AI is starting to leave the screen and move into the physical world.
For the past few years, much of the AI conversation has focused on chatbots, co-pilots and tools that generate text, images or code. The next wave is different. It is about machines that can sense their surroundings, make decisions and act in the real world.
That is the promise of physical AI, also known as embodied AI. Think humanoid robots, autonomous vehicles, warehouse machines, drones and intelligent industrial systems that can operate alongside people.
Elon Musk has claimed this technology could help solve poverty and create a world where everyone is happy. That may sound wildly optimistic, and with Musk it often is, but the underlying shift is real. Physical AI could reshape industries, change the nature of work and alter how the economy functions.
To understand where this is heading, we need to look at the companies building these machines now. Some are creating robot workers for factories and warehouses. Others are developing robotaxis, drones, domestic helpers or the software platforms that train intelligent machines at scale.
Here are some of the key players anyone interested in the future of technology, business or society should be watching.
1X Technologies
This company has differentiated itself with its “soft” robots, so called not just because they have textile, fuzzy-looking exteriors but because they focus on safe interactions with humans. It hopes that this will make them popular in homes and as companions. After all, why should robots be hard and metallic? Recently, it announced plans to build 10,000 robots during the first year of operating its new U.S. factory.
Agility
One of the earliest companies to deploy humanoid robots in working environments alongside humans. Its Digit robots have been deployed at factories operated by women’s wear brand Spanx, where they manage warehouse and logistics tasks. The robots are even paid monthly for their work, just like human employees.
Amazon Robotics
Amazon has over a million robot workers, and its global retail and services network has been described as the world’s biggest sandbox for embodied AI. It funds the development of robotics as well as projects integrating them into industrial, commercial, healthcare and domestic settings. Recent progress includes robots capable of unloading trucks at cargo bays and bipedal robots for multi-level environments.
Apptronik
Humanoid robots, including their Apollo model, are built to support humans in industrial and commercial tasks. A notable innovation that it has pioneered is hot-swappable battery systems, allowing its robots to work virtually continuously on factory floors without needing to take charging breaks.
Baidu
Baidu’s Apollo platform is already powering autonomous robotaxi and self-driving car infrastructure in around 20 Chinese cities, including Beijing and Shanghai. With firm backing from the Chinese government and deep integration across the nation’s vehicle manufacturing industry, it has shown itself to be a world-class player in the embodied AI market.
Boston Dynamics
Boston Dynamics, owned by Hyundai, became the poster child for a new generation of AI-enabled robotics thanks to viral videos of robots like Spot and Atlas. Its roots stretch back decades, though, to its time as a developer of technology exclusively for the military. It is known as a pioneer of robots that can navigate human environments, and has built partnerships with AI leaders to help establish its status as a global leader in its field.
DJI
The leading Chinese drone manufacturer is well-known for its consumer and professional drones, but it has also established itself as a leader in embodied AI for industry. Its Matrice 4 platform provides AI processing power for autonomous decision-making across agricultural, construction and public safety use cases.
Figure AI
Figure builds general-purpose humanoid robots designed for labor. It also has a partnership with OpenAI, which means that while its machines work, they’re helping to fine-tune the embodied intelligence capabilities of the next generation of foundation models. Its F.02 robot reportedly helped out in the construction of 30,000 cars at BMW’s manufacturing plants, while its latest F.03 model is designed for domestic chores.
Nvidia
Nvidia CEO Hensen Huang has said that the ChatGPT moment for robotics is just around the corner. Creating the software “backbone” for autonomous, embodied AI is central to his company’s strategy. This means exploiting the power of its AI chips to create powerful simulations and platforms for training and controlling physical AI on an industrial scale.
Tesla
Tesla’s best-known embodied AI products are its self-driving cars and robotaxis. The data and experience of mass production that has been built up is now being turned to the development of its humanoid robot Optimus. Musk’s grand plan is for a helper robot in every home and hordes of robotic workers to relieve us of the grind of day-to-day labor. The long-term goal is to produce a million humanoids a year, making it feasible to sell them at around the same cost as an affordable family car.
Unitree Robotics
This Chinese manufacturer has raced ahead of the pack when it comes to deploying robots to do real work. It was the world’s top seller of humanoid robots in 2025, shipping over 5,000 units, as well as 30,000 quadruped robot dogs. Pursuing a strategy of reduced cost through mass production has given it a lead, but what’s interesting is the steady rise in the number of its units going into commercial and industrial use versus those going into education and research.
Waymo
Alphabet’s (Google) self-driving vehicle spin-off has established itself as the leading operator of robotaxis (driverless passenger cars) outside of China, with its network reportedly completing half a million journeys each week, across more than 10 U.S. cities. Its scale means that Waymo is one of the clearest examples of an embodied AI company delivering commercial services into everyday use at scale.
The Path To AGI
Physical AI is often discussed in connection with artificial general intelligence, or AGI. The basic idea is that machines may need to understand the physical world before they can truly generalize knowledge across many tasks.
Humans learn by interacting with the world. We touch things, move through spaces, observe cause and effect, then apply those lessons in new situations. If machines are ever going to reason and act with broad flexibility, they may need something similar.
That is one reason embodied AI has become such an important field. A chatbot can process language, but a robot has to understand gravity, friction, movement, distance and risk. It has to connect intelligence with action.
Many technical challenges remain, but the bigger questions may increasingly be social and economic. How comfortable are we with autonomous machines in our homes, workplaces and streets? How will they change jobs? Who is responsible when they make mistakes? How do we make sure they are safe?
Physical AI is still in its early stages, but the direction is clear. Intelligence is moving into the world around us.
The companies building that future today could become some of the most important businesses of the next decade.
