Ori Global Cloud has launched a Private Cloud cluster made up of Nvidia H100 GPUs.

The cluster features 1,024 Nvidia H100s, and is based on the Nvidia DGX SuperPOD reference architecture.

Comprised of 128 Nvidia HGX nodes, the cluster also has one petabyte of Weka parallel storage with an aggregate of nine million read and three million write operations per second.

The nodes are connected by the 3.2 Tbps Infiniband GPU interconnect, and have a storage connectivity of 400 Gbps.

According to Ori Global, during the company’s testing of the cluster it sustained a performance of 32 petaflops over the course of several days operating at around 95 percent of its theoretical peak.

Ori Industries CEO Mahdi Yahya said: “Ori’s Private Cloud offering is designed to enable AI businesses with the best technology stack to help them realize their AI ambitions without compromising on performance, scale, and security. Our latest GPU cluster is a testament to our expertise and confidence in delivering custom private clouds for large-scale AI projects.”

Ori’s Private Cloud also offers access to Nvidia H200s, and expects the Blackwell GPUs to be available soon.

Ori Industries was founded in 2018 and is based in London, UK. The company didn’t say where the cluster was deployed, but the company has previously announced deals with data center providers EdgeConneX in Europe and Centersquare in the US. It offers services from 20 locations globally.

Other cloud providers to have recently added H100 clusters include IBM Cloud and Denvr Dataworks.

https://www.datacenterdynamics.com/en/news/ori-deploys-private-cloud-cluster-with-1024-nvidia-h100s/