Huang Renxun: AI won't take jobs away—in fact, it will lead to labor shortages
TechFlow Selected TechFlow Selected
Huang Renxun: AI won't take jobs away—in fact, it will lead to labor shortages
AI is actually a five-layered cake.
Navigating Web3 tides with focused insightsAuthor: Facing AI
At the 2026 World Economic Forum, Jensen Huang, CEO of Nvidia, sat down for a conversation with Larry Fink, CEO of BlackRock.
The discussion centered on AI’s technological evolution, the scale of infrastructure development, AI’s impact on employment markets, and its implications for the global economy.
Huang positioned AI as a profound, foundational transformation that will reshape the global economic and social structure. He described it as an unprecedented wave of infrastructure construction in human history—one that will redefine the value of labor and offer unparalleled opportunities for balanced global economic development.
Huang’s central argument is that we are undergoing a fundamental “platform shift.”
He likened the rise of AI to the emergence of personal computers, the internet, and mobile cloud computing, noting that each such shift has completely reshaped the computing stack and given birth to entirely new application ecosystems.
In his view, AI is not merely individual applications like ChatGPT or Claude, but rather a new foundational platform upon which everything else can grow. The defining breakthrough of this shift is that computers have, for the first time, acquired the ability to understand “unstructured information.”
Previous software systems—such as SQL-based databases—could only process pre-defined, structured data. AI, by contrast, can interpret complex, context-rich unstructured data in real time—images, sounds, natural language—and perform tasks based on reasoning about human intent.
This transition—from “pre-recorded” to “real-time generation”—is the essential characteristic that distinguishes AI from all previous technologies.
To help the audience better grasp the scope of this vast industry, Huang introduced a “five-layer cake” model.
1. Energy: At the very bottom is energy. Because AI processes and generates intelligence in real time, it requires power.
2. Chips and Computing Infrastructure: The second layer is my domain—chips and computing infrastructure.
3. Cloud Infrastructure: The next level up is cloud services.
4. AI Models: Above that are AI models. This is where most people think AI resides. But don’t forget—you need all the layers beneath to make these models possible.
5. Application Layer: But the most important layer, and the one currently exploding, is the application layer. Last year was incredible for AI, frankly, because progress in AI models enabled the application layer—the ultimate layer we all depend on—to flourish. This includes applications in financial services, healthcare, manufacturing, and more. Ultimately, economic value will be created here.
This model was introduced to emphasize the depth and breadth of the AI industry and to support his key assertion about infrastructure investment.
Based on this framework, Huang declared that we are witnessing “the largest infrastructure build-out in human history.” He believes that the hundreds of billions of dollars already invested are just the beginning, with trillions more expected to flow into this space.
This is not hyperbole, but a logical necessity driven by how the AI platform operates.
To process massive contextual information and generate intelligence, there will be exponential growth in global demand for energy, data centers, chip fabrication plants, computer factories, and AI factories.
He cited large-scale factory construction plans from partners like TSMC and Foxconn, along with major investments by Micron, Samsung, and others in memory chips, to underscore the reality and urgency of this infrastructure wave.
When asked about a potential “AI bubble,” Huang offered a market-based response: The rental prices for Nvidia GPUs in the cloud—across both the latest and older generations—are continuously rising, indicating that real demand far outstrips supply.
Therefore, current massive investments are not irrational speculation, but necessary construction to close a significant supply-demand gap.
On the impact of AI on employment, Huang presented a view sharply at odds with prevailing anxieties. He argued that rather than causing mass unemployment, AI could actually lead to labor shortages in certain sectors. He explained this through a distinction between a job’s “purpose” and its “tasks.”
Using radiologists as an example, a decade ago it was widely predicted that AI’s computer vision capabilities would eliminate this profession. Yet ten years later, the number of radiologists has increased.
Huang explained that AI automated the task of reading scans, allowing doctors to work more efficiently and spend more time on higher-value activities—diagnosis, patient communication, and collaboration with other clinicians—that better fulfill their core purpose.
Increased efficiency enables hospitals to serve more patients, leading to higher revenue and thus the hiring of more doctors. Similarly, nurses freed by AI from tedious documentation can focus more on human-centered care, improving both patient capacity and hospital performance.
Extending this logic, he believes AI will become a powerful tool for professionals across industries, automating repetitive tasks and thereby enhancing their ability to achieve core objectives—ultimately boosting productivity and value across entire sectors.
Moreover, the infrastructure build-out itself will create numerous skilled blue-collar jobs—electricians, construction workers, technicians—that do not require advanced degrees, helping foster more inclusive economic growth.
Huang remains optimistic, believing AI has the potential to narrow, rather than widen, the global technology divide.
His core argument lies in AI’s usability. He stated that AI is “the easiest-to-use software in history”—users no longer need to learn complex programming languages; they simply issue instructions in natural language to drive powerful AI systems.
This extremely low barrier to entry allows individuals from developing nations and those without advanced computer science education to participate in this technological revolution.
He further introduced the concept of “national intelligence” or “sovereign AI,” strongly advising every country to build its own AI infrastructure and train AI models using its native language, culture, and data.
He believes that a nation possessing its own AI capability is akin to having its own electricity grid or transportation network—it is foundational to future national competitiveness. This concerns not only economic development but also cultural preservation and technological sovereignty.
Regarding Europe, Huang noted that Europe possesses a remarkably strong industrial and manufacturing base, which may not have been fully leveraged during the U.S.-dominated software era.
However, in the AI era—especially with Nvidia’s recent advances in “physical AI”—Europe now has a once-in-a-lifetime opportunity.
He encouraged Europe to deeply integrate its robust manufacturing capabilities with artificial intelligence, shifting from a mindset of “writing AI” to “teaching AI,” thereby achieving leapfrog development in smart manufacturing and robotics. Additionally, Europe’s rich scientific research tradition can be combined with AI to dramatically accelerate the pace of scientific discovery. He urged European leaders to take energy supply and infrastructure investment seriously to lay the foundation for a thriving local AI ecosystem.
Full Translation
Larry:
Good morning, everyone. It's great to be back in the Congress Hall. I hope you all had a wonderful day yesterday and are enjoying today as well. I'm honored to introduce Mr. Jensen Huang—a man I deeply admire and have long followed, and who has been a teacher to me on my journey of learning about technology and artificial intelligence (AI).
Watching him lead Nvidia is truly remarkable. I don’t usually measure myself against others, but here’s one comparison I like: Since Nvidia went public in 1999—the same year BlackRock went public—
Jensen:
Oh, my goodness.
Larry:
Yes. Nvidia has delivered a total compounded return of 37% to its shareholders. Just imagine what that would mean if every pension fund had invested at Nvidia’s IPO—we could have secured retirement success for everyone.
Meanwhile, BlackRock has delivered an annualized total return of 21%. Not bad for a financial services company, but clearly overshadowed. Yet this powerfully demonstrates Jensen’s leadership, Nvidia’s positioning, and the world’s belief in Nvidia’s future. So, Jensen, congratulations on your journey—I know we still have many years ahead together.
Jensen:
Thank you. I really appreciate that. My only regret is that after our IPO, I wanted to buy something nice for my parents, so I sold some Nvidia stock when the company was valued at $300 million. I bought them a Mercedes S-Class. That turned out to be the most expensive car in the world.
Larry:
Did they regret it? Do they still have the car?
Jensen:
Oh yes, absolutely. They still have it.
Larry:
Alright. Now let’s get into it. The debate around AI centers on how it will change the world and the global economy. Today I want to talk about how AI can add value to the world economy, and how AI is increasingly becoming a foundational technology that each of us here can use to improve our lives and the lives of everyone on the planet.
We need to discuss how it will reshape productivity, labor, and infrastructure across nearly every industry—but more importantly, how it will reshape the world, and how more parts of the world can benefit from AI, and how we can ensure the global economy broadens rather than narrows.
I can’t think of anyone with a clearer perspective on AI itself and the surrounding infrastructure—the kind that must be built around it. Because so many major hyperscale computing companies rely on products created by Nvidia, and the entire ecosystem revolves around AI infrastructure and its potential, I believe we have a voice worth listening to this morning. So, Jensen, thank you again.
This is his first time at the Davos World Economic Forum, and I know your schedule is incredibly busy—thank you for making the time.
Jensen:
Thank you very much.
Larry:
So I’ll go straight to the point. Why do you believe AI has the potential to be such a powerful engine for growth? What makes this moment, this technology, different from past tech cycles?
Jensen:
Yes. First, when you interact with AI in various ways—of course including using ChatGPT, Gemini, Anthropic’s Claude, etc.—and see the amazing things it can do, it helps to go back to first principles and ask: What exactly is happening in the computing stack?
This is a platform shift. Platforms are the foundation on which applications are built. This platform shift is like the earlier shift to personal computers (PCs), where new applications were developed for a new type of computer; or the shift to the internet, a new computing platform that hosted new applications; or the shift to mobile cloud. In every platform shift, the computing stack is redefined and new applications emerge.
In this sense, this is a new platform shift. The ChatGPT you’re using today is itself an application, but critically, new applications will be built on top of ChatGPT, and others on top of models like Anthropic’s Claude. So this is exactly what a platform shift looks like.
If you realize that AI enables things you could never do before, it becomes easier to understand. Past software was essentially “pre-recorded.” Humans input and described algorithms or recipes for the computer to execute.
It could only handle structured information—meaning you had to enter names, addresses, account numbers, ages, residences, create structured tables, and then the software would retrieve data from them. We call this SQL queries. SQL is the most important database engine ever created, and almost everything once ran on SQL.
Now, we have computers that can understand unstructured information. This means it can read an image, comprehend a piece of text.
These are completely unstructured. It can hear sound and understand its meaning, parse its structure, and reason how to respond. So for the first time, we have a computer that isn’t “pre-recorded” but can process information in real time. It can take in current situational, environmental, contextual information—anything you give it—then infer its meaning and your intent, even if your intent is expressed in highly unstructured ways.
We call these “prompts,” but you can describe it any way you like. As long as it understands your intent, it can perform the task for you.
The key point here is that we’re reshaping the entire computing stack. The question is: What is AI? When you think of AI, you might think of AI models, but it’s crucial to understand AI from an industrial perspective. AI is actually a five-layer cake:
1. Energy: The bottom layer is energy. Because AI processes and generates intelligence in real time, it needs power.
2. Chips and Computing Infrastructure: The second layer is my domain—chips and computing infrastructure.
3. Cloud Infrastructure: The next layer is cloud services.
4. AI Models: Then comes AI models. This is where most people think AI resides. But remember, to make these models possible, you need all the layers below.
5. Application Layer: But the most important layer, and the one currently unfolding, is the application layer. Last year was incredible for AI, frankly, because progress in AI models allowed the application layer—the final layer we all depend on—to flourish. This includes applications in financial services, healthcare, manufacturing, and more. Ultimately, economic value will be created here.
But importantly, because this computing platform depends on all the layers beneath it, it has triggered the largest infrastructure build-out in human history. We’ve already invested hundreds of billions of dollars.
Larry:
Only hundreds of billions.
Jensen:
We’ve only invested hundreds of billions. Larry and I have plenty of opportunities to collaborate on projects.
Trillions of dollars in infrastructure still need to be built. This is logical because all this contextual information must be processed so AI models can generate the intelligence needed to power top-level applications.
So when you trace it back layer by layer, you see the energy sector undergoing extraordinary growth. The semiconductor industry—TSMC just announced plans to build 20 new wafer fabs.
Foxconn, Wistron, and Quanta are partnering with us to build 30 new computer factories whose products will go into AI factories. So we’re seeing chip fabs, computer factories, and AI factories being built around the world.
Larry:
And memory.
Jensen:
And memory, absolutely. Those chip fabs. Micron has started investing $200 billion in the U.S. SK Hynix and Samsung are both performing exceptionally well. You can see the entire chip layer expanding at an astonishing rate. Of course, we’re focused on the model layer now, but excitingly, the application layer above is also doing extremely well. One indicator is last year’s venture capital (VC) funding trends.
Last year was one of the largest VC investment years in history—over $100 billion globally, with most flowing into so-called “AI-native” companies. These span healthcare, robotics, manufacturing, financial services—essentially all major industries. You’re seeing massive investments in these AI-native firms because, for the first time, AI models are good enough to serve as a foundation for building applications.
Larry:
Let’s go deeper. Clearly, I believe everyone here uses their own chatbot to get information. But you mentioned that AI’s accessibility will be key. Let’s talk about the more positive visions of AI’s proliferation into the physical world. You mentioned healthcare as a great example—what transformative opportunities do you see in areas like transportation or science?
Jensen:
Last year, I’d say three major things happened at the AI technology layer—the model layer.
First, the models themselves were initially novel but produced lots of “hallucinations.” But last year, we can reasonably say these models became more grounded. They can conduct research, reason about environments they weren’t explicitly trained on, break problems into step-by-step reasoning, and devise plans to answer questions or complete tasks. So last year, we saw language models evolve into what we call “agentic systems” or “agentic AI.”
The second breakthrough was in open models. About a year ago, Deepseek emerged, which worried many. But frankly, Deepseek was a huge event for most industries and companies worldwide because it was the world’s first open reasoning model. Since then, a wave of open reasoning models has followed. Open models allow companies, industries, researchers, educators, universities, and startups to leverage them to launch projects and create specialized, domain-specific solutions.
The third area of major progress last year was the concept of “physical intelligence” or “physical AI.” This AI doesn’t just understand language—it understands nature. It can be AI that comprehends our physical world, or proteins, chemicals, physics (like fluid dynamics, particle physics, quantum physics). These AIs are now learning all these different structures and “languages”—if you will, seeing proteins as a language.
All these AIs have made such tremendous progress that industrial companies, whether in manufacturing or drug discovery, are advancing rapidly. A great example is our collaboration with Eli Lilly. They realized that due to AI’s extraordinary progress in understanding protein and chemical structures—essentially being able to interact with proteins as naturally as we chat with ChatGPT—we’re going to see some truly groundbreaking discoveries.
Larry:
All these breakthroughs raise concerns about the “human factor.” You and I have discussed this many times, but we need to address our audience: People are deeply worried that AI will replace jobs. And you’ve been arguing the opposite. Clearly, as you said, the construction of AI—the largest infrastructure build-out in history—will create jobs. Energy creates jobs, industry creates jobs, the infrastructure layer creates jobs, land, power, facilities—all of this is incredible.
So let’s dive deeper. You actually believe we’ll face labor shortages. So how do you see AI and robotics changing the nature of work, rather than eliminating it?
Jensen:
We can approach this from several angles.
First, this is the largest infrastructure build-out in human history. It will create massive employment. And the great thing is, these jobs are tied to skilled trades. We’ll need plumbers, electricians, construction workers, steelworkers, network technicians, and people to install and commission equipment. In the U.S., we’re seeing a remarkable boom in these roles. Salaries have nearly doubled.
Larry:
Yes.
Jensen:
We’re talking about six-figure salaries for people building chip fabs, computer factories, or AI factories. And we have severe shortages in these areas. I’m thrilled to see so many countries and people recognizing the importance of this field. Everyone should be able to live a good life—you don’t need a PhD in computer science to do so. I’m glad to see that.
Second, while we theoretically discuss automation of tasks and its impact on jobs, let me share some real-world anecdotes—things that have already happened.
Remember 10 years ago? One of the first jobs people thought would disappear was radiology. The reason: The first superhuman AI capability was computer vision, and one of its biggest applications was radiologists reading scans.
Ten years later, AI has indeed fully permeated every aspect of radiology, and radiologists are using AI to analyze scans. The impact is 100%, completely real. Yet surprisingly—or perhaps not surprisingly, if you think from first principles—the number of radiologists has actually increased.
Larry:
Is that due to lack of trust in AI, or because human-AI interaction leads to better outcomes?
Jensen:
The latter. The purpose of a radiologist’s job is to diagnose disease and help patients. That’s their purpose. The task includes analyzing scans. Now that they can analyze scans nearly infinitely faster, they can spend more time with patients, diagnosing, interacting, collaborating with other clinicians.
Unsurprisingly, hospitals can now see more patients—many previously waited in long queues for scans. With more patients, hospitals earn more revenue and hire more radiologists.
The same is happening with nurses. In the U.S., we’re short 5 million nurses. Using AI for medical note-taking and visit transcription—tasks that used to take half their time—now frees them up. Our partner Abridge is doing excellent work here. Nurses can now spend more time visiting patients.
Larry:
Human care.
Jensen:
Exactly. Because you can now see more patients, we’re no longer bottlenecked by nurse availability. More patients can enter hospitals faster. As a result, hospitals perform better and hire more nurses.
So AI is boosting their productivity, which is unsurprising. As a result, hospitals thrive and want to hire more people. Too many people are waiting too long to get care. These are two perfect examples.
The simplest way to think about AI’s impact on any specific job is to understand the job’s purpose versus its tasks. If you put a camera on both of us and just watched, you might think we’re typists because I’m typing all day. If AI automates word prediction and typing, we’d be out of jobs.
But clearly, that’s not our purpose. So the question is: What is your job’s purpose? For radiologists and nurses, it’s caring for people. That purpose is enhanced, not diminished, when tasks are automated. So if you can distinguish between purpose and task, I think that’s a helpful framework.
Larry:
Let’s expand beyond developed economies. Help me understand how AI benefits the whole world. Last weekend I read an article from Anthropic saying that recently, AI usage has mainly been led by educated social groups—they even observed higher usage among more educated segments within societies. Of course, they based this on their own models, so there might be bias.
So how do we ensure AI becomes a transformative technology like Wi-Fi and 5G were for emerging markets? What does it mean when AI intersects with emerging economies and their labor markets? How do we broaden the global economy? Back to jobs—robotics and AI will inevitably displace some roles. In the U.S., displacement is already happening.
We may create more plumbers and electricians, but fewer financial analysts. Law firms may need fewer associates because they can process data faster. So let’s turn to emerging and developing worlds. How do you see this playing out?
Jensen:
First, AI is infrastructure. I can’t imagine any country in the world that shouldn’t treat AI as part of its infrastructure. Every country has electricity, roads—AI should be part of that too.
You can import AI, but training AI today isn’t that hard. With so many open models available, combining them with local expertise allows you to build models that serve your country.
So I firmly believe every country should participate, build its own AI infrastructure, develop its own AI, leverage your most fundamental natural resources—your language and culture—and refine your national intelligence as part of your ecosystem. That’s point one.
Second, remember, AI is incredibly easy to use. It’s the easiest-to-use software in history. That’s why it’s growing fastest and being adopted most rapidly. In just two or three years, users are approaching one billion.
First, I think Claude is amazing. Anthropic has made huge strides with Claude. We use it everywhere in our company. Its coding and reasoning abilities are truly impressive. Anyone running a software company should be using it. On the other hand, ChatGPT may be the most successful consumer AI in history—its ease of use and friendliness mean everyone should engage with it.
Whether someone in a developing country or a student, it’s clear that learning how to use AI, guide AI, prompt AI, manage AI, set guardrails, and evaluate AI is critical. These skills aren’t different from what leaders and managers do—they’re exactly what you and I do. In the future, besides biological, carbon-based “AI” (humans), we’ll have digital, silicon-based AI, and we must manage them. They’ll become part of our digital workforce.
So I urge developing nations: Build your infrastructure, engage with AI, and recognize that AI is likely to narrow the technology gap.
Larry:
Really?
Jensen:
Because it’s so easy to use, so abundant, so accessible. So I’m actually quite optimistic about AI’s potential to uplift emerging nations. For many without computer science degrees, you can now become programmers. Before, we had to learn how to program computers. Now, you program by asking the computer, “How do I program you?”
If you don’t know how to use AI, just ask AI: “I don’t know how to use AI—how should I?” It will explain. You can say, “I want to write a program to build my own website—how do I do it?” It will ask you a series of questions about what kind of site you want, then write the code for you. It’s that easy. That’s the incredible, exciting power of AI.
Larry:
Two quick questions—we’re running out of time. We’re in Europe now. When we talk about companies, we often mention American and Asian ones. Tell us how AI intersects with Europe’s success and future, and what role you see Nvidia playing in Europe?
Jensen:
I have an advantage: Nvidia is fortunate to work with every major AI company in the world. Because we sit at the base of the infrastructure stack, we power AI across industries—language, biology, physics, and world models related to manufacturing and robotics.
For Europe, what’s truly exciting is that your industrial base is incredibly strong. Europe’s manufacturing foundation is exceptionally robust. This is your chance to leapfrog the software era. The U.S. did lead the software era. But AI is software you don’t write—you teach AI (You don't write AI, you teach AI).
So jump in early and integrate your industrial and manufacturing capabilities with artificial intelligence. This will propel you into the world of physical AI and robotics. Robotics is a once-in-a-lifetime opportunity for European nations. Every European country I’ve visited has an incredibly strong industrial base.
Another key point: Europe’s deep sciences remain very strong. Now, these deep sciences can benefit from applying AI to accelerate discovery. So I believe you must take increasing energy supply seriously, so you can invest in the infrastructure layer and cultivate a rich AI ecosystem in Europe.
Larry:
So what I’m hearing is that we’re far from an AI bubble. The question is, are we investing enough? Let’s flip the question—so many people talk about a bubble, but what I hear from you is: Are we investing enough to broaden the global economy?
Jensen:
A good test for an AI bubble is this: Nvidia now has millions of GPUs in the cloud. We’re in every cloud, everywhere. If you try to rent a Nvidia GPU today, it’s extremely difficult. GPU rental spot prices are rising—not just for the latest generation, but even for GPUs two generations old.
Why? Because the number of AI startups being created, and the number of companies shifting R&D budgets to AI, is increasing. Eli Lilly is a great example. Three years ago, most—if not all—of their R&D budget went to wet labs. But look at their investment in a large AI supercomputer, a major AI lab. Their R&D spending will increasingly go toward AI.
So the so-called AI bubble exists because investment is massive. And it’s massive because we must build the infrastructure required for all layers above AI. So I believe this opportunity is truly extraordinary.
Everyone should get involved. We need more energy. We all recognize we need more land, power, and facilities. We need more skilled technical workers. In fact, this workforce is very strong in Europe.
Larry:
Yes.
Jensen:
In many ways, the U.S. lost this over the past two or three decades. But it’s still strong in Europe. This is an extraordinary opportunity you must seize. Where Larry and I work, we see investment opportunities, and the scale is rising.
As I mentioned, last year was one of the biggest VC investment years in history—over $100 billion globally, mostly in AI-native companies. These AI firms are building the application layer—they’ll need infrastructure, they’ll need our investment, to build this future.
Larry:
I firmly believe that for pension funds worldwide, participating and growing alongside this AI world will be a great investment. It’s also one of the messages I deliver to political leaders: We must ensure ordinary retirees, ordinary savers, can be part of this growth. If they stand aside, they’ll feel left behind.
Jensen:
We want to invest in infrastructure, right? Infrastructure is a great investment. This is the largest infrastructure build-out in human history. Get involved.
Larry:
Our time is up. I hope everyone here and watching online can see the strength of Jensen Huang as a leader—not just in technology and AI, but in business, and as a leader with heart and soul, which is especially important today: leadership that comes from genuine feeling and conviction. Thank you all.
Jensen:
Thank you all.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
