300-640 DCAI
Designing Cisco Data Center AI Infrastructure (300-640 DCAI)
More study resources
Cheat sheet
Keep the key ideas within reach.
ExploreKey terms
Get comfortable with the vocabulary.
ExploreStudy plan
Give each week a clear focus.
ExploreCommon mistakes
Understand the traps and how to avoid them.
ExploreCase studies
Put your knowledge to work in context.
ExploreExam-day guide
Prepare for the day ahead.
ExploreAbout this exam
The Designing Cisco Data Center AI Infrastructure (300-640 DCAI) exam is a Cisco specialist-level certification exam focused on planning and designing data center infrastructure for AI and machine learning workloads. It targets data center architects, pre-sales and design engineers, and senior network or compute engineers who translate AI workload requirements into validated data center designs. The exam is organized around six domains: AI infrastructure architecture and design; compute and GPU fabric for AI workloads; AI/ML storage and data management; high-speed AI networking and fabric design; automation, orchestration, and operations; and security, resiliency, and sustainability. Expect scenario-driven items that ask you to choose or justify a design rather than recall isolated commands. Preparation should combine Cisco data center portfolio study (Nexus switching, UCS and GPU compute, data center management and automation tooling) with a solid grasp of AI fabric concepts: lossless Ethernet, congestion control, RoCEv2 behavior, oversubscription ratios, east-west traffic patterns, and storage throughput for training and inference. Build design fluency by working through end-to-end scenarios, sizing exercises, and failure-domain reasoning. The official Cisco site publishes current exam details, including registration, pricing, and scheduling.
What you'll study
About 300-640 DCAI preparation
What is the exam format?
Cisco delivers its specialist exams as proctored multiple-choice and multiple-response items, and reports that some items may be scenario-based. The authoritative breakdown of item types and delivery options is published on Cisco's certification site, so check it before scheduling rather than relying on secondhand descriptions.
How long is the exam and how many questions are there?
Cisco publishes the current question count and time limit for each exam on its official certification pages, and these figures change over time. Check the exam page when you register so you can plan your pacing. As a general habit, budget time to reread scenario stems carefully.
How hard is it, and how much experience should I have?
It is a design exam, so difficulty comes from judgment rather than trivia. Candidates typically need working knowledge of data center switching, compute, and storage, plus familiarity with AI workload characteristics. Without hands-on exposure to fabric design and operations, scenario items about oversubscription and congestion control will feel abstract.
What should I study first?
Start with AI infrastructure architecture and design, then move to compute and GPU fabric, storage, and high-speed networking. Automation, security, and resiliency build on those foundations. Study the smaller domains last but do not skip them, since they carry enough weight to affect a borderline result.
Do I need to memorize CLI commands?
Less than you might expect. The exam tests design decisions: why a particular fabric topology, congestion-control approach, or storage path fits a workload. Knowing configuration concepts helps you reason about trade-offs, but rote command recall is a weak use of study time compared with scenario practice.
What are the most common failure modes?
Candidates often confuse training and inference requirements, misjudge oversubscription and bandwidth sizing, or treat lossless Ethernet as a checkbox rather than a design constraint. Others overlook storage throughput and failure domains. Read each scenario for the workload described, not the one you support at work.
How should I approach scenario-style questions?
Identify the workload first: model size, parallelism strategy, and whether it is training or inference. Then determine the binding constraint, such as bandwidth, latency, or storage throughput. Eliminate options that violate stated requirements before comparing the remaining designs on operational simplicity and resilience.
Can I retake the exam if I fail?
Cisco publishes its retake and waiting-period policy on its certification site, and the terms depend on the exam and delivery method. Review that policy before scheduling so you know your options. Use the score report's domain breakdown to target the weakest areas before attempting again.
What happens after I pass?
You earn a Cisco specialist certification tied to data center AI infrastructure design. Recertification follows Cisco's published policy for specialist credentials. Confirm the current validity and recertification requirements on Cisco's site, since those details are set by the certification body, not this page.
Is this page affiliated with Cisco?
No. This is independent study material written for candidates preparing for the exam. It is not produced, endorsed, or reviewed by Cisco. Always treat the official exam blueprint and Cisco's published policies as the authoritative source for exam scope, registration, and certification rules.