US-China AI Rivalry: Power, Chips and Research

As global superpowers race to dominate artificial intelligence, a comprehensive breakdown by Crimson Reporters reveals a stark division in capabilities between Washington and Beijing. While American enterprises command the lion’s share of hardware and financial investment, Chinese laboratories demonstrate remarkable prowess in scholarly output and are rapidly narrowing the performance gap in cutting-edge algorithms.

This technological competition occupies center stage as United States President Donald Trump and China’s President Xi Jinping prepare for a pivotal summit in Washington, DC on Thursday. Diplomatic discussions are slated to navigate thorny subjects including international commerce, artificial intelligence, regional tensions concerning Taiwan, and the ongoing US-Israel war on Iran.

Ahead of these high-stakes negotiations, US Treasury Secretary Scott Bessent disclosed that Washington put forward a diplomatic framework dubbed an AI “notification mechanism.” This proposed protocol functions essentially as a dedicated emergency hotline, designed to alert both administrations immediately should unforeseen artificial intelligence anomalies pose direct threats to national security.

The upcoming bilateral encounter arrives against a backdrop of urgent warnings from prominent tech developers regarding the profound hazards tied to automated systems. Both governments continue funneling massive resources into their respective technology sectors to outpace the rival nation in this technological arms race.

The Hardware and Computing Advantage

Executing and training modern machine learning architectures demands staggering quantities of computational muscle. Nations equipped with superior processor units and expansive hardware networks accelerate model development exponentially, establishing infrastructure as a primary battleground for supremacy.

Computing capacity is typically evaluated in floating-point operations per second, quantifying how many calculations a processor handles within a single second. Aggregating the processing output of every specialized chip within a nation’s borders yields a clear picture of its aggregate technical capacity.

By these analytical metrics, the United States commands a commanding lead, claiming nearly three-quarters of global artificial intelligence processing capabilities. Conversely, Epoch AI data indicates that China controls slightly over 14 percent of worldwide computational output.

American dominance stems largely from preferential access to cutting-edge microchips. Research from Stanford University’s 2026 AI Index highlights that US-based manufacturer Nvidia commands upwards of 60 percent of total global capacity among premier chip designers, whereas Chinese telecommunications giant Huawei maintains a modest yet expanding portion of the market.

Infrastructure and Frontier Innovation

Massive processing clusters require immense physical housing facilities. The United States hosts in excess of 5,400 commercial data facilities—surpassing other nations by roughly tenfold—and maintains 84 facilities engineered exclusively for machine learning operations, outstripping the next eight countries combined.

Simultaneously, the race for elite generative systems intensifies. Frontier models represent the pinnacle of automated engineering, exemplified by OpenAI’s GPT series, Anthropic’s Claude, and specialized frameworks originating from Chinese laboratories like DeepSeek.

Fuelled by vast datasets and immense processing power, these sophisticated platforms execute complex reasoning, computer coding, and natural language tasks. Although American developers traditionally pioneered these elite systems, competing Chinese architectures are swiftly catching up in overall capability.

• US companies control nearly 75% of global AI computing power.
• China holds approximately 14% of the world’s total AI capacity.
• The US operates over 5,400 data centres globally.