Company News
General availability of Virtual Machines with NVIDIA GPUs
General availability of Virtual Machines with NVIDIA GPUs (H100, A100, V100) in Ori Global Cloud.
Meet Ori Global Cloud's new Private Cloud cluster with 1024 NVIDIA H100 GPUs, designed for massive scale AI with limitless customization.
Today, we‘re excited to share that Ori has deployed a Private Cloud cluster of 1024 NVIDIA H100 GPUs with state-of-the-art performance, storage and networking capabilities, for a large AI startup with a global footprint.
According to Forrester's 2023 Infrastructure Cloud Survey, 79% of enterprises reported that they are implementing internal private clouds. As businesses increase AI spending, private clouds are emerging as a top choice due to their superior cost-efficiency, flexible tech stack and robust data security.
Ori Global Cloud is driving the Private AI Cloud of the future by fusing simplicity with performance, competitive pricing, and unparalleled flexibility, empowering AI builders and businesses to scale AI/ML models with complete control over their data.
Purpose built infrastructure for massive scale AI
Based on the NVIDIA DGX SuperPOD reference architecture, this new Private Cloud cluster combines:
- 128x NVIDIA HGX nodes comprising of 8 H100 SXM GPUs each.
- 1 Petabytes of Weka parallel storage that delivers an aggregate of 9 million read and 3 million write operations/sec (IOPS).
- 3.2 Tbps Infiniband GPU interconnect, and storage connectivity of 400 Gbps Ethernet.
- Stelia HyperBand™, a petabit-per-second data mobility platform that connects the cluster with the rest of the world.
During our tests, this cluster sustained a performance of 32 PFlops over the course of several days (operating at about 95% of its theoretical peak performance of 34 PFlops). After going through intense testing, the cluster is now successfully used in production and delivers state-of-the-art AI and ML performance, enabling its users to train large language models (LLMs) over the course of several weeks at full speed and without failure.
Take a peek at our new Private Cloud cluster
Chart your own AI reality with Ori Private Cloud
Ori’s Private cloud is designed for AI-focused startups and enterprises to realize their AI ambitions with the AI infrastructure they need. Whether you want to train foundation models or run super-scale inference, you have the flexibility to maximize your infrastructure without compromises:
- Large clusters of top-tier GPUs such as NVIDIA H100 and H200, with NVIDIA’s next-gen Blackwell GPUs to be available soon.
- Superfast networking and storage that harnesses the power of thousands of GPUs into an AI supercomputer.
- Massive-scale AI computing without the stress of huge capital expenditures, and large infrastructure teams.
