Next-generation spotlight
Early buy — delivery March 2027
VR200 NVL72
Plan your next cycle with the Rubin generation
- Memory per GPU
- 288 GB HBM4
- Inside the node
- NVL72 rack · liquid-cooled
- Node interconnect
- ConnectX-9 · 1.6T XDR
GPU COMPUTE
NVIDIA GPUs for every stage of AI.
Run training, fine-tuning, and inference on dedicated systems. Compare 8-GPU nodes and rack-scale configurations with high-speed networking and transparent reservation pricing.
06From research to results
Your next discovery. Your next product. Your next scale.
Train what comes next
Give your foundation models the capacity they demand. Reserved clusters, dedicated networking, and spare nodes support the infrastructure behind your most ambitious training runs.
Explore B200, GB200, and GB300Turn knowledge into an advantage
Adapt models to your data and business challenges. Dedicated 8-GPU nodes and persistent storage give your team room to test, fine-tune, and evolve.
Find your H100 or H200Put your AI to work
Bring large models to production with memory-rich GPUs. Serve long-context applications on a dedicated infrastructure foundation you can plan around.
Discover H200 and B300
01Find your compute
Your next AI breakthrough starts with the right GPU.
Training, fine-tuning, or inference: every workload has its own demands. Find the capacity that fits your team, your model, and your budget.
See all pricing
NVIDIA NVL72
GB300 NVL72
Capacity for MoE models and reasoning AI
288 GB HBM3e/GPUMemory per GPU
- Inside the node
- Whole or half rack · liquid-cooled
- Node interconnect
- Quantum-X800 XDR between racks
Explore the configuration
GPU interconnect72-GPU coherent NVLink 5 domain
Rack-scale
NVIDIA HGX
B300
Memory-rich compute for large models and inference
288 GB HBM3eMemory per GPU
- Inside the node
- 8 GPU · 30 TB NVMe · 2 TB RAM
- Node interconnect
- Quantum-X800 XDR 800G
Explore the configuration
GPU interconnectNVLink 5 · 1.8 TB/s per GPU
Blackwell Ultra
NVIDIA NVL72
GB200 NVL72
A 72-GPU foundation for your largest models
186 GB HBM3e/GPUMemory per GPU
- Inside the node
- Whole or half rack · liquid-cooled
- Node interconnect
- Quantum-X800 XDR between racks
Explore the configuration
GPU interconnect72-GPU coherent NVLink 5 domain
Rack-scale
NVIDIA HGX
B200
Frontier model training and FP4 inference
180 GB HBM3eMemory per GPU
- Inside the node
- 8 GPU · 30 TB NVMe · 2 TB RAM
- Node interconnect
- Quantum-X800 XDR 800G
Explore the configuration
GPU interconnectNVLink 5 · 1.8 TB/s per GPU
New — commissioning
NVIDIA HGX
H200
More memory for long-context model serving
141 GB HBM3eMemory per GPU
- Inside the node
- 8 GPU · 30 TB NVMe · 2 TB RAM
- Node interconnect
- 8×400G IB NDR · 3.2 Tbps
NVIDIA HGX
H100
Training and fine-tuning with your budget in mind
80 GB HBM3Memory per GPU
- Inside the node
- 8 GPU · 15 TB NVMe · 2 TB RAM
- Node interconnect
- 8×400G IB NDR · 3.2 Tbps
Prices in USD per 8-GPU node-hour. Rack-scale systems are quoted separately. Subject to capacity and contract terms.
- 0%
- Hypervisor overhead
- 3.2 Tbps
- InfiniBand per node
- 97.4%
- Cluster goodput, 90 days
TESTED BEFORE HANDOVER
Start with a cluster you can trust.
Every cluster goes through 72 hours of load, memory, network, and thermal tests. Your team reviews the results before approving delivery.
72h
ACCEPTANCE TESTING
Load · memory · network · thermal checks
03Infrastructure behind your ambition
Every breakthrough needs a foundation to match.
Power, cooling, GPUs, and networking. We design and operate every layer together to give your AI a foundation ready to grow.
- 01Cooling
PUE 1.12
Direct-to-chip liquid cooling in 130 kW racks. Engineered for the demands of intensive AI workloads.
- 02GPU capacity
24,000+ GPUs
Capacity expands in six-week cycles, with readiness checks before each handover.
- 03Useful work
97.4%
Cluster goodput over the last 90 days: the share of capacity translated into useful work.
04Networking and storage
Your GPUs. Working as one.
Bring compute, data, and storage together in infrastructure dedicated to your model, from the first training batch to the final checkpoint.
Plan my clusterDedicated networking
A network built for your GPUs.
Dedicated InfiniBand connects your cluster with bandwidth for distributed training and data exchange between GPUs.
800Gb/s
per InfiniBand XDR link, Quantum-X800 for Blackwell
Network specifications
- Training fabric — Hopper
- 8×400 Gb/s InfiniBand NDR per node (3.2 Tbps), rail-optimized non-blocking fat-tree, SHARP in-network reduction
- Training fabric — Blackwell
- Quantum-X800 InfiniBand XDR, 800G per link; GB200 racks add a 72-GPU coherent NVLink 5 domain at 1.8 TB/s per GPU
- Oversubscription
- 1:1, all tiers. No blended east-west fabric, no shared spine with other tenants
- Front-end network
- Dual 100 GbE per node, DDoS-protected, BYO-IP supported
Storage
Keep your data in stride.
Local NVMe, WEKA, and S3-compatible storage to load datasets and save checkpoints at the scale your project needs.
Up to
720GB/s
aggregate read throughput per pod, with dedicated WEKA
Storage specifications
- Local scratch
- 30 TB NVMe Gen5 per node, ~55 GB/s read — checkpoint staging without touching the network
- Parallel filesystem
- Managed WEKA, dedicated per cluster: up to 720 GB/s aggregate read per pod, POSIX + GPUDirect Storage
- Object storage
- S3-compatible, NVMe-cached, co-located with compute — $0.055/GB-month hot tier
Data transfer
Take your data further.
Free ingress and traffic between nodes, with an included egress allowance to help you plan your operation.
20TB
egress included per node, per month
Data transfer terms
- Data transfer
- Ingress $0 · 20 TB/node-month egress included, then $1.25/TB · inter-node $0. Free 100G Direct Connect on yearly reservations
EXPLORE TENSORBAY
Compute for what comes next.
Explore the GPU systems, infrastructure, and pricing that can move your AI project forward.
EARLY RESERVATIONS
NVIDIA VR200 NVL72
Review the Vera Rubin rack-scale configuration and its early reservation timeline.
NVIDIA VR200 NVL72
GPU COMPUTE
Find the right NVIDIA GPU
Compare TensorBay GPU configurations from the newest generation to the previous ones.
Find the right NVIDIA GPU
PRICING
Plan with published rates
Compare monthly and yearly reservations for GPU nodes and rack-scale systems.
Plan with published rates05Invest in your next move
Compute to grow. Clarity to invest.
Choose your reservation term, compare capacity, and plan your investment. Here, the conversation starts with open pricing.
| GPU and memory | Monthly termCapacity reserved on a monthly term | Yearly termLower hourly rate |
|---|---|---|
NVIDIAHGX B300288 GB HBM3e per GPU | Monthly from Quote monthlyUS$ 83.20/h | Yearly from Quote yearlyUS$ 58.50/h |
NVIDIAHGX B200180 GB HBM3e per GPU | Monthly from Quote monthlyUS$ 72.80/h | Yearly from Quote yearlyUS$ 50.70/h |
NVIDIAHGX H200141 GB HBM3e per GPU | Monthly from Quote monthlyUS$ 44.20/h | Yearly from Quote yearlyUS$ 31.20/h |
NVIDIAHGX H10080 GB HBM3 per GPU | Monthly from Quote monthlyUS$ 28.60/h | Yearly from Quote yearlyUS$ 19.50/h |
The hourly rate covers the complete 8-GPU node. Choose a term to request your quote.
- SLA
- 99.9% monthly + 15-min node replacement + goodput credits
- Scale
- 128–10,000+ GPUs · dedicated fabric per tenant
- Billing
- Per-minute metering inside the reservation · $0 ingress
Every detail accounted for
Your infrastructure. No blind spots.
See what is included and explore rates for storage, data transfer, and networking.
Talk to a specialistStorage
- Local NVMe scratch — 30 TB per nodeIncluded
- Parallel filesystem — managed WEKAdedicated per cluster$0.09/GB-mo
- Object storage — hot, S3-compatibleNVMe-cached, co-located$0.055/GB-mo
- Object storage — archive$0.015/GB-mo
- Snapshots$0.021/GB-mo
Data transfer
- Ingressalways$0
- Egress included20 TB/node-mo
- Egress overageall sites$1.25/TB
- Inter-node / east-westInfiniBand fabric$0
Networking
- East-west InfiniBand fabricdedicated per clusterIncluded
- Private VLAN / VPCIncluded
- Additional public IPv4IPv6 free$4/IP-mo
- BYO-IP — /24 v4 or /48 v6$200 setupFree
- Managed firewall$4/node-mo
- DDoS protectionIncluded
- Direct Connect 10G / 100G100G free on yearly terms$450 / $1,800-mo
USD prices per 8-GPU HGX node-hour; rack-scale systems are quoted per rack. Taxes excluded. Managed Slurm and Kubernetes control planes are included. Storage and Direct Connect follow the rates above. Ask about long-term contract options.
Get my proposalIs GPU pricing per GPU or per node?
Published HGX prices are in USD per hour for a complete node with 8 GPUs, within a monthly or yearly reservation. NVL72 systems are quoted per rack. Taxes are not included.
Which NVIDIA GPU should I choose for my AI project?
Compare GPU memory, network requirements and your training or inference workload. The hardware section lists H100, H200, B200, B300 and full rack systems. Our team can help size a configuration based on your model, dataset and capacity requirements.
Can I use a dedicated cluster for training, fine-tuning and inference?
Yes. TensorBay offers dedicated NVIDIA GPU infrastructure for model training, fine-tuning and inference. You can work with Slurm, Kubernetes or SSH and configure the software environment for your workload.
What happens after I request a quote?
Share the GPU, capacity and reservation term you need, or describe your project. Our team reviews the requirements and prepares a proposal within 48 hours, including configuration and delivery timing. Capacity and final terms are confirmed in the proposal.
08Give your idea room to grow
Your next AI breakthrough starts with a conversation.
Bring us your challenge. Our team will help define your GPUs, capacity, and timeline. Get a proposal within 48 hours to launch your project or take it further.
- A configuration built around your project
- A proposal with a delivery timeline
- Acceptance testing before you start