NVIDIA GB300 NVL72
Blackwell · 5-Year Contract
/ GPU / hr
- Memory
- 20.7 TB
- Interconnect
- 800 Gb/s NVLink 5
- Cooling
- Liquid DLC
- Location
- Penang, Malaysia
Compute Resources
GPU Instances & Clusters
On-demand, reserved, and bare-metal cluster options across data centers in Tokyo, Yokohama, Penang, Georgia and Seoul. Powered by NVIDIA Blackwell and Ada Lovelace architectures.
Live
Live GPU Availability
Real-time compute inventory across all data centers
Available
NVIDIA B300
Blackwell · Reserved Instance
/ GPU / hr
- Memory
- 8 TB
- Interconnect
- 800 Gb/s IB
- Cooling
- Liquid / Air
- Location
- Yokohama · Georgia · Seoul
NVIDIA RTX PRO 6000 Blackwell
Blackwell · On-Demand
On request
- Memory
- 96 GB GDDR7
- Interconnect
- PCIe Gen5
- Cooling
- Air
- Location
- Tokyo, Japan
NVIDIA L40S
Ada Lovelace · On-Demand
On request
- Memory
- 48 GB GDDR6
- Interconnect
- PCIe Gen4
- Cooling
- Air
- Location
- Tokyo, Japan
Compute Resources
View Specs
GB300
NVIDIA GB300 NVL72
$6.80
/ GPU / hr
5-Year ContractPenang, MalaysiaAvailable
- Architecture
- Blackwell
- Memory
- 20.7 TB
- Interconnect
- 800 Gb/s NVLink 5
- Cooling
- Liquid DLC
- GPUs / node
- 72
- PUE
- 1.15
B300
NVIDIA B300
$5.10
/ GPU / hr
Reserved InstanceYokohama · Georgia · SeoulAvailable
- Architecture
- Blackwell
- Memory
- 8 TB
- Interconnect
- 800 Gb/s IB
- Cooling
- Liquid / Air
- GPUs / node
- 8
- PUE
- 1.15
RTX PRO 6000
NVIDIA RTX PRO 6000 Blackwell
On request
On-DemandTokyo, JapanAvailable
- Architecture
- Blackwell
- Memory
- 96 GB GDDR7
- Interconnect
- PCIe Gen5
- Cooling
- Air
L40S
NVIDIA L40S
On request
On-DemandTokyo, JapanAvailable
- Architecture
- Ada Lovelace
- Memory
- 48 GB GDDR6
- Interconnect
- PCIe Gen4
- Cooling
- Air
GPU Compute
Full-Stack AI Infrastructure
One connected foundation for inference.
On-demand
Nexora Public Cloud
Developer-friendly on-demand compute. Launch Jupyter Notebooks and ML environments in under 60 seconds.
- Launch time
- Under 60 s
- Environment
- Jupyter Notebook · ML
- GPU
- RTX PRO 6000 · L40S
Bare metal
Private AI Clusters
Bare-metal clusters for Tier-1 clients. Scale from 3,200 to 5,000 GPUs with 800G InfiniBand fabric.
- Scale
- 3,200 – 5,000 GPU
- Network
- 800G InfiniBand
- GPU
- GB300 · B300
Fully managed
Managed AI Factory
Fully managed operations. 800G IB networking, NVMe tiered storage, and Slurm orchestration included.
- Network
- 800G InfiniBand
- Storage
- NVMe tiered
- Orchestration
- Slurm