Demand comes in short bursts a few times per year. What is the main appeal of the cloud?
Why B is the answer
You pay for capacity only when you use it
Elastic capacity suits bursts, while steady use often favors owned systems.
In detail
- Elastic capacity suits short bursts.
- Steady use may favor owned systems.
Use case
An illustrative scenario, not a real incident.
A retailer trains models a few times a year, and rents GPUs only for those weeks.
See it in different ways
flowchart TD
A{"Is demand steady and high?"} -->|Yes| B["On-premises can cost less over time"]
A -->|"No, it comes in bursts"| C["Cloud gives flexibility"]
D{"Are there data residency limits?"} -->|Yes| BstateDiagram-v2 [*] --> Demand Demand --> OnPremises: steady and high Demand --> Cloud: short bursts OnPremises --> DataRules Cloud --> DataRules DataRules --> [*]
flowchart TD
Q{"How often are GPUs busy?"}
Q -->|"A few times a year"| A0["Cloud"]
Q -->|"All year"| A1["Compare with owning"]Key point: Cloud suits bursts.
Why the other options are wrong
- A. It guarantees the lowest cost for steady workloads
- Steady use often favors owned systems.
- C. It removes the need to think about data security
- Security stays your concern too.
- D. It removes all model training time
- Training still takes time.