Learn AI free, straight from the companies building it — all 10 links checked

Every large AI company now runs a free training site. Not a trial, not a lead magnet with a paywall three lessons in — actual courses, written by the people who built the models, given away because a developer who knows your tools is worth more than the course costs to make.

We published these ten as a list. Turning that list into a page meant checking it, and checking it found problems: one link went to a GPU toolkit rather than a course, one returns an error, and three of the ten describe something less free than a list of free things implies. All of that is below, because a link you paste into a browser and find broken is worse than no link at all.

The two links to fix first

NVIDIA — the link goes to a toolkit, not a course

Worth knowing before you go: DLI is not entirely free. Its own self-paced page says “Many of our popular self-paced courses are offered for free. These courses have top ratings, can be completed in a day or less, and are designed for beginners.” Use the Free Courses filter in the course finder; the deeper certification tracks are paid.

• Correct link: nvidia.com/en-us/training
• Filter to Free Courses, or you will be looking at paid certification tracks.
• Strongest for CUDA, inference optimisation and anything close to the hardware.

Meta — the link errors, and the site has moved

ai.meta.com/resources/ returned an HTTP 400 every time we asked for it, as did ai.meta.com itself. Meta blocks automated clients aggressively, so this is not proof the page is gone for a human — but llama.com now redirects to developer.meta.com/ai, which says the AI developer material has been reorganised regardless.

The three that are not as free as the list implies

Free is doing some work in all three of these. None of them is a bait-and-switch; all three give away a genuinely useful amount. But if you plan a study schedule around them you should know where the wall is.

• AWS Skill Builder — free account gives access to hundreds of digital courses. The Individual subscription, at around $29 a month or $299 a year, is what unlocks practice exams, Builder Labs and CloudQuest. The free tier is real; the hands-on labs are the paid part.
• NVIDIA DLI — many self-paced courses are free, in NVIDIA's own words. Not all of them.
• DeepLearning.AI — the short courses are free and excellent. The full specialisations run on Coursera, where you can audit for free but pay for the certificate.

The seven that are exactly what they say

1. Anthropic — Claude Academy

The link on the list, anthropic.skilljar.com, works and is headed “Anthropic Courses”. There is now a newer front door: anthropic.com/learn redirects to academy.claude.com, which describes itself as “Free courses, tutorials, and use cases from Anthropic’s education team.” Either gets you there; the second is the one to bookmark.

Content runs from prompt engineering to MCP — the courses cover connecting Claude to external services through tools, resources and prompts — plus an AI fluency track aimed at educators. Start here if you are building with Claude rather than studying machine learning generally.

2. Google — grow.google/ai

Titled “AI Training to Grow Your Career”, and organised by who you are rather than by technology: AI Essentials, AI for Students, AI for Small Businesses, AI for Educators, plus a Professional Certificate. The least technical list here and the best starting point for someone who is not a developer.

3. Microsoft — Microsoft Learn

Learning paths and modules across Azure AI, Copilot and the rest of the stack, with a student section for anyone wanting a route into a first technology job. The most structured of the ten: modules have estimated times and build on each other.

4. OpenAI — OpenAI Academy

academy.openai.com, with a strong education strand — applying AI to teaching, academic work and career preparation. Best if you want to use the models well rather than build on the API.

5. IBM — SkillsBuild

skillsbuild.org states it plainly: “IBM SkillsBuild offers 100% free learning from a global leader in technology.” It also issues digital credentials, which is unusual in this list and useful if you need something to put on a CV.

6. Hugging Face — huggingface.co/learn

The most hands-on of the ten. Course material on NLP, deep reinforcement learning, diffusion models and agents, all written against open models you can actually run. If you want to understand what is happening inside rather than call an API, start here.

7. DeepLearning.AI

Andrew Ng's. The free short courses are the best hour-long introductions to specific techniques anywhere, and the free resources section includes a career roadmap worth reading before you pick a direction.

An order that actually works

Ten links is a list, not a plan. If you are starting from nothing:

1. Google AI Essentials, for vocabulary and a sense of what the tools do.
2. A DeepLearning.AI short course on whatever you found most interesting there.
3. Hugging Face's NLP course, if you want the mechanics — or Claude Academy and OpenAI Academy if you want to build with the hosted models instead.
4. Microsoft Learn or IBM SkillsBuild once you want a credential rather than knowledge.
5. AWS or NVIDIA last, when you know which infrastructure you are actually going to use.

Sources

All ten URLs were fetched on 14 September 2026 and the destination pages read. NVIDIA's free-course wording is quoted from its self-paced courses page; IBM's from skillsbuild.org; Anthropic's from academy.claude.com. AWS Skill Builder subscription pricing is from AWS's own published training pricing. These pages move — if a link here fails, search the company's own domain rather than trusting a round-up, including this one.