Cloud AI: what you rent, what it costs, what it locks in
The cloud providers sell dozens of AI services with overlapping names and unclear boundaries. Underneath there are only four things on offer, and knowing which one you need makes the catalogue navigable.
What this course sets out to do: let you read a cloud AI catalogue and pick correctly, and predict the bill before it arrives.
What this course does not do: walk through consoles. The premium catalogue covers AWS SageMaker, Azure Machine Learning and GCP Vertex AI hands-on.
What you are about to discover
Course contents
| # | Lesson | Main goal | Time |
|---|---|---|---|
| 1 | What you are actually renting | The four categories, and which one solves your problem | 8 min |
| 2 | The three platforms compared | SageMaker, Azure ML, Vertex AI, and hosted model APIs | 9 min |
| 3 | Controlling cost | GPU hours, idle time, inference at volume, egress | 9 min |
| 4 | Security, residency and compliance | Where data lives, who can see it, what to document | 8 min |
| 5 | Lock-in and portability | What binds you, what does not, and what is worth accepting | 7 min |
| 6 | Recap and FAQ | Synthesis, a decision guide, and 12 common questions | 6 min |
| 7 | Quiz and attestation | Validate what you learned with 5 corrected questions | 3 min |
Is this course for you?
- You have to choose a platform and the documentation is not helping you compare.
- You received a cloud bill larger than expected and want to know why.
- You are asked whether data can go to a hosted AI service and need to answer properly.
- You want to know whether managed services or your own infrastructure suits your situation.
Recommended first: MLOps.
What you will be able to do at the end
- Distinguish the four categories of cloud AI offering and pick the right one.
- Compare the three platforms on the criteria that actually differ.
- Explain when a managed API beats training your own model, and when it does not.
- Predict where cost will accumulate and name the levers that reduce it.
- Answer questions about data residency and retention accurately.
- Judge which forms of lock-in are worth accepting.
Estimated time
Around 40 to 50 minutes of reading.
Prerequisites and next steps
Prerequisites: MLOps makes this considerably more useful.
Natural continuation:
- Large Language Models — the hosted model side in depth
- Ethics of AI — what your obligations are regardless of provider
Frequent questions, answered in one line
Is the cloud cheaper than buying hardware?
For variable and experimental work, clearly yes, because you pay only for the hours you use. For steady high-volume inference running continuously, owned or long-term reserved hardware is frequently cheaper, and the crossover point arrives sooner than cloud pricing pages suggest.
Can I train a serious model without a cloud account?
Yes. Classical models train on a laptop, and a single consumer graphics card handles a great deal of fine-tuning. The cloud becomes necessary for large models, large datasets, or when several people need shared infrastructure.
Are the managed AI APIs any good?
For general tasks — transcription, translation, document text extraction, standard image labels — they are better than what most teams would build and require no training data. For anything specific to your domain, they will disappoint, which is exactly the boundary lesson 1 draws.
Which provider should I learn?
The one your employer or client uses. The concepts transfer almost entirely; only the console layouts and service names differ, and those are learnable in days once the concepts are in place.
The premium catalogue covers AWS SageMaker, Azure Machine Learning and GCP Vertex AI hands-on, with a verifiable certificate after a 40-question examination. Included in every paid plan.
Ready? Start with lesson 1 →