Cloud infrastructure provider Vultr revealed a plan to deploy 24,000 AMD Instinct MI355X GPUs at a new US data centre, a move which would significantly expand its AI compute capabilities.

Vultr is launching an AI supercluster at a data centre in the US state of Ohio.

The companies stated the 50-megawatt cluster of AMD processors is designed to deliver high performance and cost efficiency for AI training and inference.

Vultr CEO J.J. Kardwel stated by investing in the development of racked GPU capacity at scale, the company is “enabling enterprises to push the boundaries of what’s possible with AI and bring next-generation applications”.

Bloomberg reported Vultr will spend more than $1 billion on the data centre supercluster, which is slated to come online in Q1 2026.

The news agency noted Vultr’s cluster operates on a smaller scale than large data centres operated by Microsoft, Google and Meta Platforms, but offers computing power at lower rates.

In January, Verizon Business stated it is using Vultr’s GPU-as-a-service capabilities to power its AI Connect platform.

AMD Ventures co-led a financing round for Vultr in December 2024, which valued the cloud provider at $3.5 billion. 

HPE deal
In other positive news for AMD, HPE announced it reached an agreement to use its Helios AI rack-scale architecture to speed up the deployment of at-scale AI training and inferencing for cloud service providers.

HPE plans to combine AMD’s Helios system with Broadcom’s Tomahawk 6 networking chip to support traffic from trillion-parameter model training, high inference throughput and massive model sizes

Through the partnership, HPE stated its services team can support customers with flexibility, interoperability, energy efficiency, and faster deployments as the demand for AI compute capacity continues to increase.