immediate development is the agreement itself and the decision to build around NVIDIA’s HGX B300 systems and Spectrum-X networking. If the deployment arrives as planned, it will give Together AI a purpose-built IBM Cloud platform for serving open-source models at the scale this $240 million commitment implies.