Microsoft certifications are among the few credentials Indian recruiters verify by ID — which makes the AI certification roadmap Microsoft publishes unusually valuable for students. But the map changed in 2026: the famous AI-900 retired on June 30 and AI-901 took its place. Here is the current, correct path from zero to certified, with costs, free prep and voucher tricks.
AI-900 retired on June 30, 2026 — any guide still telling you to take it is outdated. Its replacement, AI-901, covers the same fundamentals with a generative-AI refresh: AI workloads, ML principles on Azure, computer vision, NLP and generative AI workloads. No coding or data-science background required, though basic cloud concepts and Python familiarity help.
Preparation is 100% free on Microsoft Learn, including sandbox labs that give you temporary Azure access without a credit card. Take the free practice assessment first — if you score above 80%, schedule the exam; if not, the assessment tells you exactly which of the weighted domains (generative AI now carries 20–25%) needs work.
Exam practicalities: passing score 700, price varies by country (confirm with the exam provider before booking), register with a personal Microsoft account — school accounts lose records when you graduate. One focused month at 6–8 hours a week is enough alongside college.
AI-102 is where fundamentals become employable skill: designing and implementing Azure AI solutions — cognitive services, language understanding, knowledge mining, document intelligence and agentic workflows on Azure AI Foundry. This is the certification hiring managers name in AI engineer postings.
Prerequisites are practical, not formal: comfort with Python and REST APIs, plus the conceptual base from AI-901. Plan two to three months, and pair every module with a build — a notes-summarizer on Azure for Students, a syllabus search bot, a document-intelligence pipeline for your department. Exam plus project beats exam alone in every interview.
Microsoft's own Q&A guidance for students says the same thing: fundamentals first, then hands-on, then AI-102 as the resume credential at graduation. Follow that order and each step compounds instead of expiring.
From AI-102, branch by target role. Aspiring data scientists take DP-100 (Azure Data Scientist Associate). Builders who want proof without another full exam earn Applied Skills credentials — scenario-based lab assessments (Copilot Studio, AI agents, automation) that are Microsoft-verified and shareable on LinkedIn.
Career changers and entry-level applicants should also know the Essentials Professional Certificates (e.g. Azure AI Essentials): complete a LinkedIn Learning path plus a free summative assessment. Lighter than an exam cert, but legitimate third-party validation for your first internship applications.
Rule: one fundamentals plus one role-based cert plus one project portfolio is the complete student package. Beyond that, each additional exam has diminishing returns until you have work experience to attach it to.
Learning costs nothing — Microsoft Learn paths, sandboxes and practice assessments are free. Exams are paid (country-based pricing, taxes extra), which is exactly why timing matters: Microsoft Build and Ignite Cloud Skills Challenges regularly grant free exam vouchers for completing learning playlists, and AI sweepstakes offer 50%-off and free-exam draws for Applied Skills earners.
Azure for Students gives you free credits and services for building the projects each exam demands — claim it in semester 4 so your project runway is ready when certification season starts. Watch the official challenge pages, not resellers: vouchers come only from Microsoft event flows.
Budget plan for a student year: ₹0 for learning, one paid exam (AI-901) as the anchor investment, vouchers covering AI-102 via a challenge. Total cash outlay can stay under a single exam fee with disciplined challenge participation.
Semester 4: AI-901 fundamentals — one month of Learn modules plus practice assessment, exam in month two. Parallel: Azure for Students signup and one tiny build (notes summarizer) to make concepts stick.
Semester 5: AI-102 track — two to three months of modules with one project per service area, exam before semester-end. Publish each project on GitHub with architecture notes; interviewers ask about these, not your score.
Semester 6: branch — DP-100 or two Applied Skills credentials plus placement applications that cite cert ID numbers. Add credentials to LinkedIn the day you earn them with the verification link; recruiters filter on exactly that. Combined with a TaskVeda accelerator project, this stack reads as certified plus proven — the combination that shortlists.
Mistake one: preparing for retired AI-900 from old YouTube playlists — check the exam code on learn.microsoft.com before a single hour of study. Mistake two: treating courses as credentials — Copilot course completions are not certifications; only passed exams and assessments count with employers.
Mistake three: registering with a school account and losing records at graduation — always use a personal account. Mistake four: collecting exams without projects — a certified student who cannot demo a build loses to an uncertified student who can.
Mistake five: paying full price without checking challenges first — Build and Ignite voucher windows open yearly, and planning your exam calendar around them is the highest-ROI scheduling a student can do.
Booking: schedule through the official exam provider with a personal Microsoft account, and confirm the India price plus taxes at checkout before paying — pricing is country-based and changes without notice. Choose a morning slot; fundamentals exams reward fresh recall over cramming, and 700 out of 1000 is the passing bar.
Preparation week: stop new modules three days out and switch to the free practice assessment on repeat plus your weakest domain (most students: generative AI workloads at 20–25% weight). Skim the study guide linked on the exam page the night before — Microsoft updates exam objectives periodically and the guide is the source of truth.
If you fail: you get a score report by domain, a mandatory waiting period, then a retake. Treat the report as a syllabus — most fails come from one weak domain, not overall readiness. Students who retake within three weeks focusing only on flagged domains pass at very high rates.
AI-901 (Azure AI Fundamentals) — it replaced the retired AI-900 in mid-2026, needs no coding background, and its free Learn path plus sandbox labs make it a one-month project alongside college.
AI-900 retired on June 30, 2026. Existing certifications remain valid, but new candidates must take AI-901 instead. Verify the current exam code on learn.microsoft.com before preparing.
Pricing is country-based and confirmed at booking (taxes extra). Learning, sandboxes and practice assessments are free, and Build/Ignite challenges plus sweepstakes regularly offer free or 50%-off vouchers.
AI-901 proves fundamentals; AI-102 (Azure AI Engineer Associate) is the employable credential hiring managers list. Take AI-901 first, then AI-102 with one project per service area.
Often yes via event vouchers: complete Build or Ignite Cloud Skills Challenge playlists by the deadline for a free exam, or earn Applied Skills credentials during sweepstakes windows for discount draws.
Yes when paired with projects — recruiters verify cert IDs, and AI-102 plus a deployed Azure project is a strong fresher signal. Certificates alone, without anything built, carry little weight.
Microsoft uses country-based pricing confirmed at booking, plus applicable taxes. Learning paths, sandbox labs and practice assessments are completely free — budget only the exam fee itself, and check Build/Ignite challenge windows for voucher coverage first.
Yes — online proctored delivery is available with ID verification, a clear room scan and a stable connection. College hostel rooms usually pass the environment check; confirm system requirements a day early to avoid rescheduling.