Physical AI and VLA Models: The Humanoid Race Between China and the US, and What Russia Has

In July 2026, the BotHub company blog on Habr published the second part of a large breakdown of Physical AI, the "physical intelligence" of robots. Part one explained the technology. Part two covers economics, industrial deployment, government programs, and the race between China and the US for humanoids.

What the second part covers

The second part focuses on how "physical intelligence" works in practice: in the economy, on factory floors, and in national strategies, not just in research labs. The authors move from fundamentals to practice: who is spending money and on what, where robots already work, and which countries are setting the pace.

• The economics of humanoid robots and the cost of deploying them
• Industrial use of VLA models in real production
• Government robotics support programs in China and the US
• Russia's place in the race for "physical intelligence"
• The human factor: how people accept robots

The key question of the whole series is whether "physical intelligence" can repeat in the world of atoms the leap that generative models have already made in the world of bits.

Why did classical robotics hit a ceiling?

Classical robotics hit a ceiling because of rigid programming: a traditional industrial robot executes a pre-written sequence of movements and cannot adapt to unfamiliar surroundings. Change a part, the lighting, or the position of an object, and the script breaks.

The breakthrough the breakdown describes is VLA (Vision-Language-Action) models: an architecture in which the robot gains the ability to generalize from large language and vision models. It does not simply repeat a memorized motion. It "understands" the task in natural language and transfers the skill to new situations.

"Classical robotics hit a ceiling, and language models evolved into VLA," is how the authors of the

BotHub breakdown on Habr describe the key shift in the industry.

The race between China and the US for humanoids

Two countries lead the race for humanoid robots, China and the US, and the breakdown examines their government programs and industrial ecosystems in the most detail. Both are betting on "physical intelligence" as the next big technology wave after generative AI.

The bet on a human-like form has a simple explanation: the world around us, including factories, warehouses, and homes, is built for people, so a humanoid robot can work in the same places humans do without any changes to the infrastructure.

A separate thread in the second part asks what Russia has in all of this: which developments, teams, and groundwork exist, and how realistic it is to join a race in which the leaders have already invested significant resources. The authors take an honest look at the economics, the industry, and human nature, explaining why robot adoption is held back by more than technology.

What this means

Physical AI is moving out of the lab-demo stage and into economics and government strategy. Humanoids and VLA models are becoming a subject of industrial competition between China and the US, and the question for Russia is whether it can claim a place in this race in time. For business, this is a signal: robotics is no longer a niche for engineers but a field where capital, manufacturing capacity, and government support decide the outcome.

Frequently asked questions

What is a VLA model?

VLA (Vision-Language-Action) is a class of models that combine vision, language, and action: the robot perceives its surroundings through cameras, understands the task in natural language, and builds the sequence of movements itself. This approach replaced the rigidly programmed scripts of classical robotics.

What is Physical AI?

Physical AI, or "physical intelligence," is the transfer of large AI models' capabilities from the digital world to the physical one: robots and humanoids that act in a real environment. The Habr breakdown is devoted to exactly this, along with its economics, industry, and government programs.