NVIDIA co-founder and CEO Jensen Huang, in a recent interview at the Hillen Valley Forum, provided an in-depth insight into his perspective on AI development. He positioned AI as an entirely new industrial revolution, painting a grand blueprint of how AI will reshape various industries in the 21st-century economy and how human society will adapt to this transformation.
AI Factory: The Wisdom Source of the New Era
Huang first explained what an "AI Factory" is. He pointed out that AI is not just a new technology, but its construction is fundamentally different from past software, capable of performing tasks that previous software could not.
More importantly, AI's production mode has undergone a transformation. "In the past, software was produced by humans inputting code," Huang said, "Now, we have a new industry where software is produced by machines." These machines, namely large supercomputers, operate through electricity, producing Tokens that can be recombined into various forms of intelligence, such as numbers, text, protein structures, images, videos, and 3D models.
"I call it an AI Factory because it does only one thing every day: producing Tokens."
These intelligence Tokens produced by AI Factories will permeate various fields, including healthcare, financial services, engineering, supply chain management, and especially the education sector, which Huang is particularly optimistic about.
He believes that just as electricity played a role in past industrial revolutions, the intelligence produced by AI Factories will fundamentally transform and innovate every existing industry. For example, future automotive companies will not only manufacture physical cars but also establish AI Factories to produce the intelligence (Tokens) that drive these cars.
"In ten years, every car company will also produce Tokens that run in those cars." This transformation suggests that any company manufacturing physical products may need an AI Factory to produce the "brain" required for their products. The direct market impact is a surge in demand for computing power, energy, and related infrastructure, positioning companies like NVIDIA that provide underlying technologies at the crest of the wave.
AI's Evolution Wave and the Future of 'Physical AI'
Huang reviewed the development of modern AI, dividing it into several stages.
About 12-14 years ago, the breakthrough in computer vision represented by AlexNet initiated the "Perception AI" wave, enabling machines to understand the meaning of images, sounds, vibrations, and temperatures. Subsequently, around five years ago, "Generative AI" became the focus, with AI models learning to understand and transform information, such as translating English to French or generating images based on text prompts, like a universal translator.
Currently, we are in the era of "Reasoning AI". This AI not only understands and generates but can also solve problems and identify unprecedented situations. They use reasoning capabilities similar to humans, gradually breaking down problems and applying learned rules and principles to solve them.
"We call it Agent AI, which has agency," Huang explained. These digital robots can understand tasks, learn, use computers, browsers, and other tools to execute tasks for humans, such as accessing SAP systems to handle supply chain issues or accessing Workday for human resources matters.
He foresees that future CEOs will manage both physical and digital workforce, and IT departments might transform into "human resources departments" for agent AI.
The next wave will be "Physical AI".
This requires AI to understand physical laws such as friction, inertia, causality, and common sense in the real world. For example, objects cannot pass through tables, balls that roll off tables will land on the floor, not disappear into another dimension. AI with these physical reasoning capabilities, when placed in physical robots, will give birth to "robotics".
Huang believes this is crucial for the future of manufacturing in the US and globally. "When we build new factories and plants across the US, we want to use the latest technology. We hope that in the next decade, these new generation factories and plants will be highly robotized to help us address the severe global labor shortage." This vision suggests massive market opportunities for the robotics industry, sensor technology, and related software development.
Global AI Competition and the US Response
Facing the global AI competition, Huang offered his view on how the US government should respond. He emphasized first understanding the nature of this competition as an "infinite game" rather than a time-limited contest. NVIDIA's 33-year development journey, spanning the personal computer revolution, internet revolution, mobile revolution, and now the AI revolution, embodies this long-term thinking.
He analyzed the winning keys from three aspects of AI:
- Technical Level: Intellectual capital is crucial. Jensen Huang reminds that 50% of global AI researchers are of Chinese descent, a factor that must be considered strategically. This means talent attraction, cultivation, and international cooperation will be key.
- AI Factory Level: Energy is core. AI factories operate by converting electricity into digital Tokens, similar to how past industrial revolutions transformed energy into physical products or electricity itself. Therefore, sufficient and cost-effective energy supply is the foundation for developing AI factories.
- Infrastructure and Application Level: Jensen Huang points out that the winners of the previous industrial revolution were not the countries that invented the technology, but those that applied it fastest, with the United States being a prime example. Therefore, for AI, the key is active application, not fear. This includes enhancing workforce skills to apply AI and encouraging societal adoption of AI technology.
Jensen Huang's perspective undoubtedly provides clear guidance for policymakers and market investors. In the AI race, focusing solely on technological R&D is insufficient; energy policies, talent strategies, and promoting industrial applications will be equally, if not more, critical. Short-term market reactions may be reflected in increased attention to energy stocks, AI infrastructure concepts, and educational technology fields.
The Real Impact of AI on the Job Market: Transformation, Not Replacement
Addressing concerns about potential mass unemployment due to AI, Jensen Huang offered a more nuanced view: "New jobs will be created, some jobs will disappear, but every job will be changed." He emphasizes avoiding extremes and analyzing the issue from first principles.
At the fundamental technological level, AI's development itself creates new employment opportunities. Using San Francisco as an example, Huang points out how the city has been revitalized by AI. "AI has created a new type of work because software development methods have changed. Software that was previously coded by humans and ran on CPUs is now generated by machine learning and runs on GPUs." This means that from tools, compilers, and methodologies to data collection, management, and AI security protection, new technologies and positions are emerging at every layer.
At the AI factory level, massive employment opportunities are emerging. Huang uses a 1 GW AI factory as an example, with construction costs reaching $60 billion, equivalent to Boeing's annual revenue.
"Building it requires financing, creating numerous jobs; it needs factory construction, including carpenters, steelworkers, and masons." Subsequently, mechanical, electrical, and plumbing engineers, as well as IT, network, and operational maintenance professionals, with the entire construction and startup cycle taking about three years. "The demand for these new technical workers will be enormous."
He believes that in the past computer industry transformation, software engineers were the critical bottleneck; in the AI factory era, technical workers will be the most crucial element. "I think this is fantastic. Our country needs to recognize that technical workers are respectable, critical work necessary for building our nation." This statement suggests that vocational education and skill training systems need corresponding adjustments to meet market demands for new types of technical workers.
At the application level, AI agents will change the work methods of professionals like doctors, financial service professionals, and customer service representatives. Using NVIDIA as an example, each software engineer is equipped with an AI assistant, significantly increasing code submission volume and overall productivity. "Our productivity has soared, and we are hiring more people because AI enables us to create more of what the world needs, increasing our revenue and hiring capacity."
His advice is: "It's not AI that will steal your job, nor will AI destroy your company, but companies and individuals using AI will steal your job." This statement emphasizes the importance of actively embracing and learning AI, which is a race for future competitiveness for both individuals and enterprises.
Manufacturing Returning to the US?
Recently, discussions about manufacturing returning have been very heated. Apple CEO Tim Cook previously stated that one of the main bottlenecks in moving iPhone manufacturing back to the US is the lack of excellent and precise robotic arm technology. Huang believes AI will be the key enabling technology for manufacturing upgrade and repatriation.
"Advanced manufacturing is now software-driven, with the entire factory being a massive robot coordinated by a master robot operating numerous smaller robots." He praised the US government's efforts to encourage and support manufacturing localization, believing it offers excellent opportunities for high-quality, high-tech work for the nation.
He particularly emphasized the concept of "Digital Twins". NVIDIA itself designs complex chips entirely in a digital twin environment, with these chips existing in the virtual world and undergoing detailed simulation months before their physical manufacture. "We should do the same for digital factories. For these large factories, we should completely create digital twins, use AI to create these digital twins, perform virtual integration, operations, optimization, and use them for output planning."
He predicts that in the future, every factory, person, car, building, and city will have its digital twin version. This trend will drive massive demand for simulation software, sensor networks, data analysis, and AI algorithms, forming new economic growth points.
Regarding when AI robots will become prevalent in daily life, Huang offered an optimistic expectation. Using autonomous driving cars as an example, Waymo, after about ten years of development, now provides services in multiple US cities. "The time needed for robots will be even shorter."
He explained that robot operating environments can be more constrained compared to cars, not needing to adapt to all streets and conditions like automobiles. "From a functional prototype to a mass-produced product, it takes about five years. We already have quite powerful robots today."
Therefore, he anticipates that in the next five years, we will see robots mass-produced from factories, and existing automotive manufacturers, by mastering software and AI components, can effectively transition to robot manufacturing.
Please refer to the following video for the complete interview:
