H100 Node

Flagship Hopper performance for frontier training and low-latency serving.

Compute Plans billed per month Most Popular

A dedicated NVIDIA H100 80 GB SXM node — the standard for frontier training runs and sub-100ms serving of large models. NVLink and an InfiniBand-ready fabric keep multi-node scaling an option, not a rewrite.

What is included

Dedicated NVIDIA H100 80 GB (SXM)
48 vCPU · 512 GB RAM · 4 TB NVMe
NVLink · InfiniBand ready
99.9% uptime SLA
Lowest-latency serving for large models

Specification

GPU: NVIDIA H100 SXM
VRAM: 80 GB HBM3
CPU: 48 vCPU
Memory: 512 GB
Storage: 4 TB NVMe

Other Compute Plans

RTX 4090 Instance $349.00

A dedicated RTX 4090 for inference, LoRA fine-tuning, and development work.

A100 Node $899.00

Production-grade A100 capacity for fine-tuning and batch inference.

H100 8-GPU Cluster $9,999.00

An 8×H100 NVLink cluster for distributed training at scale.

Order summary

Plan H100 Node

Billing per month

Unit price $1,499.00

Payment is arranged after the order: bank transfer, corporate card, or a secure payment link from our billing team. Online card checkout opens to customers shortly.

Not sure this is the right size?

Tell us the workload and we will recommend the smallest plan that fits it — over-buying credits is the most common mistake we see.

Talk to an engineer

Start routing in minutes

One endpoint. Every frontier model.

Create an account, pick a plan, and point your existing OpenAI client at our base URL. No SDK rewrite, no lock-in.

Browse plans Talk to sales