Reference Solutions

Proven cluster architectures for AI training and inference — from single-node development to 512-GPU production deployments.

1-Node

1-Node Production Pod

8 GPUs

Compact production-ready pod with full networking stack.

1
Nodes
8
GPUs
800G InfiniBand
Network
100 TB NVMe
Storage
Power: ~14kWLead Time: 4-6 weeks
Start Consultation for This Config
8-Node

8-Node Pod

64 GPUs

Compact high-performance pod for mid-size training workloads.

8
Nodes
64
GPUs
800G InfiniBand
Network
200 TB
Storage
Power: ~112kWLead Time: 8-12 weeks
Start Consultation for This Config
32-Node

32-Node Pod

256 GPUs

Multi-rack scale-out cluster for large-scale training.

32
Nodes
256
GPUs
800G InfiniBand
Network
500 TB
Storage
Power: ~448kWLead Time: 12-16 weeks
Start Consultation for This Config
64-Node

64-Node Production Cluster

512 GPUs

Maximum-density production cluster with next-gen B300 GPUs.

64
Nodes
512
GPUs
800G InfiniBand
Network
1000 TB
Storage
Power: ~900kWLead Time: 16-20 weeks
Start Consultation for This Config

Need a Custom Configuration?

Our solutions engineers can design a bespoke cluster tailored to your workload, facility constraints, and budget.