Humanoid Robots Could Have Their “ChatGPT Moment” by 2027: Is Physical AI Finally Ready?

Artificial intelligence has spent the past few years transforming what computers can do with text, images and software. Now, the industry is preparing for a potentially bigger shift: AI that can understand and act in the physical world.

The latest prediction comes from Wang Xiaogang, chairman of Chinese embodied-AI startup ACE Robotics, who believes humanoid robots could experience a breakthrough similar to ChatGPT by the end of 2027. The idea is that advances in world models, environmental data and AI reasoning could allow robots to perform unfamiliar tasks using natural-language instructions.

The prediction may sound ambitious, but the underlying technology is already developing rapidly.

From Generative AI to Physical AI

The first generation of generative AI primarily operated inside digital environments.

Chatbots could write text.

Image models could generate pictures.

Coding assistants could create software.

Physical AI takes the concept further.

These systems combine AI models with cameras, sensors, motors and robotics hardware. Instead of simply generating an answer, they need to understand their surroundings and determine what action should happen next.

For a humanoid robot, that could mean identifying an object, understanding a spoken instruction, planning a sequence of movements and physically completing the task.

That is significantly harder than generating text.

Why Robots Need a “World Model”

One of the major challenges in robotics is that the real world is unpredictable.

A computer program can operate within clearly defined digital rules.

A robot has to deal with objects moving unexpectedly, changing lighting, uneven surfaces, people walking around it and countless other variables.

World models are intended to help AI systems build an internal understanding of how the physical environment works.

ACE Robotics believes advances in this area could help create a major leap in robot intelligence.

If successful, robots could become much more capable of handling unfamiliar situations rather than simply repeating pre-programmed movements.

China Is Moving Quickly

China has become one of the world's most important markets for humanoid robotics.

At the World Robot Conference in Beijing, more than 300 companies displayed over 2,000 exhibits, with manufacturers demonstrating humanoids sorting parcels, packing products and performing household tasks.

The focus is increasingly shifting from impressive demonstrations to commercial applications.

Factories and warehouses are among the most obvious early markets because environments can be structured and repetitive.

But companies are also exploring hospitality, retail and other industries.

The Data Problem

The biggest obstacle may not be hardware.

It could be data.

Large language models benefited from enormous amounts of digital text.

Robots need real-world physical data.

ACE Robotics currently has around 100,000 hours of real-world training data and wants to increase that to tens of millions of hours within two years.

That is a massive challenge.

Robots need to learn how objects behave, how people move and how physical environments change.

Collecting that data requires actual robots operating in real environments.

Commercial Reality Is Still Far Away

The optimism around physical AI needs to be balanced with reality.

Even if robots achieve a major software breakthrough in 2027, widespread commercial deployment could take several additional years.

Hardware remains expensive.

Battery life remains a limitation.

Robots must also become safer and more reliable before they can operate around people at scale.

The difference between successfully performing one demonstration and reliably performing thousands of tasks every day is enormous.

The Bigger AI Opportunity

Nevertheless, the potential market is huge.

A truly capable humanoid robot could eventually assist with manufacturing, logistics, healthcare, retail and household work.

That would make physical AI one of the most significant applications of artificial intelligence.

The AI industry has already experienced its "ChatGPT moment" in software.

The next one may happen in the physical world.

The question is no longer whether AI can generate.

It is whether AI can perceive, reason and act.

And according to ACE Robotics, that breakthrough may be closer than many people think.