From "Computing Power Competition" to "National Capability Competition": Jensen Huang and Ro Khanna discuss how the U.S. can win the AI era

Writing: Techub News Compilation

In this public dialogue centered on “America’s Leadership in Artificial Intelligence,” NVIDIA founder and CEO Jensen Huang, U.S. Congressman Ro Khanna, and moderator H.R. McMaster discussed not just chips, models, and export controls, but a bigger question: when AI becomes a new general-purpose technology, what does a country rely on to stay ahead? The answer is not just the technology itself, but a comprehensive capability involving talent, energy, manufacturing, university systems, policy design, social trust, and national narratives.

From the content, this dialogue contains at least three main threads: first, AI is not a single-point technology but a multi-layered industrial system; second, if the U.S. wants to maintain its lead, it must not only emphasize frontier innovation but also rebuild manufacturing capacity, expand technology diffusion, and benefit more ordinary workers; third, in the face of global competition, especially with China’s complex industrial ties, the U.S. cannot simply “de-risk” to stifle innovation nor allow uncontrolled globalization to erode domestic industries and social cohesion.

What’s more noteworthy is that this discussion did not stay in the binary opposition of “tech optimism” or “AI panic.” Huang emphasized that AI will reshape industries, but “task automation” does not mean “jobs will disappear”; Ro Khanna reminded that even if long-term productivity gains create more jobs, the transition period of technology diffusion may still be accompanied by significant unemployment, income polarization, and regional imbalance. Therefore, what truly matters is not whether to develop AI, but how to develop it in a more socially inclusive manner.

AI is not just a model but an entire set of industrial infrastructure

Huang repeatedly emphasized in the dialogue that one of society’s biggest misunderstandings about AI is to see it as a single model or product. According to him, AI is fundamentally a “five-layer stack” industrial system: the bottom layer is energy, above that are chips, then cloud and AI factories and other infrastructure, then models, and at the top are applications.

This judgment is very critical because it expands “AI competition” from model capability competition to a national-level infrastructure competition. That is, whether a country can stay ahead in the AI era depends not only on a few star model companies but also on whether there is sufficient electricity, sustainable chip supply, robust data centers and cloud infrastructure, a thriving model ecosystem, and most importantly—whether AI applications truly penetrate industries and society, forming large-scale usage.

Huang especially stressed that if the U.S. is strong in the first four layers but the application layer cannot diffuse, the entire industrial flywheel cannot turn, and the technology’s value cannot be truly amplified. His concern is not that the technology is not advanced enough, but that society might overreact with fear and excessively regulate AI, even “regulating out industry and society.” If application diffusion is artificially suppressed, then even if the U.S. pioneers this industrial revolution, it may not fully reap its dividends.

From this perspective, the core of AI policy is not just “risk control,” but “reducing barriers to effective use.” These barriers can be institutional, psychological, or public opinion-based. If a country constructs AI as a mere threat at the societal level rather than a tool to learn and master, it risks losing the window for technology diffusion through self-doubt.

America’s advantage lies not only in enterprises but also in open talent and university systems

Ro Khanna’s answer to “why does the U.S. still have a chance to lead in AI” complements Huang’s industry perspective. He believes the U.S.’s greatest comparative advantage is its ability to attract global talent to study, research, start businesses, and collaborate in the U.S., followed by a strong research university system, then a culture of academic freedom and questioning authority, and a relatively mature mechanism for translating research into industry between universities, government, and private sectors.

In this discussion, Khanna specifically mentioned that many AI startups are founded by immigrants, and many AI researchers did not complete their undergraduate education in the U.S., yet they ultimately come to the U.S. to innovate. This “absorbing global talent and forming high-density collaboration locally” mechanism is one of the fundamental sources of America’s technological leadership.

He also pointed out that the importance of research universities cannot be underestimated. America’s long-term accumulation in basic research, talent cultivation, and technological spillovers is not accidental but closely related to sustained public investment. In other words, when discussing America’s AI advantage today, one must acknowledge the foundational role of national research funding and university systems, not just capital markets and leading companies.

This is why, although the dialogue was led by a star entrepreneur and a congressman, its underlying logic is not “enterprise万能” or “government万能,” but a tripartite collaboration: the government provides long-term guidance and institutional environment, universities supply talent and basic research, and enterprises drive industrialization and large-scale application.

Reindustrialization becomes a new keyword in AI competition

While previous AI discussions focused more on computing power, models, and capital, this dialogue’s most vivid reality is its clear inclusion of “reindustrialization” into the AI agenda. Khanna openly stated that a major mistake in the U.S. over the past decades was the illusion that it could focus solely on finance and innovation hubs without maintaining a strong industrial base. This not only harms national security but also weakens social cohesion and leaves long-term deprivation in many regions.

He mentioned that the decline of traditional manufacturing is not an abstract macro trend but has concretely destroyed the dignity, employment, and intergenerational identity of many communities. Cities and families once supported by factories, steel, and industrial chains are forced to face the reality that “if you can’t get into finance or tech, you’re out.” This rupture is ultimately reflected in American political anger, division, and distrust.

Therefore, Khanna advocates a “21st-century Marshall Plan” style new economic patriotism: the U.S. cannot just posture with tariffs but must truly rebuild key industries, forming new industrial investment capabilities around rare earths, critical minerals, pharmaceuticals, robotics, and advanced materials, and reorganize efforts among government, industry, technology, and labor.

Huang added a practical complement here. He believes the AI industry itself is becoming a driver of America’s reindustrialization. As AI factories, chip plants, and computing infrastructure land in the U.S., related construction has created demand for many manufacturing, construction, electrical, piping, and precision tool jobs, pushing up wages in related fields. He also said that companies are planning large-scale investments in U.S. manufacturing, and the prerequisite for this trend is that the U.S. must maintain a sufficiently vibrant, profitable, and investment-friendly industrial environment.

This means AI is not just a “labor substitute” technology but could also be an opportunity to rebuild the real economy and regional employment. But whether it can be an opportunity depends on policies guiding capital into long-term development rather than short-term arbitrage.

“Will AI take jobs?” is not a simple yes-or-no question

Regarding AI’s impact on employment, the most widely circulated part of this dialogue is Huang’s direct rebuttal to the “AI will destroy jobs” narrative. He believes portraying AI as a massive job destroyer is not only inaccurate but also harms society’s acceptance of technology.

He cited a famous example: years ago, some leading AI scholars predicted that as AI penetrated image interpretation, radiologists would become “irrelevant” within ten years. Huang admits the first part was correct—AI has indeed penetrated nearly every aspect of radiology; but the second part was wrong—radiologists did not decrease; they increased.

Why? His explanation is that the “purpose” of a profession and the “specific tasks” within it are not the same. AI can automate certain tasks but does not necessarily eliminate the profession itself. On the contrary, when AI improves efficiency, organizations can serve more patients, handle more demands, and generate higher income, which in turn requires more professionals for higher-level judgment, collaboration, and service.

He extends this logic to software engineering. At NVIDIA, software engineers widely use AI assistant tools, and the result is not that engineers are replaced but that “more AI-savvy engineers” become more valued and successful, and teams can push more projects faster. In other words, AI first changes the organization of work and productivity boundaries, not simply reducing jobs by headcount.

However, Khanna offers a necessary correction. He does not deny that long-term technological progress will create new demands and jobs, but history also shows that productivity growth from the Industrial Revolution onward does not automatically or fairly distribute benefits to everyone. The diffusion process often involves significant unemployment, widening income gaps, and long-term exclusion of certain groups from benefits.

Therefore, responsible policies are not just repeating slogans like “technology will eventually create more jobs,” but considering at the adoption stage: Do workers have bargaining power? Can they share in productivity gains? How can young people and entry-level workers access new opportunities? Can those first hit hardest by change receive training, protections, and reemployment support during the transition?

This also explains why Khanna positions himself as an “AI democratist,” not a “AI doomer” or “AI accelerationist.” His core stance is not against AI but against the concentration of AI benefits in capital while costs are borne by ordinary workers.

The real danger is not AI itself but that only a few will know how to use it

Huang offered a very representative judgment on employment: most people may not “lose to AI,” but they are likely to “lose to those who know how to use AI.” The emphasis is not on creating anxiety but on pointing out the direction of technology diffusion—rather than fearing AI, it’s better to get more people to learn how to use it.

He sees AI as one of the fastest-adopted technologies in history mainly because its usage barrier is much lower than many foundational technologies of the past. Ordinary people do not need to be chip engineers or algorithm researchers; they can turn AI into an augmenting tool in their professional scenarios. He even gave an example: a carpenter, with AI, can better express designs and thus elevate to a service provider closer to architectural or interior design.

The logic behind this is: the most important social value of AI is not to keep professional knowledge confined to a few institutions but to partially spill over high-threshold cognitive and expressive abilities to more people. When more ordinary workers, entrepreneurs, and students can use AI to accomplish more complex work, the technology dividend can truly diffuse.

Khanna further elevates this “diffusion” to a social contract level. He believes that the reason Americans are highly skeptical of AI today is not just because they don’t understand the technology but because many ordinary people no longer trust elite institutions and no longer believe that the next wave of technological revolution will automatically bring opportunities for them. To restore this trust, it’s not enough to just promote; visible employment plans, skills training, regional investments, and public commitments are needed.

Between China, globalization, and regulation, the U.S. needs a “middle way”

Another highly sensitive topic in this dialogue is how the U.S. should handle its relationship with China and the global supply chain. Huang’s stance is very clear: the world is interconnected, and AI and its related industrial chains are not systems that a single country can fully isolate. Any approach of “closing everything off and cutting others out” could lead to serious unintended consequences.

He repeatedly emphasized that AI is not a single product but a deeply embedded complex industry within the global supply system. From energy, minerals, equipment to manufacturing, the U.S. is deeply dependent on other countries, including China. Therefore, policy-making cannot be simplified or emotional but must carefully evaluate long-term consequences, chain reactions, and overall systemic balance.

Khanna agrees and adds that the U.S. cannot fully “decouple,” but the unrestrained globalization of the past has proven unsustainable. The U.S. needs to rebuild a “more bounded openness”: recognizing the risks of Chinese monopolies in key resources and components, pushing for rebalancing and domestic capacity building; and also acknowledging that the U.S. should not fall into a political mood of anti-China, anti-immigrant, or anti-international talent.

Huang also issued a very important reminder: opposing China’s national competitors should not slide into anti-Chinese, anti-immigrant, or anti-international talent sentiments. Because one of America’s core assets is “the world’s best talent wants to come here.” If the U.S. treats competition as identity enemies, it damages not others but its own most valuable asset—the attraction of top global talent and the “American Dream” brand.

On regulation, their views are not as divergent as some might think. Khanna advocates for establishing moderate, detailed rules to keep U.S. AI competitive and trusted globally; Huang emphasizes regulating application scenarios and usage, warning against premature, rigid regulation of foundational technologies still evolving rapidly.

In summary, neither supports the extremes: unbounded laissez-faire or overly restrictive safety measures. The feasible path is to maintain a dynamic balance among risk governance, industrial development, and global competition.

This dialogue is truly about a new national narrative

If you only see this as an “AI policy seminar,” you underestimate its significance. More deeply, it discusses whether, at a moment when AI is reconstructing the economy and social order, the U.S. can rebuild a national narrative that most people believe they are part of.

Khanna repeatedly mentioned that one of America’s most serious issues today is the loss of confidence. People no longer believe they can share in growth, trust that national institutions prioritize ordinary workers, or that the “American Dream” still applies to the next generation. Therefore, he advocates using AI as an opportunity to rethink what should be the guiding star of technological development—not just technological breakthroughs but building a more cohesive, diverse society where more people have a sense of security and upward mobility.

Huang, from an entrepreneurial perspective, offered another inspiring response. He believes that now is one of the best times for young people to enter society, use AI, participate in entrepreneurship, and reshape industries. Because this wave of technological revolution is not about patching the old world but resetting the entire computing industry and nearly all sectors built on computational foundations. For students and young workers, this means unprecedented equal starting opportunities.

In this sense, the most important consensus of this dialogue is not that “America will definitely win,” but that “if America wants to win, more people must win.” Leadership in the AI era is not just about having the best chips, most capital, and most advanced models, but about whether it can weave technology, industry, education, manufacturing, governance, and social trust into a shared community project.

Perhaps this is the most valuable takeaway this dialogue leaves for the outside world: future AI competition, on the surface, appears as a technological race among companies and nations, but in essence, it is a contest over who can build a more complete, open, and resilient national capability system. And the true determinant of victory is often not the loudest slogans but whether one can answer three questions simultaneously: who will innovate, who will manufacture, and who will benefit.