Humanoid Robots Are Entering the Real World (Faster Than Expected!)

A few days ago, Logan Kilpatrick, the lead product for Google AI Studio and Gemini API, posted something that really stuck with me. He said, "2026 is going to be a huge year for embodied AI. " Then he followed it up with, "Put differently, we're going to see a lot more robots in the real world soon. " That statement matters because this isn't coming from hype culture or sci-fi speculation accounts. It's coming from people who are actually building these systems. And when you zoom out and look at what's happening across the AI ecosystem, it becomes clear that this shift isn't theoretical anymore and it's about to come in 2026. So, let's get into it. Before we get into the video, I

want to show you a clip from China, which is Fibbot playing badminton against a human. What matters here isn't the sport, it's the coordination. The robot is tracking a fastmoving object, predicting its trajectory, adjusting balance, and timing its swing in real time. This is not a presscripted motion. So, this shows that China seems to be ahead in the robotics race compared to the US. What's also crazy is that in China, these humanoids are actually now performing as tourist attractions and they're pretty much everywhere. So, we can see an example of here where this robot is dressed as Santa Claus and is doing some martial arts with a performer on the right. Then, if we scroll down a little bit more, we also see a similar thing at the bottom here. You're seeing more and more tourist attractions in China featuring humanoids as part of their shows. And this kind of shows where the future of embodied AI is. Obviously in the early stages is just for laughs and giggles like we saw with chat GPT when it first was open. People used to ask it basic questions and everything like that. And over time we saw chat GPT evolve into a real agent AI assistant and coding capabilities came to play. And now chat GPT is a very serious tool. And this is probably the same thing that's going to happen with humanoids as well in my opinion. at first are going to be used for things like this martial arts with members over here, but then we'll see it evolve into something much more meaningful. And this is an example in China where humanoids are playing some instruments. So crazy, right? Then we also have this Christmas themed robots dancing around the manufacturing plant dressed up as Santa and snowman. So this person is saying that these robots are cute, but decorations and accessories might also be a business that robotics might enter. We are seeing a lot of change coming in. So, let's talk about 2026 specifically and what embodied AI is. For most of

AI's history, intelligence lived inside software. We typed prompts, models responded, and the interaction stopped there. Embodied AI changes that completely. Embodied AI means intelligence that can see the physical world, reason about it, and take action inside it. It's the combination of vision, language, planning, and motor control in a single system. Instead of answering questions, the AI performs tasks. Instead of generating just text, it moves objects. And instead of living on a screen, it exists in space. That transition is what unlocks robots that are actually useful. It is said that

2026 is the year AI giants go allin on humanoid robots while racing to own the full stack from chips to models to devices. That framing is critical because this is no longer just a model race. It's an infrastructure war. When you look at an AI stack across the major players, a very clear pattern emerges. The companies pushing hardest into robots are the ones that already control multiple layers of the stack. Google stands out immediately. They control custom TPUs, massive training clusters, cloud infrastructure, state-of-the-art Gemini models, consumer devices, and now a direct push into humanoid robotics. Microsoft and Open AI together form another full stack contender. Microsoft brings cloud enterprise distribution and compute. Open AAI brings frontier models, multimodal reasoning, and early steps towards physical embodiment. Amazon already runs the largest warehouse automation operation on Earth. For them, humanoid robots aren't futuristic. They're efficiency upgrades. Meta is approaching this through wearables and spatial computing which still feeds directly into robotics because perception and interaction are core problems for embodied intelligence. Tesla is one of the most vertically integrated player here though they design chips, train models, manufacture hardware, collect real world data and deploy robots inside their own factories. Even Nvidia, which doesn't ship consumer robots, sits at the center of the entire ecosystem by controlling the compute layer. What this chart really shows us is that humanoid robots are not a standalone product. They're the conversion point of chips, models, cloud, software, and devices. And when a single company controls multiple layers, iteration speeds up dramatically. That's why timelines are compressing and the competition is heating up. This is also

why Gemini Robotics matters so much. Gemini Robotics is not just a robot project. It's the physical extension of Google's full stack AI strategy. Instead of training robots to perform one narrow task, Gemini is designed to generalize. It can see an object, understand a spoken instruction, and figure out how to manipulate that object, even if it hasn't seen the exact scenario before. In demonstrations, Gemini powered robots adapt when objects move, recover from mistakes, and continue executing tasks without being reprogrammed. That's important because the real world is messy. Traditional robots fail when conditions change. Generalist models don't. Gemini robotics shows what happens when foundation models stop being purely cognitive systems and start becoming physical agents. Tesla Optimus

represents a different but equally important approach. Optimus is actually not being built for consumers first. It's being deployed internally inside Tesla's factories. That matters because factories are controlled environments with clear tasks and measurable ROIs. Optimus uses the same AI stack as Tesla's self-driving system. Vision, neural networks, and real world learning are already baked into the company. The goal isn't to build a flashy humanoid. It is to replace repetitive labor with softwaredriven machines that can be updated continuously. The footage you're

looking at here is Figure's third generation humanoid robot. And what's important is not just what it's doing, but where it's doing it. This is actually not a factory. This is not a warehouse. This is a homelike environment. That matters because homes are unstructured, unpredictable, and constantly changing. If a robot can operate here, it can operate almost anywhere. Figure three is performing multi-step household tasks by interpreting natural language instructions, visually understanding its surroundings, and physically manipulating everyday objects. This is no fixed script. The robot is reasoning through each step as it goes. This is a huge shift from traditional robotics where every moment has to be predefined and tightly controlled. What figure is demonstrating here is generalization. The robot can adapt when objects are placed differently, when tasks change, and when the environment isn't perfectly predictable. That's exactly what makes humanoid robots viable long term. A home is one of the hardest environments for a robot to operate in. Objects vary in shape, texture, and placement. Lighting changes, and spaces are cluttered. If embodied AI can work here, scaling to warehouses, offices, and commercial settings becomes much easier. Figure isn't saying these robots are ready for mass consumer deployment tomorrow. They're showing that intelligence layer is finally capable of handling real world complexity. And that's the key difference in this wave of robotics. The hardware is important, but the intelligence is what unlocks everything. This also sets the stage for why Gemini robotics is such a big deal because it's attacking the same generalization problem from the foundation model side.

If you enjoyed this video, this is what we do here. fast, clear updates on the biggest moves in AI. If you want to stay ahead of everything happening in this space, make sure you're subscribed. And if you want the hands-on side, demos, tools, workflows, and everything developers can actually build, check out the World of AI. We also run a simple no noise newsletter that gives you the most important AI tools and updates in just a couple of minutes. Subscribe here. Follow World of AI. Join the newsletter.