Twenty Free AI Courses, Four Branches: Pick the One Closest to Your Work
You do not need an expensive degree to learn AI. The part nobody mentions is that the companies actually building it already teach it themselves, for free: Anthropic, OpenAI, Google, Microsoft, NVIDIA, Meta, IBM, AWS – alongside Stanford, MIT and Harvard. No tuition, no trial, no card. You sign in and start.
Twenty of them in one place is genuinely useful. It is also twenty open tabs and a strong feeling of being behind, which is why the instruction printed at the bottom of the card matters more than the list: start with one today, not twenty, and pick the branch closest to the work you already do.
Four branches, not twenty courses
The twenty come from four kinds of organisation, and what they teach follows from what they sell.
| Model labs → | Clouds → | Universities → | Data platforms |
|---|---|---|---|
| How to use the thing well Prompting, working habits, what the model can and cannot do |
How to build and deploy Their platform, their services, production patterns |
Why it works The maths, the methods, the history behind the words |
Where the data lives Storage, retrieval, memory for AI applications |
| Who it suits: Anyone who works with AI daily |
Who it suits: People shipping software |
Who it suits: People who want the foundations |
Who it suits: People building AI apps |
| The bias: Teaches their product first |
The bias: Teaches their platform first |
The bias: Vendor-neutral |
The bias: Teaches their database first |
None of that bias is a problem as long as you know it is there. A lab’s course on using its own assistant is the best possible guide to that assistant. What it will not tell you is when a different tool is better, which is a question only you can answer – the argument in generalist versus expert tools.
All twenty, by branch
Here is the full list with what each one teaches and an honest note on whether it earns your time if you run a channel.
| Branch | Course | What it covers | Useful for a creator? |
|---|---|---|---|
| Model labs | Anthropic Learn | Practical Claude use, prompting and safe AI at work | Yes – start here |
| Model labs | OpenAI Academy | AI basics, prompt craft and real working workflows | Yes |
| Model labs | Google AI | Beginner-friendly AI skills for everyday work | Yes |
| Model labs | Meta AI Resources | Open research, tools and learning material | Only if you are curious about open models |
| Model labs | DeepLearning.AI | Generative AI, LLMs and core machine learning | Yes, after the basics |
| Model labs | Hugging Face Learn | Transformers and open models, hands-on | No, unless you build |
| Clouds | Google Cloud AI | Generative AI, cloud tools, enterprise use cases | No |
| Clouds | Microsoft AI Skills | AI fundamentals, Copilot and responsible AI | Partly – the fundamentals |
| Clouds | AWS Skill Builder | AI and ML on cloud infrastructure | No |
| Clouds | IBM SkillsBuild | Machine learning, deep learning and business AI | No |
| Clouds | NVIDIA DLI | Accelerated computing, deep learning, deployment | No |
| Universities | Stanford Online | Academic and practical foundations of AI | If you want depth |
| Universities | MIT OpenCourseWare | Full lectures on AI theory and methods | If you want depth |
| Universities | Harvard CS50 AI | Search, learning and optimisation in AI | Good first university course |
| Universities | fast.ai | Hands-on deep learning with fast results | If you code |
| Universities | Kaggle Learn | Micro-courses: Python, data skills, ML | Good for short sessions |
| Data platforms | MongoDB University | AI apps and modern developer workflows | No |
| Data platforms | Databricks Academy | Data engineering, ML and AI platforms | No |
| Data platforms | Snowflake University | Data analytics and AI-powered workflows | No |
| Data platforms | Redis University | Vector search and memory for AI apps | No |
Count the yeses and the twenty becomes four or five. That is not the list failing – it is the list doing its job, which is to let you find your branch and ignore the rest, exactly as in eighty tools sorted by the job.
The post this came from
Watch the reel
All twenty official learning hubs on one card. Posted on the BANI Academy page.
Pick your branch by the work you already do
The card says to choose the branch closest to your existing work. Here is what that looks like in practice.
| If your work is… | Your branch | Start with | What you should have afterwards |
|---|---|---|---|
| You write, film or design every week | Model labs | The course from the assistant you already use | A prompt library of five shortcuts you keep |
| You manage people or projects | Model labs, then a cloud fundamentals course | Prompting, then responsible use | A one-page policy for your team |
| You build small tools or scripts | Universities, hands-on branch | A practical coding course | One script that does a real job |
| You handle spreadsheets and reports | Kaggle-style micro-courses | Python and data basics | A cleaned-up version of your own monthly sheet |
| You want to understand the field | Universities | One full lecture series, start to finish | Being able to explain a model’s limits in two minutes |
The right-hand column is the part worth insisting on. A course you finished with nothing to show for it is indistinguishable, a month later, from a course you never opened. Decide the artefact before you start – a prompt library, a working script, a cleaned-up sheet – and the material stops being entertainment.
For a channel, that artefact is usually a small set of saved instructions you actually trigger each week. Turning what you learn into shortcuts is the fastest way to keep it, as in save once, call in one word, and the structure that makes each one reliable is in vague in, vague out.
Why free courses go unfinished, and what to do about it
Access has not been the problem for years. Completion is. A free course asks nothing of you – no deadline, no money down, nobody noticing you stopped – and anything that asks nothing tends to get nothing back.
Three things fix most of it. Put the sessions in the calendar, as three fixed hours rather than “when I have time”. Tell one person what you are doing and when you will be done. And define the artefact first, so finishing means producing something rather than reaching the last video.
That is the same gap we wrote about in nine free Amazon AI courses – excellent material, no accountability – and the same trap as consuming endlessly without shipping, described in self-taught and still guessing. If you would rather learn by watching, the channels worth opening in order are in same ten channels, different first click.
Frequently asked questions
Are all twenty really free?
The learning hubs listed are free to access. Some hands-on labs use cloud services that can cost money, and certificates are sometimes paid, so check the page you land on.
Do I need a certificate?
It proves attendance, not skill. What convinces anyone is the work: a prompt library you use, a script that runs, a video that got better.
Which one should a complete beginner open?
A model lab course on the assistant you already use. It pays off the same week, which is what keeps people going.
Is a vendor course biased?
It teaches its own product first. That is fine when you want to use that product, and a reason to add a university course when you want the fundamentals.
Open one today, not twenty
Pick your branch from the table. Open one course. Write down the artefact you will have at the end and the three hours you will spend on it this week. Close the other nineteen tabs; they will still be there, and you will know exactly where to find them.
The courses cost nothing. The only currency here is finishing one – and one finished course beats twenty saved links every time.
Learning a lot does not make anyone good. Doing a lot does.
If you want the structure that turns that into a habit – the deadline, the accountability, the people doing it with you – that is what our free challenge is built for: no barrier to entry, a genuine commitment to action, and a refundable commitment fee you get back when you finish. You are not paying for the knowledge. You are betting on yourself finishing, and we hold the stake.
Which number are you starting with? Tell us in the comments on the original reel and we will tell you what to do first.
For more on building channels for English-speaking audiences, that is what we work on at mmoyoutube.com.
