The AI revolution is not just about developing increasingly powerful models but also about the widespread adaptation and adoption of these technologies across various sectors of the economy. This dual focus aims to lower costs of existing products and services while fostering new or improved offerings that can drive economic and social development. While the United States and China lead in AI model development, the real challenge lies in ensuring that the benefits of AI diffusion reach all corners of the globe.
Experts anticipate that AI will follow a J-curve trajectory. Initially, significant investments in physical infrastructure, software, business-model adaptation, data consolidation, and human-capital development may not yield immediate results, resulting in a temporary dip in productivity, broadly defined. However, as the technology’s value-creation potential materializes, it is expected to ascend. The exact nature of this upswing remains uncertain, with investors largely betting on substantial returns, although skepticism persists about whether AI will meet these lofty expectations. The outcome will depend substantially on how effectively AI is diffused across various sectors.
Currently, AI adoption varies significantly among industries. Sectors like technology, finance, and professional services have embraced AI more rapidly, while larger employment sectors such as healthcare and construction lag behind. If these disparities continue, they could lead to a flatter J-curve, reflecting muted returns on investments and delays in growth and productivity enhancements. Whether the current AI investment climate represents a bubble will largely hinge on the diffusion patterns observed in the coming years.
Software-as-a-service (SaaS) providers are seen as crucial channels for AI diffusion. Companies like Google, Microsoft, Salesforce, and Adobe are already integrating AI into their products. The technology can also quickly enhance scientific processes, aided by major developers offering application programming interfaces (APIs) for the rapid creation of tailored AI models. Open-source models, which have gained traction particularly in China, further encourage specialization and competition from smaller firms and countries that may lack the infrastructure for large-scale models. However, barriers like reliable electricity, computing capacity, and mobile-internet access remain significant hurdles to broad adoption.
Trade in advanced semiconductors and human capital are also vital for successful AI diffusion. Economies must provide adequate access to skills ranging from advanced AI engineering to user-level capabilities through education and reskilling initiatives. Additionally, the efficacy of AI models is contingent upon the availability of quality data; fragmented, incomplete, or inaccessible data systems can impede training efforts.
While private-sector initiatives play a major role in AI diffusion, policy frameworks and regulatory environments are equally important. China, for instance, has adopted a proactive stance on this front. Ren Zhengfei, founder of Huawei, noted that China’s approach focuses on leveraging AI to tackle real-world challenges. The government encourages the development of open-source models and directs large tech platforms to build applications across specific sectors such as healthcare, autonomous driving, and green technologies, even sponsoring developer competitions to spark innovation.
This strategic direction has yielded results; China accounted for over 30% of total global manufacturing output in 2024 and 54% of all robot installations worldwide. In fact, the country now hosts nearly half of the world’s installed robots, totaling just over 2 million. Compared to the U.S., China’s policies are more engaged and oriented toward guiding applications and adoption across various sectors. In contrast, U.S. tech firms and well-funded AI startups are primarily focused on pushing the boundaries of large models, often chasing artificial general intelligence and superintelligence. Despite open channels for diffusion, the responsibility for application remains largely with the private sector.
This reliance on private actors may work for sectors like technology and finance, where resources and expertise are abundant. However, it is unlikely that such entities alone can address the multiple obstacles—data fragmentation, regulatory hurdles, capacity deficiencies—that inhibit AI adoption across other sectors. The resulting two-speed diffusion pattern could lead to suboptimal economic growth, unfair distributional outcomes, and weakened national security foundations.
The U.S. government has long recognized the necessity of some level of state guidance in ensuring that innovation aligns with public goals. A similar approach is required for AI diffusion. Implementing a hybrid, pragmatic, and sector-specific strategy is essential to facilitate broader economic benefits, prevent negative distributional effects, and maintain the economic underpinnings of national security. As the landscape evolves, merely observing and hoping for progress is not a viable strategy.
