With AI compute capacity in high demand, even IBM is now cashing in and selling access to servers. This week, Big Blue signed a $240 million deal with Together AI to rent Nvidia hardware via IBM Cloud.

The deal will see the neocloud offer open source model inference deployments hosted on Nvidia HGX B300 servers via IBM Cloud from early 2027, a needed boost after a troubled quarter for IBM.

IBM Cloud General Manager Alan Peacock said the “scalable, economical, enterprise-grade AI infrastructure” would serve companies racing to scale agentic AI as the industry shifts away from a focus on training models.

IBM described the deal as “multi-year” but did not provide details on how long it will last. It will deploy a “large cluster” of the HGX B300 systems from Q1 2027 in IBM Cloud’s first large-scale dedicated cluster for inference.

A Hybrid Cloud strategy

Get the full story: Subscribe for free

Join peers managing over $100 billion in annual IT spend and subscribe to unlock full access to The Stack’s analysis and events.

Subscribe now

Already a member? Sign in