Domain 2 — AI Infrastructure (40%)

The heaviest domain — covers everything you need to spec, build, connect, and store for an AI data center. Rich in official NVIDIA course screenshots.

Subdomains

# Topic File
2.1 Hardware requirements for specific AI training use cases hardware-requirements
2.2 Scaling GPU infrastructure gpu-scaling
2.3 Power and cooling requirements power-and-cooling
2.4 On-prem vs cloud onprem-vs-cloud
2.5 Cluster components of accelerated infrastructure cluster-components
2.6 Facility requirements facility-requirements
2.7 Networking requirements for AI workloads networking-requirements
2.8 Data center networking protocols and key concepts networking-protocols
2.9 High-speed data center network options and use cases high-speed-network-options
2.10 DPU purpose and benefits dpu

Study tip: 2.7–2.10 have the most screenshots and the most specific product names. Know the difference between AI Fabric (E-W) and control/user-access network (N-S). Know which hardware (BlueField-3 + Spectrum-4) enables RoCE adaptive routing and congestion control.

Table of contents

Back to top

Licensed under CC BY 4.0. Notes based on NVIDIA course materials and original field experience. Not affiliated with or endorsed by NVIDIA Corporation. No exam-dump material — all practice questions are original.