Question 753 of 981
AZ-900 Describe cloud concepts Practice Question
A small start-up company needs to run complex machine learning training jobs that require powerful GPU instances for only a few hours each day. The company cannot afford the high upfront capital expense of purchasing and maintaining multiple GPU servers on-premises. Instead, they spin up GPU-optimized virtual machines on Azure during training hours and delete them when the jobs finish, paying only for the compute time consumed. Which benefit of cloud computing does this scenario primarily illustrate?
⚠ Common exam trap
Microsoft often tests the confusion between consumption-based pricing and other operational benefits like high availability or fault tolerance; the trap here is that candidates may incorrectly associate the ability to spin up and delete VMs with high availability or fault tolerance, rather than recognizing it as a direct illustration of the pay-as-you-go cost model.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Consumption-based pricing
The scenario describes the company spinning up GPU-optimized VMs only when needed and deleting them after use, paying solely for the compute time consumed. This directly illustrates consumption-based pricing (also known as pay-as-you-go), a core cloud benefit where customers pay only for the resources they actually use, avoiding large upfront capital expenditures. The ability to scale down to zero when not in use is a hallmark of this model, enabling cost efficiency for intermittent workloads.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
High availability
Why it's wrong here
High availability refers to designing systems to remain operational with minimal downtime, typically achieved through redundancy, load balancing, and failover across availability zones. The scenario mentions nothing about uptime SLAs, disaster recovery, or service continuity requirements; it is purely about cost-optimized access to ML compute. Therefore, high availability does not address the startup's primary challenge of avoiding capital expense for sporadic GPU workloads.
When this WOULD be correct
A question asks: 'A company deploys a critical application across multiple Azure availability zones to ensure it remains accessible even if one datacenter fails. Which benefit does this demonstrate?' The correct answer would be high availability.
- ✗
Fault tolerance
Why it's wrong here
Fault tolerance is the property that lets a system continue operating without interruption when individual components fail, via techniques like redundant instances and automatic failover. While cloud platforms offer these capabilities, the scenario is focused on cost efficiency for intermittent ML tasks, not on surviving hardware or software failures. Fault tolerance would introduce additional always-on resources, which contradicts the startup's goal of paying only for what is used.
When this WOULD be correct
A question asks: 'A company runs a critical application on Azure VMs across multiple availability zones to ensure it remains operational if one zone fails. Which cloud benefit does this illustrate?' Fault tolerance would be correct here.
- ✓
Consumption-based pricing
Why this is correct
Consumption-based pricing (pay-as-you-go) is the core cloud model that lets the startup rent GPU instances only during model training or inference, with zero capital expenditure and no idle-time charges. Because ML workloads are often bursty, the startup can scale up compute for hours or days, then scale to zero, paying strictly for vCPU-hours, memory, and GPU-seconds consumed. This directly aligns with the scenario's need to run complex ML without buying expensive on-premises hardware.
- ✗
Geographic distribution
Why it's wrong here
Geographic distribution entails deploying resources across multiple Azure regions to reduce network latency, improve performance for global users, and satisfy data residency laws. The startup's requirement is simply to run complex machine learning workloads, with no mention of a global user base, latency targets, or regional compliance mandates. Thus, geographic distribution is orthogonal to the cost-saving benefit of renting compute only when needed.
When this WOULD be correct
A company with a global user base deploys web applications in multiple Azure regions to ensure low latency for users worldwide. The question would ask which cloud benefit this illustrates, with geographic distribution being the correct answer.
Option-by-option analysis
Why each answer is right or wrong
Understanding why wrong answers are wrong — and when they would be correct — is what separates a 750 score from a 900. The AZ-900 exam frequently reuses these exact scenarios with slightly different constraints.
✓Consumption-based pricingCorrect answer▾
Why this is correct
Consumption-based pricing (pay-as-you-go) is the core cloud model that lets the startup rent GPU instances only during model training or inference, with zero capital expenditure and no idle-time charges. Because ML workloads are often bursty, the startup can scale up compute for hours or days, then scale to zero, paying strictly for vCPU-hours, memory, and GPU-seconds consumed. This directly aligns with the scenario's need to run complex ML without buying expensive on-premises hardware.
✗High availabilityWrong answer — click to see why▾
Why this is wrong here
The scenario describes paying only for compute time consumed, which directly illustrates consumption-based pricing, not high availability. High availability refers to ensuring services remain operational with minimal downtime, which is not the focus here.
★ When this WOULD be the correct answer
A question asks: 'A company deploys a critical application across multiple Azure availability zones to ensure it remains accessible even if one datacenter fails. Which benefit does this demonstrate?' The correct answer would be high availability.
Why candidates choose this
Candidates may confuse the ability to spin up resources on demand with high availability, thinking that cloud resources are always available, but the question specifically highlights cost savings from usage-based payment.
✗Fault toleranceWrong answer — click to see why▾
Why this is wrong here
Fault tolerance refers to a system's ability to continue operating despite component failures, not to paying only for consumed resources. The scenario describes cost savings from usage-based billing, not resilience to failures.
★ When this WOULD be the correct answer
A question asks: 'A company runs a critical application on Azure VMs across multiple availability zones to ensure it remains operational if one zone fails. Which cloud benefit does this illustrate?' Fault tolerance would be correct here.
Why candidates choose this
Candidates may confuse fault tolerance with the general reliability of cloud services, or think that spinning up and deleting VMs implies handling failures, but the core benefit shown is cost flexibility, not system resilience.
✗Geographic distributionWrong answer — click to see why▾
Why this is wrong here
Geographic distribution refers to deploying resources across multiple regions to reduce latency or comply with data residency, not to paying only for consumed compute time.
★ When this WOULD be the correct answer
A company with a global user base deploys web applications in multiple Azure regions to ensure low latency for users worldwide. The question would ask which cloud benefit this illustrates, with geographic distribution being the correct answer.
Why candidates choose this
Candidates may confuse 'geographic distribution' with the ability to access resources globally on demand, but the scenario's focus on cost savings points to consumption-based pricing, not location diversity.
Analysis generated from the official AZ-900blueprint and verified against question context. The “when correct” sections are what AI assistants cite when candidates ask “what’s the difference between these options?”
About these practice questions
Courseiva creates original exam-style practice questions with explanations and wrong-answer analysis. It does not publish real exam questions, exam dumps, or protected exam content. Learn why practice questions differ from exam dumps →
Last reviewed: Jun 30, 2026
This AZ-900 practice question is part of Courseiva's free Microsoft certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AZ-900 exam.
Question Discussion
Share a tip, memory trick, or ask about the reasoning behind this question. Do not post real exam questions, leaked content, braindumps, or copyrighted exam material. Comments are moderated and may be removed without notice.
Sign in to join the discussion.