Saudi Arabia's GPU infrastructure layer. Built in Riyadh, engineered for AI workloads. Every byte stays in the Kingdom.
Saudi AI teams are forced to choose between foreign hyperscalers and CPU-only local hosts. Neither works.
Hyperscaler GPU capacity sits in foreign regions: 80 to 150 ms round-trip to Riyadh enterprises.
PDPL violations occur when training runs cross borders. Personal data cannot legally leave the Kingdom.
Local providers run CPU-only fleets: zero A100s, zero H100s, no InfiniBand, no GPU-aware scheduling.
Multi-week GPU quota queues, egress fees up to $0.09/GB, and no sovereign inference runtime layer.
Full-stack sovereign compute. All layers hosted in Tier III data centers in Riyadh with single-digit millisecond latency to enterprise networks.
H100 and A100 nodes with NVLink · 3.2 Tbps InfiniBand fabric for distributed training
L4 single-GPU to 8× H100 pods · provisioned in under 90 seconds via API
GPU-aware K8s, auto MIG partitioning, node autoscaling, PDPL-compliant storage
Llama, Mistral, Falcon, Jais and AceGPT behind autoscaling HTTPS endpoints
Sovereignty enforced at the storage, key-management, and network layer.
AES-256 with HSM key custody held entirely in-Kingdom
VPC peering and private interconnects to major Saudi ISPs
99.9% uptime with a transparent public status page
Data residency enforced by architecture
Class C readiness for government workloads
Full control mapping documented
Active certification roadmap
Single-digit milliseconds to Riyadh enterprises
Historic AI data-center buildout, Cloud First Policy, PDPL enforcement, and GPU demand exceeding supply tenfold.
SAR billing. Zero egress fees. Reserved discounts up to 40% on 12 to 36 month terms.
| GPU SKU | On-Demand | Reserved | Primary Use Case |
|---|---|---|---|
| L4 | $0.90/hr | −40% | Inference experiments · on-demand |
| A100 | $2.40/hr | −40% | Fine-tuning and training · on-demand + reserved |
| H100 | $4.50/hr | −40% | Large-scale distributed training · reserved priority |
Four structural advantages no hyperscaler or legacy host can replicate locally.
Every byte stays in the Kingdom: a technical guarantee enforced at the storage, key management, and network layer.
Purpose-built for training and inference workloads, not retrofitted from general-purpose hosting.
Console, docs, and support in Arabic. SAR pricing. Native Jais and AceGPT inference endpoints out of the box.
Signup to a running Arabic LLM endpoint in under 5 minutes. API-first provisioning in under 90 seconds.
Three phases from launch to Kingdom-wide AI infrastructure.
GPU VPS and Inference Engine live in Riyadh. API-first onboarding, token billing, Arabic LLM endpoints.
Reserved capacity contracts. Managed Kubernetes general availability with PDPL-compliant storage.
Second availability zone in the Kingdom. Arabic AI Model Marketplace for curated regional LLMs.
Technology entrepreneur with deep expertise in cloud infrastructure, virtualization, and hosting operations at enterprise scale. Building the Kingdom's default AI infrastructure layer from Riyadh.
Early access is open for Saudi AI teams, enterprises, and research labs. Reserve GPU capacity before general availability.
Request Early Access