GPUs by the hour.
Sovereign by default.
Reserved and on-demand access to NVIDIA H100, H200, and B200 clusters in dedicated immersion-cooled pods, with data residency in the Kingdom and sovereign-grade access controls.
The offering
AI compute, productized.
For sovereign AI initiatives, AI labs, and cloud resellers, we offer GPU compute as a productized service: reserved and on-demand cluster capacity in our immersion-cooled pods, with KSA data residency, in-region object storage, parallel filesystems, and the AI fabric (InfiniBand HDR/NDR) you need for serious training. Bring your own software stack, run our managed Kubernetes, or run our managed Slurm — your choice.
- Reserved capacity from 8 to 1,024+ GPUs per cluster.
- On-demand cluster spin-up for development and elastic inference.
- KSA data residency by default. Sovereign access controls available.
- Bring-your-own-software or managed Kubernetes/Slurm.
Technical specifications
What you're committing to.
- GPU types
- H100 SXM, H200, B200 (rolling availability)
- Per-GPU memory
- 80 GB (H100), 141 GB (H200), 192 GB (B200)
- Inter-GPU fabric
- NVLink within node, InfiniBand HDR/NDR cluster-wide
- Storage
- Lustre / BeeGFS parallel FS, S3-compatible object storage
- Pricing model
- Reservation (1m/3m/12m), on-demand hourly, spot
- Access controls
- OIDC SSO, IAM, role-based, audit-logged
- Compliance
- KSA resident, SAMA-aligned, NCA-aligned
- Sovereign tier
- Air-gapped option, Saudi-cleared operations
What's included
The full AI compute stack.
GPU cluster
NVIDIA H100, H200, or B200 in dedicated immersion pods. Sustained boost clocks. No throttling.
InfiniBand fabric
Non-blocking InfiniBand HDR/NDR fabric across the cluster. RDMA-grade latency for distributed training.
Parallel filesystem
Lustre or BeeGFS available. Tens of GB/s sustained throughput for training data ingestion.
Software stack
Bring-your-own, managed Kubernetes (NVIDIA GPU Operator), or managed Slurm with priority queues and accounting.
Spot and reservation
Reserved capacity for production workloads. Spot capacity for development and bursty inference at significant discount.
Data residency
All compute, storage, and metadata KSA-resident by default. Sovereign access tier available with Saudi-cleared operations.
Cluster sizes
Sized for what you actually need.
8–32 GPUs
Development clusters, small-team fine-tuning, evaluation pods. On-demand or weekly reservation.
64–256 GPUs
Mid-scale training clusters for fine-tuning foundation models, domain adaptation, and large-scale inference.
512–1,024+ GPUs
Foundation-model training scale. Multi-month reservations with dedicated technical account management.
Capacity is being committed daily.
Reserve yours.
Tour the facility. Talk to engineering. Get a precision quote. We respond to qualified inquiries within one business day.
Olaya District, Riyadh · Open for tours by appointment