Skip to main content

1-Click Clusters

Deploy GPU clusters from 16 to 1,500 GPUs with a single click. Designed for large-scale AI training, distributed computing, and research at scale.

Overview​

1-Click Clusters provide self-serve access to large GPU deployments without the complexity of traditional HPC provisioning. Select your cluster size, choose your commitment level, and launch.

Cluster Configurations​

NVIDIA B200 Clusters​

CommitmentPrice per GPU/hrCluster Sizes
On-demand (week)$3.7916 - 512 GPUs
On-demand (month)$3.4916 - 1,024 GPUs
1-year reserved$3.2932 - 1,500 GPUs
3-year reserved$2.9964 - 1,500 GPUs

NVIDIA H100 Clusters​

CommitmentPrice per GPU/hrCluster Sizes
On-demand (week)$2.6916 - 512 GPUs
On-demand (month)$2.2916 - 1,024 GPUs
1-year reserved$1.8532 - 1,500 GPUs

Features​

High-Speed Interconnect​

  • 400 Gbps per GPU with NVIDIA Quantum-2 InfiniBand
  • Non-blocking, full-bisection bandwidth
  • Optimized for distributed training frameworks (DeepSpeed, FSDP, Megatron)

S3-Compatible Storage​

  • Built-in S3-compatible storage adapter
  • Store datasets, checkpoints, and model weights
  • Accessible from all nodes in the cluster

Real-Time Inference​

  • Serve models at 10,000+ tokens per second
  • Low-latency inference endpoints
  • Auto-scaling capabilities

Pre-Configured Software​

  • CUDA, cuDNN, NCCL pre-installed
  • PyTorch, TensorFlow, JAX ready
  • Slurm or Kubernetes orchestration included

Use Cases​

Use CaseRecommended Config
LLM pre-training (70B+)64-512x H100 or B200
Distributed fine-tuning16-64x H100
Multi-node inference16-32x H100
Research at scale32-128x A100 or H100

Getting Started​

1-Click Clusters are available for qualified accounts. Contact us via X / Twitter to discuss your requirements.