Vultr Orders HPE 'AMD Helios' Systems in $1.2 Billion Deal

Vultr has placed a $1.2 billion order for 'AMD Helios AI Rack' systems manufactured by HPE, market data showed. This marks the first order of its kind for AMD's new 'Helios' platform, and it will be deployed in Vultr's cloud data centers in the United States.

The systems include scale-up networking from HPE, which connects chips to one another within a single rack to accelerate data exchange between them. The order targets AI model training and inference workloads—the most compute-intensive tasks in modern data centers.

The size of the order highlights the accelerating spending on AI infrastructure among specialized cloud service providers, who are expanding their capacity to accommodate the growing demand for running models. The deal also gives the 'Helios' platform its first foothold in the competitive market for integrated AI racks.

The order comes at a time when manufacturers and chip providers are vying for market share in AI hardware, which has become one of the largest areas of capital expenditure in the technology sector. Such deals typically impact supply chains, hardware pricing, and the available capacity for enterprises to run their own models.

What do these terms mean?

AI Rack: An integrated unit containing compute chips, networking, and cooling in a single enclosure, installed in a data center to run AI models instead of assembling individual components. Training: The phase of teaching the model using vast amounts of data. Inference: The phase of using the trained model to respond to requests. Scale-up Networking: High-speed interconnects that link chips within the same system so they operate as a single unit, as slow connections waste processing power.