US and China AI race: How the superpowers are battling for supremacy

How is the US and China AI race being measured?

The determination of who leads the global artificial intelligence competition depends entirely on the metrics used to define success. Currently, the United States holds a distinct advantage in the development of 'frontier models'—the most advanced, high-performing systems capable of complex reasoning. According to researchers at Stanford University, as of March 2026, the top model from Anthropic led the field by a margin of just 2.7%.

American dominance is fueled by massive capital inflows into leading AI laboratories. Firms such as OpenAI, Google, and Anthropic are investing enormous sums into the infrastructure required to train next-generation models. This investment has turned AI into a primary driver of the US economy, with the stock market's expansion heavily reliant on the projected profitability of these frontier companies.

The struggle for frontier dominance

While the US leads in raw capability, the margin of victory is narrowing. In February 2025, a model from the Chinese firm DeepSeek briefly matched the performance of top US models, demonstrating that the technological gap is not insurmountable. The US strategy focuses on proprietary, high-cost systems, whereas the Chinese approach seeks to democratize capability through different technical means.

What advantages does China hold in the AI competition?

China is pursuing a strategy of economic and geopolitical leverage by making capable AI cheap, open, and ubiquitous rather than focusing solely on the 'moon landing' style achievement of frontier models. Instead of the proprietary subscription models favored by US firms, Chinese developers are releasing 'open weight' models. These models, such as Moonshot AI's Kimi K3, are downloadable and adaptable, allowing for rapid integration across the Chinese economy.

Beyond software, China possesses significant structural advantages in the physical requirements of AI deployment:

• Energy Costs: According to Rebecca Arcesati of the Mercator Institute for China Studies, Chinese AI companies benefit from significantly lower electricity costs, a critical factor for energy-hungry data centers.
• Robotics Leadership: The International Federation of Robotics estimated that over two million robots are currently working in Chinese factories.
• Humanoid Production: According to Chatham House, China accounted for more than 90% of global humanoid robot shipments in the first half of 2026.

Despite US-led restrictions on advanced microchips, Chinese developers have shown resilience. Analyst Lian Jye Su from Omdia notes that Chinese firms have innovated despite the shortage of high-end hardware, with some US officials alleging that Chinese developers use 'distillation' techniques to learn from advanced American platforms.

How do chip restrictions impact the technological landscape?

Sweeping government restrictions on the sale of advanced microchips to Chinese firms remain a primary tool in the US strategy to maintain a technological edge. These chips are essential for training the cutting-edge models that define the current frontier. By limiting China's access to the most sophisticated hardware, Washington aims to slow the pace of Chinese breakthroughs in Artificial General Intelligence (AGI).

However, these restrictions have not halted Chinese progress. The rapid release of new models in both nations suggests that the 'finish line' of the AI race remains undefined. The competition has shifted from a simple race for hardware to a complex battle of software efficiency and industrial application. While the US attempts to gatekeep the most advanced tools, China is focusing on the mass deployment of capable, accessible technology.

Why do the US and China differ on AI safety and regulation?

The two superpowers are following fundamentally different philosophies regarding the governance and safety of artificial intelligence. The US approach, particularly under the Trump administration, prioritizes rapid innovation and deregulation to ensure American labs do not lose their competitive edge. President Trump has dismissed existential risk concerns as a 'hoax' and suggested that the US already possesses sufficient rules, arguing that the primary goal should be preventing China from closing the gap.

In contrast, China has adopted a more centrally coordinated and cautious regulatory framework. President Xi Jinping has emphasized that AI must remain under human control. Beijing has introduced guidelines on data usage and algorithms well ahead of many other nations. According to Jayant Dave from Check Point Software Technologies, this approach aims to manage risks while providing a structured environment for developers.

Divergent risk perceptions

The motivations behind these regulatory stances are also distinct. While the US debate often centers on the existential risks posed by AI to humanity, Chinese authorities appear more focused on the technology's potential to threaten domestic stability or be misused in cyber warfare. This difference in priority makes international cooperation on AI governance exceptionally difficult.

Will the US and China cooperate on AI governance?

Cooperation remains uncertain, even as the two nations face the shared challenges of managing a transformative technology. While US and Chinese officials, including US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng, have discussed establishing a notification mechanism for serious AI incidents, the underlying tension remains high. Historically, the two nations have cooperated on existential threats like nuclear weapons, but AI lacks a similar shared understanding of the threat.

The current climate is characterized by a 'race narrative' that many analysts fear could escalate into an unhealthy rivalry. As Leia Wang from the Carnegie Endowment for International Peace notes, while the inclusion of AI on the Trump-Xi summit agenda is a positive step, the risk of an escalating confrontation remains a significant concern for global stability.

Frequently asked questions

What is the difference between frontier and open weight models?

Frontier models are the highest-performing, most advanced AI systems, typically owned by companies like OpenAI or Google and accessed via paid platforms. Open weight models, which China focuses on, are downloadable and can be modified by users, making them cheaper and more widely adaptable for various industries.

How are chip restrictions affecting China's AI development?

US restrictions prevent Chinese firms from purchasing the most advanced microchips required to train cutting-edge AI. While this has slowed their ability to create top-tier frontier models, Chinese companies have responded with innovation, such as using distillation techniques to learn from existing advanced models.

What is Artificial General Intelligence (AGI)?

AGI is a theoretical form of artificial intelligence that would possess the ability to learn, reason, and perform any intellectual task at or above the level of a human being. Both the US and China are currently racing to achieve this milestone.

Why is energy consumption a factor in the AI race?

Training and running massive AI models requires immense amounts of electricity. China has a competitive advantage in this area because its AI companies can access electricity at a significantly lower cost than many of their international competitors.

How does China's robotics industry relate to the AI race?

Robotics is a physical application of AI. China has become the world's largest manufacturer of robots, including humanoid robots, which allows them to integrate AI technology directly into the global manufacturing supply chain more effectively than many other nations.