For years, humanoid robots existed only in sci-fi movies or in spectacular demonstrations on the stages of conferences. In 2026, they took the next step: they left the lab and entered real factories, where they work alongside humans. At Altanet Craiova we consider this one of the most spectacular changes that AI is bringing to the physical world – and it is only just beginning.
What is physical AI?
Until now, AI lived exclusively in the digital world: it answered questions, generated text, analysed data. Physical AI is the next step – artificial intelligence systems that perceive the real world through cameras and sensors, make decisions and act through mechanical arms, wheels or legs.
The company Deloitte describes this shift simply: "AI is becoming physical." We are no longer talking about a program on a server, but about a system that lifts a box, screws in a part or pushes a cart in a warehouse.
Atlas's first real job
In January 2026, the Atlas robot made by Boston Dynamics began its first field test in a real factory – the Hyundai plant in Savannah, Georgia. This moment was broadcast by CBS on the "60 Minutes" program and drew the attention of the whole world.
Why does it matter? Because Atlas is not a fixed mechanical arm mounted on an assembly line. It is a robot that walks on two legs, climbs stairs, lifts objects of different shapes and adapts to unstructured spaces – exactly as a human worker does.
Which other companies are in the race?
Boston Dynamics is not alone. In 2026, several major companies presented or tested humanoid robots in industrial environments:
- NVIDIA Alpamayo: a 10-billion-parameter AI model created specifically for robotics and autonomous driving, able to understand the physical environment and make decisions in real time.
- Figure AI, Agility Robotics, Apptronik: American startups that are already delivering humanoid robots to Amazon warehouses, BMW factories and other large companies.
- Unitree, Fourier Intelligence: Chinese manufacturers offering humanoid robots at significantly lower prices than their Western competitors.
At CES 2026, the technology exhibition in Las Vegas, humanoid robot demonstrations dominated the press's attention. They are no longer curiosities – they are commercial products.
Where they are used now and where they will be soon
The chart below shows the main application areas of AI robotics in 2026 and the deployment estimates:
8%
Mature commercial deployment
Growing / pilot
Why now and not earlier?
Classic industrial robots have existed for decades. What changed in 2026 is the convergence of three technologies that matured at the same time:
- Perception: cameras and sensors are now good and cheap enough for a robot to "see" and understand the space around it in real time.
- Simulation: before being sent to a real factory, a robot can be trained for millions of hours in a virtual environment. NVIDIA has created platforms dedicated to this type of training.
- AI models: the brains that coordinate everything are now capable enough to make good decisions in unexpected situations, not just in preset scenarios.
What comes next?
IBM Research is direct in its estimates: "Robotics and physical AI will definitively accelerate." Estimates show that by 2028 AI robots will be standard equipment in logistics and in factories. Household applications are still limited, but in the industrial environment the ROI (that is, the money recovered from the investment) is already measurable.
For Romania, this trend is especially relevant in the automotive industry and in logistics, two sectors in which our country has a significant presence. Companies that start getting familiar with these technologies now will have a clear advantage over those that have not taken the step.
Do you want to know how AI robotics could affect your sector? The Altanet Craiova team follows these developments and can help you understand what they mean in concrete terms for your business. Visit our contact page and let's talk.
This article is part of the Altanet series on AI trends in 2026. Next article: AI in Healthcare: How an algorithm detects disease earlier than the doctor. See also the complete guide to the series.