The U.S.-China AI competition is shifting from software and semiconductors to a new frontier: who can build AI systems that operate in the physical world. While America leads in large language models (LLMs), which are AI systems trained on vast amounts of text to understand and generate language, China has constructed a structural advantage in the infrastructure needed to train the next generation of AI that learns by doing, not just by reasoning about data. This divergence could reshape which nation emerges dominant as artificial intelligence moves beyond chatbots and into factories, warehouses, and autonomous systems.

What Is the Physical AI Race, and Why Does It Matter?

The AI field is developing along two distinct paths. The first extends the current wave of large language models, where progress depends on access to massive computing power and training data. The second involves "world models" or "embodied AI," systems that learn by perceiving and acting in real environments, adjusting their behavior based on sensory feedback from cameras, touch sensors, and other inputs.

The distinction matters because language models excel at reasoning about the world on paper, but they cannot learn from physical experience. A language model can describe how to assemble a machine or optimize a workflow in theory, but it cannot feel friction, adapt to unexpected obstacles, or improve through trial and error in a real factory setting. World models, by contrast, accumulate knowledge through deployment, with every movement, adjustment, and correction feeding back into learning.

When German Chancellor Friedrich Merz visited China in February, President Xi Jinping demonstrated this advantage firsthand, showcasing humanoid robots dancing, sparring, and moving in perfect synchrony. The display reflected more than technological prowess; it illustrated what a coordinated industrial and political system can build and deploy at scale.

Where Does Each Country Lead in the AI Competition?

The United States maintains a commanding position in the first path. According to Stanford's 2026 AI Index, U.S.-based institutions produced 50 notable AI models in 2025 compared to China's 30, narrowing from a 40-to-15 gap just one year earlier. In private investment, the asymmetry is even starker: U.S. private AI funding reached $285.9 billion in 2025, roughly 23 times China's $12.4 billion.

For the inputs that large language models demand, the United States holds a decisive edge. NVIDIA controlled an estimated 80 to 90 percent of the AI graphics processing unit (GPU) market in 2025, a dominance built on hardware and software ecosystems that have made its chips the default for frontier model training across the industry. The U.S. position is strengthening further: the OpenAI-Oracle-SoftBank Stargate venture is building toward 10 gigawatts of data center capacity, and Taiwan Semiconductor Manufacturing Company (TSMC) is expanding its $165 billion investment in Arizona to anchor the computing infrastructure on American soil.

China, however, has built a structural advantage in the substrate needed for world models: streams of sensory data generated by physical systems working in real environments. According to the International Federation of Robotics, China installed a record 295,000 industrial robots in 2024, representing 54 percent of global deployments and nearly nine times as many as the United States. The gap in the installed base is even starker: roughly 2,027,200 factory robots operate in China, approximately five times the 393,700 in the United States.

Chinese firms such as Unitree are commercializing humanoid platforms at prices that no Western equivalent can match, accelerating the accumulation of real-world training data for embodied AI systems.

What Are the Key Factors Determining AI Dominance?

• Computing Infrastructure: The United States controls GPU production and data center capacity, essential for training large language models, while China dominates industrial robot deployment, critical for gathering sensory data needed to train world models.
• Critical Minerals and Energy: China is the dominant refiner of 19 of the 20 minerals in the International Energy Agency's Global Critical Minerals Outlook 2025, with an average market share around 70 percent. In permanent magnets, China's share of sintered magnet production has climbed from roughly 50 percent two decades ago to 94 percent today, giving it near-total control over components essential to robot actuators and data center cooling systems.
• Hybrid System Integration: The frontier is moving toward systems that combine language models with perception, planning, and control. Whichever nation can integrate both paths and translate them into deployed capability at scale will gain a structural industrial and military advantage.

How Are Capital Markets Signaling the Shift to Physical AI?

Major venture capital investors are already moving resources toward world models and embodied AI. Yann LeCun, formerly Meta's chief AI scientist, raised $1.03 billion at launch for his new venture AMI, describing its mission as building world-model-based AI "that understands the real world," with applications in industry, robotics, healthcare, and automation. World Labs announced a separate $1 billion round in February to advance spatial intelligence and build world models for robotics and scientific discovery.

These are substantial commitments to an emerging technical direction, signaling that world models and embodied AI are now treated as strategically consequential by the investment community.

What Constraints Remain on China's Physical AI Advantage?

China's dominance in robotics deployment does not mean the terrain is uncontested. Software for training, sensing, and control remains dominated by firms based outside China, most importantly NVIDIA, whose ecosystem underpins nearly every major robotics company in the country. High-precision mechanical components represent another dependency: Bosch Rexroth, Schaeffler, THK, and NSK control roughly 90 percent of the high-end ball screw and precision motion market on which Chinese manufacturers rely.

At the level of autonomous function, the gap is wider still. Chinese humanoids are mostly deployed in narrow, site-specific tasks and trials, a far cry from full autonomy in unpredictable environments.

How Does This Shift Affect Diplomatic Negotiations?

As President Trump hosts Chinese President Xi Jinping at a state dinner this week, the topic of AI is expected to rank high on the agenda. The two nations have already begun exploring limited cooperation: Treasury Secretary Scott Bessent said that the U.S. has proposed a new "notification mechanism" for AI incidents that could affect national security.

"We think that just like any cross-border activity, that moving from opaque to more transparency between the No. 1 and the No. 2 AI powers in the world is very important," Bessent stated.

Scott Bessent, U.S. Treasury Secretary

However, mutual trust remains fragile. The U.S. is already restricting exports of the most powerful AI chips to China and has accused Chinese developers of tapping into American AI models at industrial scale to extract their capabilities, a practice known as distillation. Beijing argues that distillation is widely used, including by American companies, and has criticized the U.S. for attempting to monopolize the industry.

The Trump administration has also launched a campaign to isolate Iran from its remaining economic partners, but Beijing has insisted on its right to trade with Tehran. China is Iran's biggest oil buyer and largest trading partner, and has been building ways to make payments outside the U.S.-led financial system, reducing Washington's leverage.

For years, trade restrictions on high-performance AI chips and rare earth elements have been recurring points of contention. Whether mutual trust is strong enough to lead to reciprocal concessions on AI transparency will likely only become clear at the summit.

What Is Trump's New AI Policy Direction?

In a shift from his earlier dismissal of AI safety concerns, President Trump announced over the weekend that he would establish an "AI Force" and appoint a special envoy to address the issue. Trump promised that he would by no means slow down the industry's growth, but stated that "we will also be looking for BAD, and we can do that, very easily, with our already existing Criminal and Civil Justice System".

"To that end, I will be announcing, in the near future, the AI 'Czar', Only High I.Q. individuals need apply!" Trump declared.

Donald Trump, U.S. President

Trump left open exactly what the individual's duties and powers would be and what resources would be allocated to the position. The announcement could be an attempt to undermine an initiative by California's Democratic Governor Gavin Newsom, who announced plans to establish a commission to regulate Silicon Valley-based tech companies and require emergency shut-off switches in AI systems.

The current "AI race" between the U.S. and China is often compared to the space race between the U.S. and the former Soviet Union in the 1960s. To ensure American companies remain industry leaders, Trump intends to continue giving them free rein while maintaining export controls on advanced chips and critical minerals.

As the competition evolves beyond language models into embodied systems that operate in the physical world, the question of who controls the respective substrates,computing infrastructure for the U.S., robotics deployment and critical minerals for China,will increasingly determine which nation achieves strategic primacy in artificial intelligence.