For years Google's learning content had a discovery problem rather than a quality problem. Cloud Skills Boost was solid, Grow with Google was solid, Google for Education had good material, DeepMind published excellent explanations, and finding the right one meant knowing which of four properties happened to host it. Skills, which launched in October 2025, fixes that by putting close to three thousand courses, labs and credentials behind one search box. That sounds like a boring administrative change and it is, but it is also the single most useful thing Google has done for learners in a while, because material you cannot find is material that does not exist.
The hands-on labs remain the strongest part and the clearest reason to pick this over a generic video platform. Being dropped into a real, temporary cloud environment with a task to complete teaches things that watching someone else's screen recording never will, and Google has been doing this longer and better than most. Skill badges attached to those labs are also more meaningful than a completion certificate, because you had to make something work rather than reach the end of a video. The addition of DeepMind content is the genuinely new part and the bit I would flag to anyone deciding whether to bother.
You can move from an applied lab on deploying a model to research-flavoured material on how the models work, inside the same platform, and very few places let you cross that line without switching provider entirely. Now the honest problems. Consolidating four content libraries means inheriting four content libraries, and the quality range is wide enough to be annoying. Serious certification preparation sits next to ten-minute AI Boost Bites that are closer to marketing than teaching, and nothing in the interface tells you which is which before you start.
Expect to abandon a few things. The bias is also unavoidable and worth stating plainly. This is Google teaching you Google, so Vertex AI and Gemini are the answer to almost every question, and the transferable concepts are real but you are learning them inside one company's product decisions. That is fine if your job runs on Google Cloud and much less fine if you are trying to work out which stack to learn in the first place.
The pricing is the part that catches people out. Hundreds of introductory courses genuinely cost nothing, and Cloud customers get the entire library, but the labs are where the learning happens and lab credits are metered. Developers get a monthly allocation that disappears quickly once you are working through anything substantial, and at that point you are on a paid plan or waiting for the calendar to turn over. The gamification I could take or leave.
Streaks and shareable achievements clearly exist to help someone justify a team subscription, and I have never once been motivated by a leaderboard. My recommendation is straightforward. If your work touches Google Cloud, start here, use the free foundational courses to orient yourself and spend your lab credits on the certification paths rather than sampling widely. If you are vendor-neutral, take the free AI fundamentals material, which is good, and pair it with something that is not built by a cloud vendor before you decide what to specialise in.