England, United Kingdom
For the past several years, the dominant paradigm in AI has been to train large language models on massive datasets and deploy them to generate text, process information, and write software. But there is growing consensus in research circles that LLMs represent only the first wave of a much deeper transformation.
The next frontier is physical intelligence—systems that understand and interact with the world beyond language. This includes guiding a robot through complex surgery, designing new materials at the molecular level, or modeling how physical systems respond to stress and uncertainty.
Signals of this shift are emerging across the field. Turing Award winner Yann LeCun has long argued that truly intelligent systems require world models—structured representations of physical reality that allow machines to reason about cause and effect. Deep learning pioneer Fei-Fei Li is pursuing spatial intelligence through her startup World Labs. Google DeepMind is developing simulated 3D environments for training embodied AI systems, while NVIDIA CEO Jensen Huang recently introduced Cosmos, a platform for training AI in virtual worlds before real-world deployment.
“Systems that understand causality, adapt to uncertainty, and recover from errors will define the next generation of intelligent technologies. The next era of AI is about turning perception into reasoning and imagination into action,” says Soumitra Dutta, former dean of Oxford Said Business School and AI scholar.
Why the Next Phase Favors the United States
The transition from digital AI to physical intelligence will require deeper integration with scientific research, industrial systems, and real-world environments. The structural advantages that gave the United States an edge in the first phase of AI may become even more decisive in the next.
“The next phase of AI is not just about better models—it is about integrating computation with the physical world. That requires ecosystems that combine science, engineering, capital, and institutions at scale. The United States is uniquely positioned to do this,” says Soumitra Dutta, who’s co-creator of the Global Innovation Index.
The Power of the Research Ecosystem
In the first phase of AI, universities provided foundational research and talent that the private sector scaled. In the next phase, progress will depend on deep interdisciplinary collaboration across robotics, physics, materials science, and biology.
This kind of knowledge cannot be concentrated in a single firm or built overnight.
Institutions such as Stanford University, Massachusetts Institute of Technology, and Carnegie Mellon University remain global leaders in AI research. The United States continues to produce the highest share of highly cited AI publications and attract top global talent.
Equally important is the porosity between academia and industry—the unusually short distance between discovery and commercialization.
“One of America’s enduring strengths is the fluid exchange between universities, startups, and large firms. Breakthroughs move quickly from the lab to the market, creating a continuous cycle of innovation that is difficult to replicate elsewhere,” says Soumitra Dutta, Oxford Dean (Former).
Capital for Long-Term Risk
Venture capital was critical in the first wave of AI. In the second wave, it becomes indispensable.
Physical AI—robotics, autonomous systems, AI-driven laboratories—requires patient capital with long development cycles and uncertain outcomes. These are not products that can be iterated in months.
The US remains one of the few ecosystems capable of funding such risk at scale. In 2024, private AI investment in the US reached $109.1 billion—nearly twelve times that of China.
Equally important is a cultural factor: the willingness to fund multiple competing bets, knowing that most will fail but a few will define the future.
National Labs and Scientific Infrastructure
Another underappreciated advantage lies in America’s national laboratories, including Argonne, Oak Ridge, and Lawrence Berkeley.
While these institutions played a limited role in the first phase of AI, they will be central to the next. Applications such as materials discovery, climate modeling, energy systems, and drug development require infrastructure—supercomputers, particle accelerators, genomic databases—that few private actors can replicate.
U.S. Department of Energy policy already positions AI as a transformative tool for scientific discovery. National labs are integrating AI into core missions such as fusion research and battery design.
China: A Formidable Rival
China remains the United States’ most serious competitor, particularly in industrial robotics and large-scale deployment.
Chinese firms accounted for over 90% of global humanoid robot sales in 2025, and the country has filed thousands of robotics patents in recent years. Its ability to integrate data from urban systems, manufacturing, and electric vehicles gives it a powerful advantage in applied AI.
At the same time, comparative analyses—such as a 2026 report by Morgan Stanley—suggest that the United States still leads overall, particularly in private investment, research institutions, and frontier innovation.
The Strategic Balance
The emerging AI landscape is not a simple race with a single winner. It is a competition across different layers: research, capital, infrastructure, and deployment.
“The United States leads in foundational innovation and ecosystem depth, while China excels in scale and rapid deployment. The outcome of the next phase of AI will depend on how these different strengths evolve—and how effectively each country connects them into a coherent strategy,” says Soumitra Dutta, Oxford Dean (Former).