NVIDIA H100 Rental

Hopper-class training, rented by the month.

HGX H100 SXM5 with NVLink/NVSwitch for tensor-parallel training, or PCIe Gen5 H100 for high-throughput inference. Multi-node clusters available over NDR 400G InfiniBand.

3,958 TFLOPS FP8

Per-GPU FP8 sparse throughput on Hopper — the fastest path through trillion-parameter pretraining.

640 GB HBM3

8 × 80GB HBM3 in an HGX node, all connected by 900 GB/s NVLink for true tensor parallelism.

InfiniBand multi-node

Scale out to 32, 64, or more H100s over NDR 400G InfiniBand with NCCL-tuned rail topology.

Specifications

GPU (HGX)
8 × NVIDIA H100 SXM5
80GB HBM3 each · 900 GB/s NVLink
CPU
2 × Intel Xeon Platinum 8480+
112 cores total · PCIe 5.0
Memory
2 TB DDR5
32 × 64GB @ 4800 MT/s
Storage
2 × 3.84TB + 8 × 7.68TB NVMe
Local NVMe scratch + boot
Fabric
8 × NDR 400G IB
1 IB per GPU · rail-optimized
Power
6 × 3,000W PSU
208V/415V redundant
Best fit: foundation-model pretraining

If you're training a 70B+ model from scratch, you want H100 SXM5 with NVLink. PCIe variants are better for inference, fine-tuning, and rendering.

Transformer Engine

Native FP8 with automatic mixed-precision scaling — typical 2–3x speedup over A100 FP16 for transformer training.

Multi-node tuning included

Our SRE team tunes NCCL, GPU Direct RDMA, and topology files before handover so you're not debugging fabric on day one.

Frontier-scale LLM pretraining

HGX H100 SXM5 is the platform of record for trillion-token pretraining runs. 8 GPUs share 640GB of HBM3 over a 900 GB/s NVLink/NVSwitch fabric, and our NDR 400G InfiniBand rails extend tensor- and pipeline-parallelism cleanly to 32, 64, or 256 GPUs.

Dense 70B–405B training

FP8 with Transformer Engine cuts memory pressure roughly in half vs BF16 and unlocks larger micro-batches. Llama-class dense models train end-to-end without offloading.

MoE and pipeline-parallel

NVLink + NVSwitch keeps all-to-all expert routing on-chassis. Cross-node pipeline stages ride NDR 400G IB with rail-optimized NCCL we tune before handover.

Megatron, NeMo, DeepSpeed

Validated stacks for Megatron-LM, NVIDIA NeMo, PyTorch FSDP, and DeepSpeed ZeRO-3. We ship working config files, not just a base image.

Checkpoint throughput

Local NVMe Gen5 scratch plus optional Weka/VAST tiers keep 70B+ checkpoints under a minute, so you actually use your save_steps cadence.

Production inference and agentic workloads

The same Hopper compute that trains your model serves it. FP8 inference via TensorRT-LLM, vLLM, and SGLang turns H100 into the highest-throughput option for any context window that fits in 80GB HBM3 — and PCIe variants give you a cheaper serving tier when NVLink isn't required.

vLLM + TensorRT-LLM

PagedAttention, continuous batching, and speculative decoding stacks are pre-validated. Hit production p99 latency targets without weeks of kernel tuning.

Agentic and tool-use

Long-running agent loops with high prompt churn benefit from H100's prefill throughput. Pair with Redis/KV-cache offload for multi-turn workloads at scale.

Fine-tune and serve in one rack

LoRA/QLoRA fine-tunes finish in hours on a single H100 node, then serve from the same hardware — no migration, no re-quantization surprises.

Multi-tenant model hosting

MIG partitioning splits each H100 into up to 7 isolated instances for model-as-a-service products that need hard tenant boundaries.

Privacy, security, and compliance

Frontier training runs touch proprietary data — research IP, customer transcripts, code repos, medical imaging. We rent bare metal, not slices of a multi-tenant cloud, so the only software on your node is yours.

Single-tenant bare metal

No hypervisor, no neighbor workloads, no shared GPU memory. Confidential Computing modes on H100 are available for attested workloads.

Customer-held keys

Full-disk encryption with keys you generate and rotate. We never escrow plaintext keys, and drives are sanitized to NIST 800-88 between tenants.

Air-gapped & on-prem options

For HIPAA, ITAR, FedRAMP-track, or contract-sensitive workloads we can deploy identical H100 builds in your facility or a dedicated cage you control.

Private fabric

Dedicated VLANs, no internet egress by default, and optional Direct Connect / ExpressRoute peering keep training data off the public network.

Frequently asked

Do you offer H100 SXM5 or PCIe?+

We build both. SXM5 in HGX 8-GPU configurations with NVLink/NVSwitch for tensor-parallel training, and PCIe Gen5 for inference and rendering workloads where NVLink isn't required. Pricing and lead times differ — request a quote.

What is the lead time on an H100 server?+

H100 supply remains tight. We typically quote 4–8 weeks for HGX H100 SXM5 builds and 2–4 weeks for PCIe configurations. Booking a longer term often unlocks earlier delivery slots.

Can I rent month-to-month?+

Yes, but H100 month-to-month carries a ~25% premium over a 12-month term given supply constraints. Most customers commit 12–36 months.

Do you support multi-node training over InfiniBand?+

Yes — we deploy NDR 400G InfiniBand fabrics for multi-node H100 training. Talk to a sales engineer about your topology and scale targets.

Get an H100 quote in 24 hours.

Tell us your model size, batch shape, and timeline — we'll spec the right node and fabric.

Request a quote