On April 8, numerous committees within the United States House of Representatives held hearings on AI, examining China’s growing capabilities, the release of DeepSeek’s R1 reasoning model, and potential implications for U.S. security and economic interests. These conversations attempted to untangle what is more important for U.S. strategic interests: building the most advanced and capable AI technology, potentially at the cost of widespread global adoption, or following China’s approach by focusing on building a new global technology ecosystem where potentially less capable models could be adopted and deployed rapidly at scale.
Recent evidence suggests it may be beneficial for the United States to pursue the latter strategy. Smaller, more resource-efficient, and localizable models are gaining significant traction globally, potentially rivalling the impact of compute-intensive frontier systems in user adoption metrics. This is exemplified by the recent releases of models from Chinese firms like DeepSeek and Alibaba; smaller in size and, therefore, more efficient to run. They have quickly achieved high rates of international adoption despite, or perhaps because of, their relatively modest size.
Although dozens of countries participate in AI development at various stages, only a few countries, notably the United States and China, are able to scale and produce the most compute, data, and talent-intensive AI models due to the immense amount of resources required. This gap may only widen as these two countries continue to pour investment into frontier model development, AI applications, computing infrastructure, and energy systems. Therefore, the AI ambitions of most countries will be interlinked and dependent on developments in the United States or China.
Due to this dynamic, AI competition between the United States and China is often framed in terms of their state-of-the-art AI capabilities. However, this view is misleading and overlooks critical dimensions of AI leadership. Different approaches, such as promoting reliable and user-friendly AI systems in international markets, developing practical business or government AI applications, and creating AI that functions effectively across varied contexts, offer strategic advantages that often go unnoticed in policy debates on international AI competition.
For the United States to maintain its current competitive edge and global influence in AI, it must acknowledge this reality and actively export and promote its AI products to the world. Getting this right will require any AI promotion strategy to pay sufficient attention to three key attributes: quality, reach, and adaptability.
The United States and China: Diverging Strategies for Global AI Leadership
The U.S. and China are each pursuing their own distinct strategies to secure their positions as global leaders in AI, putting different emphasis on technological dominance versus diffusion, the global adoption of technologies. Yet, when it comes to technological innovation, diffusion matters. History shows that “being the first” to achieve a given technological breakthrough does not necessarily translate into lasting market leadership. What matters more is how widely diffused and adopted the technology becomes. The same is likely to be true for AI. Simply reaching new thresholds of frontier capabilities, creating the world’s largest model, or building the world’s largest compute cluster may not produce a definitive or long-lasting strategic advantage. The current approaches of the United States and China toward AI engage with this dynamic in different ways.
To date, the U.S. strategy for global AI leadership has largely centered on the concept of control, particularly of computing resources via export controls. When coupled with a strong tendency towards proprietary models by U.S. firms, this gives rise to a relatively closed ecosystem. Models developed and released by the U.S. AI industry currently remain the most advanced globally and enjoy high market penetration in developed economies. Additionally, the United States has leveraged its significant advantages in computing to effectively determine which states can and cannot develop cutting-edge AI. In the short term, this approach guarantees that the United States will maintain its lead at the frontier of AI development by prioritizing technological advantage over broad adoption. In the long term, this strategy may lead other countries to look elsewhere for their technology needs, namely to China. This scenario is already playing out today. Fearing their dependence on the U.S. technology ecosystem, some countries are developing new sovereign digital capabilities and seeking alternatives for their AI needs.
In contrast to the United States, and despite U.S. export controls, recent Chinese AI advancements, such as those released by DeepSeek, Alibaba, Huawei, Zhipu, and Tencent, have showcased substantial progress in the country’s AI ecosystem and global competitiveness. Many of these releases are especially well-suited for localized adoption at a low cost to users. Combining these technological advances with longstanding government-led efforts to export Chinese-produced digital infrastructure globally has created a strong foundation for the widespread adoption of Chinese AI solutions, both domestically and internationally. This may prove to be more significant in the long term than advances in frontier capabilities alone. For example, open source repositories already indicate that Chinese models are achieving notable global download rates, with lightweight versions of DeepSeek and Qwen frequently ranking high in adoption metrics.
These divergent yet nascent approaches to AI development and deployment reflect broader strategic choices about how technological influence will spread globally.
AI Diffusion: The Importance of Quality, Reach, and Adaptability
At present, it remains uncertain which strategy — the current U.S. focus on technological superiority and control or the Chinese approach of global coalition building and diffusion — will be the most successful for achieving and maintaining AI hegemony. However, if the United States wants to seriously compete with China and guarantee that U.S. AI systems enjoy global adoption, any new AI strategy must focus on three essential attributes for effective technology promotion: technical accuracy and reliability (quality), global user accessibility (reach), and the ability to respond and adapt to the diverse needs of businesses and communities worldwide (adaptability).
Quality represents an AI model’s actual capabilities, performance, and reliability. Excelling in AI quality signifies being at the forefront of development in ways that truly matter to users and institutions. Ensuring that U.S. AI is consistently of the highest quality will require advancements in assurance mechanisms: the specific governance processes, evaluation methodologies, and verification systems that substantiate performance claims and risk mitigation strategies. Additionally, high-quality AI must reliably work in different environments and under diverse sets of conditions. This will build trust and, in turn, increase the likelihood of others adopting the technology. This is particularly important in markets where multiple systems compete for integration and adoption. Institutions like the National Institute of Standards and Technology in the United States will be instrumental for continued leadership in this domain.
Reach explains how widely adopted and accessible an AI system is. Fundamental to ensuring higher levels of reach is the presence of the necessary underlying digital infrastructure that enables access to AI capabilities in the first place. Without the necessary infrastructure, developing and deploying even the most basic AI systems may remain out of reach for significant portions of the global population that lack access to computing resources, potentially leading to a rapid increase in inequality.
Similarly, AI systems unable to function effectively across diverse environments and resource-constrained contexts will fail to achieve widespread reach and adoption. Therefore, the most successful AI systems will be those that are able to demonstrate compatibility with existing digital ecosystems and technological infrastructures. Due to China’s work on building digital infrastructure, it enjoys numerous advantages in exporting its AI systems.
Adaptability refers to an AI system’s ability to function effectively across diverse linguistic, cultural, and operational contexts. Open source represents one of the clearest ways to achieve high levels of adaptability by enabling communities to customize and tailor AI systems to meet their unique needs, though there are likely security tradeoffs to this approach. Adaptability will also be heavily influenced by the steps developers take during model training to ensure that a wide variety of use cases, languages, and contexts are considered. Many Chinese AI companies are actively working to compete with U.S. models by not only open sourcing their solutions, but also ensuring that their training data includes support for a number of typically underserved languages and cultures at a rate that surpasses that of leading U.S. companies.
A strategy grounded within these principles will be more likely to succeed in improving and expanding current diffusion efforts. As the United States navigates evolving global AI competition, balancing these elements will be crucial in determining whose AI systems — and by extension, whose approaches, values, and standards — shape the global technological landscape for decades to come.
Toward Meaningful Technological Leadership
Emphasizing the interconnected attributes of quality, reach, and adaptability will provide U.S. policymakers with a clearer perspective for conceptualizing the country’s global technological influence. By balancing technical excellence with deployment breadth and contextual adaptability, this approach recognizes the multidimensional nature of leadership in AI.
For U.S. policymakers, this highlights several strategic priorities:
First, the adoption of AI will depend heavily on trust in a given system. This will require investments in mechanisms, such as risk mitigation strategies, that reinforce trust in the quality and capabilities of a specific AI system, in addition to supporting innovation.
Second, the United States should invest in the necessary institutional capacities to support global AI deployment and benefit sharing that aligns with commercial and national security interests. This might include more robust engagement on the topic of digital public infrastructure with the global community or supporting existing government organisations, such as the Export–Import Bank, International Development Finance Corporation, or the Trade and Development Agency.
Finally, the United States should support a regulatory agenda that enables and facilitates new mechanisms and practices for AI deployment, placing adaptability at the forefront. This approach will ensure that American-made AI products become the preferred choice worldwide across a wide variety of business and societal use cases.
As the global AI landscape evolves, the systems that achieve widespread integration will not necessarily be the most technically advanced, but rather those that best balance quality, reach, and adaptability. This multidimensional understanding of competition does not diminish the importance of frontier innovation, but complements it with equally crucial considerations about how technologies spread and where they gain traction.
