Core42 AI Cloud is a full-stack, AI-native cloud platform built for the full intelligence lifecycle, from training and fine-tuning to production-grade inference. The platform is anchored by two core services: GPU as a Service, providing direct access to a diverse range of accelerators via bare metal, Kubernetes, or Slurm orchestration, and Compass Inference as a Service, enabling teams to deploy and scale models with low latency, enterprise-grade performance, and built-in scalability.
The platform is engineered for accelerator choice without lock-in, supporting NVIDIA, AMD, Cerebras, Qualcomm, and Microsoft silicon so teams can match the right hardware to each workload. AI-optimized storage delivers fast, reliable access for AI and HPC data, designed for large-scale training and high-concurrency workloads without becoming a bottleneck and engineered with enterprise-grade resiliency for operational continuity.
Performance is validated by independent benchmarks. The Core42 Maximus-01 (AMD MI300X) system in the US ranks #20 worldwide on the Top500 HPC list and #3 worldwide on the IO500 storage benchmark. In the UAE, the Core42 NVIDIA DGX system ranks #37 globally (#1 in the UAE) and the Core42 AMD MI210 system ranks #38 globally (#2 in the UAE). These rankings reflect sustained system balance across compute, networking, and storage at production scale.
The brochure walks through the full architecture stack: GenAI services (agents, RAG, guardrails, fine-tuning, evaluation), model hosting and inference (model catalog, model-as-a-service), AI Ops (training, model customization, model governance), and infrastructure-as-a-service (compute, ultra-fast storage, high-speed networking, managed Kubernetes and Slurm, vector data management, access management, billing, and metering), all underpinned by a unified security and compliance layer.
The platform supports two primary consumption models. On-demand GPU instances give teams immediate access to diverse GPUs via the Core42 AI Cloud console with pay-as-you-go pricing and no long-term commitments, ideal for ML experimentation and inference. Large-scale GPU clusters provide reserved capacity for sustained training workloads, with managed Kubernetes and Slurm orchestration and high-speed networking over InfiniBand and Ethernet.
Built for global scale, Core42 AI Cloud operates 86K+ GPUs across sovereign data centers in the US (Buffalo, Minneapolis, Stockton, Sunnyvale, Dallas), UAE, Southern Europe, with Kenya, India, and SE Asia in active deployment. The brochure closes with the MBZUAI case study: how the Mohamed bin Zayed University of Artificial Intelligence trained frontier models including Jais, K2, and Jais Climate on a sovereign, heterogeneous compute environment combining NVIDIA DGX SuperPod with AMD MI210 GPUs hosted within the UAE, accelerating research timelines while keeping all nationally significant models under UAE jurisdiction.
