AI Wisdom

NVIDIA associate level

NVIDIA Certified Associate: AI Infrastructure and Operations

Code
NCA-AIIO
Length
60 minutes
Price
$125
Valid for
2 years, renewed by retaking the exam
Study time, our estimate
25 to 40 hours

The best on-ramp for infrastructure engineers moving into AI. If you know data centers but not GPU clusters, this is a few weeks of work that gets you the vocabulary. It is also a sound first step before the professional infrastructure exams.

Take it if

You run infrastructure and AI workloads are coming your way.

Skip it if

You already operate GPU clusters. Go to NCP-AII or NCP-AIO.

Before you start

Check yourself against this list first. If something is missing, close that gap before you pay for an exam.

  • A basic understanding of data center infrastructureNVIDIA's stated prerequisite.

Next: NCP-AIINCP-AIO

All exam details
Format
50 questions
Length
60 minutes
Price
$125
Valid for
2 years, renewed by retaking the exam
Delivery
Proctored, English
Experience
Basic understanding of data center infrastructure
If you fail
Wait 14 days

What the exam covers

  • AI Infrastructure40%
  • Essential AI Knowledge38%
  • AI Operations22%

What to focus on

  • Why GPUs suit AI, and how training differs from inference.
  • The NVIDIA stack: DGX, networking, and software.
  • InfiniBand against Ethernet for AI clusters.
  • Power, cooling and data center needs.
  • Running workloads: Kubernetes, Slurm, MIG and monitoring.
Everything on one mapEvery topic and key idea in this certification.
mindmap
  root((NCA-AIIO))
    AI infrastructure
      NVLink
      InfiniBand
      MIG
      Facilities
    Essential AI knowledge
      Why GPUs
      Training
      Inference
    AI operations
      DCGM
      Slurm
      Kubernetes

Lesson 1 of 3, 40% of the exam

AI infrastructure

The largest domain at 40%.

NVLink
High-bandwidth GPU-to-GPU links.
InfiniBand
Low latency with RDMA for training clusters.
MIG
Splits a GPU into isolated instances.
Facilities
Power and cooling often set the limit.
Inside an AI clusterHow the pieces connect.
flowchart TD
  A["GPU"] -->|"NVLink"| B["GPU"]
  A --> C["Server"]
  C -->|"InfiniBand or Ethernet"| D["Network fabric"]
  D --> E["Other servers"]
  D --> F["Storage"]

Common trap

Sizing compute and forgetting power and cooling.

Try it yourself

Sketch a small cluster with compute, network and storage.

Official docs: MIG User GuideDGX platform

Lesson 2 of 3, 38% of the exam

Essential AI knowledge

You need to understand the workloads you run.

Why GPUs
Training is parallel matrix math.
Training
Long jobs that learn weights.
Inference
Serves requests with latency targets.
Training and inferenceSame model, very different workloads.
flowchart LR
  subgraph T["Training"]
    T1["Huge dataset"] --> T2["Many GPUs for hours"] --> T3["Weights"]
  end
  subgraph I["Inference"]
    I1["One request"] --> I2["Weights"] --> I3["Fast answer"]
  end

Common trap

Planning inference capacity like training capacity.

Try it yourself

List how training and inference differ on your infrastructure.

Official docs: DGX platform

Lesson 3 of 3, 22% of the exam

AI operations

Covers daily running.

DCGM
Monitors GPU health at cluster scale.
Slurm
Common batch scheduler for training.
Kubernetes
GPU Operator automates the software.
Two ways to schedule GPUsOften decided by the workload.
flowchart TD
  A{"Type of work"} -->|"Long batch training"| B["Slurm"]
  A -->|"Services and inference"| C["Kubernetes with GPU Operator"]
  B --> D["DCGM monitors the GPUs"]
  C --> D

Common trap

Assuming Kubernetes knowledge covers GPUs.

Try it yourself

Read the GPU Operator overview and list what it installs.

Official docs: DCGM DiagnosticsNVIDIA GPU Operator

Study plan

  1. Weeks 1 and 2: the course

    Take the self-paced AI Infrastructure and Operations Fundamentals, which takes about seven hours.

  2. Week 3: the study guide

    Work through NVIDIA's exam study guide and the practice questions in learning mode.

  3. Exam week

    Take the mock below. One hour for 50 questions leaves about a minute per question.

Exam day and after

  • NVIDIA certifications last two years, and you renew by retaking the exam.
  • Exams are proctored and in English. NVIDIA's FAQ covers both remote and in-person testing, and remote exams allow no breaks.
  • Check the exam page before you plan around a date. NVIDIA has been adding hands-on labs to professional exams.
Test yourself8 practice questions, scored by exam topic.
Start the mock exam

Keep going

Exam facts checked on October 31, 2026 against NVIDIA certification page. Lessons are our own explanations, so use the official docs as the final word, and confirm price and dates before booking.