Jensen Huang Predicts Again: The Market Will Reach $90 Trillion When AI Meets the Real Economy

On March 10, 2026, local time, Jensen Huang, the founder of NVIDIA, published a long signed article, mentioning the "five - layer cake" framework of AI (Artificial Intelligence). The article reviewed the past year, and Jensen Huang summarized that AI has crossed an important threshold - the model performance has been significantly improved and can be put into large - scale use; the reasoning ability has been enhanced, the hallucination phenomenon has been reduced, and the ability to implement applications has been greatly improved; the applications built based on AI have created real economic value for the first time.

Sorting out Jensen Huang's statements on different occasions in the past three months, the creation of this "real value" first stems from the implementation of agents. At NVIDIA's Q4 earnings conference in February this year, Jensen Huang said, "Agent AI has reached an inflection point in development. Its practicality has been widely verified among global enterprises, and the resulting demand for computing power has increased explosively."

Since February, the rapid implementation of OpenClaw (an open - source AI agent) in multiple industries around the world has confirmed the judgment of this inflection point. The popularization and maturity of agents have enabled the interaction between AI and humans to move beyond the single dialogue window and start to handle people's real tasks in the virtual world through computers and networks.

However, beyond the wave of agents, a technological wave with great potential and whose impact has not been fully released has begun to be included in the task lists of global AI giants - the world model or physical model oriented towards the real world. This model is considered capable of understanding the spatial and physical logic of the real world, allowing AI to truly integrate into real - life and production scenarios.

From the CES speech in January this year to the above - mentioned earnings conference, Jensen Huang has repeatedly emphasized the potential of this route. In February this year, NVIDIA also announced a long - term strategic partnership with Dassault Systèmes, a French industrial software giant, to jointly build a shared industrial AI architecture for key business scenarios in various industries. SOLIDWORKS and CATIA under Dassault Systèmes are both mainstream design software in the industrial field.

At the Q4 earnings conference, Jensen Huang said, "We are witnessing the development wave of agent AI, and the next wave will be physical AI - applying AI and agent systems to physical fields such as manufacturing and robotics. This field will bring us huge development opportunities."

AI Moving towards the Real World

What capabilities does an AI that can understand the real world need to have?

Jensen Huang gave the answer in his keynote speech at the 2026 CES. The real world has some basic characteristics. For example, it has constancy, "If I put something here and look back, it's still there"; it has a cause - and - effect relationship, "If I push it, it will fall."

However, for an AI to understand the real world, it also needs to master physical laws such as friction, gravity, and inertia, and know that a heavy - duty truck needs a longer braking distance. These are common sense to humans but completely unfamiliar to AI.

The past AI revolution was essentially a breakthrough in the "symbolic space". From BERT to ChatGPT, large models have learned to understand the grammar, semantics, and context of language and can even complete complex reasoning tasks, but they know almost nothing about the physical laws of the real world - gravity, friction, inertia, and causality. A large model that can write beautiful prose doesn't know what will happen when a stone rolls down a hillside. "Physical AI" was born to fill this gap. Jensen Huang defines it as "AI that can understand the laws of nature", and its core is to enable AI to not only process language symbols but also truly understand and interact with the physical world.

He believes that this will be the next stage of AI development and will also have a huge impact and bring about great changes to the world. From an industrial perspective, AI that understands the real world will reshape the huge automobile, transportation, and manufacturing industries through autonomous driving and robotics. In the first three months of this year, NVIDIA has launched multiple architectures and products centered around the physical world.

It's not just NVIDIA. In the first three months of 2026, while the Agent (agent) craze swept the world, Silicon Valley has been actively making arrangements around world AI.

Google DeepMind's Genie3 was officially opened to the public at the end of January. Users can input text to let Genie3 generate an interactive 3D environment in real - time. Waymo then transformed it into a dedicated simulation tool for autonomous driving in February, which is used to generate extreme scenarios that a fleet is hardly likely to encounter on real roads, such as tornadoes, flooded roads, or an elephant suddenly appearing at an intersection.

More symbolically, Yann LeCun, a Turing Award winner, left Meta at the end of 2025, founded AMI Labs, and announced the completion of a $1.03 billion seed financing on March 10, with a pre - investment valuation of $3.5 billion, becoming the largest seed - round financing in European history. NVIDIA and Samsung are also among the investors.

LeCun's judgment is that large language models are a dead end because they cannot truly understand how the physical world works. His new company is betting on the JEPA architecture - an AI framework that doesn't predict every pixel detail but learns to understand the world structure, with target application areas including healthcare, robotics, and industrial automation.

Meanwhile, World Labs founded by "the mother of AI", Fei - Fei Li, completed a new round of $1 billion financing in February, with a valuation approaching $5 billion. Its first product, Marble, has been launched, focusing on generating 3D worlds with correct physical laws. The CEO of AMI Labs predicted after the financing that within six months, every company will claim to be a "world model company" to raise funds. This prediction itself shows the popularity of this track.

It has become quite clear in 2026 that AI should better understand the real world.

The "Inflection Point" of AI in Manufacturing

On February 3, 2026, local time, after finishing his Asian trip, Jensen Huang rushed to Houston, the United States, and appeared at the Dassault Systèmes 3DEXPERIENCE World conference.

There, he and Pascal Daloz, the CEO of Dassault Systèmes, jointly announced a strategic cooperation called "the largest in 25 years" - deeply integrating NVIDIA's accelerated computing and AI capabilities with Dassault Systèmes' virtual twin platform. Dassault Systèmes has been established for more than forty years, and its 3DEXPERIENCE platform serves more than 45 million users and 400,000 customers. This French company is one of the deepest - rooted software suppliers in the global manufacturing industry - its influence can be seen behind almost everything from aircraft engines to consumer product casings.

This cooperation points to an important field: industry. In the past, although large language models have played an important role in some fields, their applications in the industrial field have still been very limited due to a lack of reliability and understanding of the real world. The core goal of the cooperation between the two parties is defined as building an "industrial world model" - an AI system that is scientifically verified and rooted in physics, which can serve as a key - task platform in the fields of biology, materials science, engineering, and manufacturing. At the same time, through its cloud brand OUTSCALE, Dassault Systèmes deploys "AI factories" on three continents around the world based on NVIDIA's latest AI infrastructure, aiming to provide enhanced functions for AI models on the 3DEXPERIENCE platform while ensuring the privacy and sovereignty of customer data.

Gian Paolo Bassi, the global senior vice - president of the professional customer division of Dassault Systèmes, said in an interview with media such as Economic Observer that many large - model companies cannot create new atomic structures, nor can they develop new alloys, new types of aircraft, or aerospace equipment because their focus is mainly on language models and they do not have the professional knowledge to develop a drug or a new device. The core advantage of Dassault Systèmes lies precisely in the fact that this "hard knowledge" has been precipitated in the software. Bassi said, "Our knowledge is in the software. There is some industrial - related professional knowledge, and in this field, Dassault Systèmes has a unique advantage."

This means that artificial intelligence must reconstruct the system based on professional knowledge. Take the medical device industry with strict regulatory requirements as an example. The traditional verification cycle is long and costly, but with the assistance of AI, engineers can simulate many different situations, complete products more effectively and quickly, and with higher quality.

The greater ambition is that with the assistance of AI, the tests that used to require repeated production of physical prototypes can now be completed in the digital world with extremely low costs through a large number of parallel iterations. This means that the entire supply - chain process from raw materials to assembly and the production process can be reconstructed in the virtual world.

Gian Paolo Bassi, the global senior vice - president of the professional customer division of Dassault Systèmes, said in an interview with media such as Economic Observer, "Dassault Systèmes has built highly realistic virtual twin models over the years. The cooperation with NVIDIA enables these models to run on a large scale, with high precision, and close to real - time, and be directly used by AI, thus upgrading the virtual twin from an engineering tool to a sustainable system - level capability."

The concept of Virtual Twin is not new. It describes the use of digital models to accurately map physical systems, allowing engineers to test and iterate in the virtual world and then apply the conclusions to the real world. Dassault Systèmes has been working in this direction for many years, and its technical system is quite mature. However, for a long time, the large - scale implementation of this technology has faced a fundamental bottleneck: computing power. Sufficiently realistic and complex physical simulations require far more computing power than before. Now, this bottleneck is being broken.

Jensen Huang said that in the past, industrial enterprises spent one - third of their time on design and digitization and more time on building physical forms. In the future, 100% of the time can be spent on digitization. From producing a pair of tennis shoes to producing a car, whether it's design, description, simulation, or operation, "it's all defined by software."

After this precise virtual reconstruction is completed, the combination of artificial intelligence and robotics can almost reshape the manufacturing process and efficiency.

A factory is not a single entity but a collection of millions of objects. Artificial intelligence can help simulate all parts of this collection in the digital world and arrange the production line more reasonably, organizing the robots to operate in the factory.

In large - scale manufacturing enterprises, this scenario has long appeared: the maximum - extent virtual twin and widely deployed robots. However, the cost of achieving all this is high, which also means that robots can only survive in industries with highly repetitive tasks and large task volumes, such as the automotive industry, where a robot is specially programmed to do one thing.

In Jensen Huang's view, this is also the value of AI entering the industrial field. Through the improvement of simulation modeling efficiency and the intelligence level of robots by AI technology, small and medium - sized enterprises that account for the majority of the global supply chain can also use these cutting - edge technologies, which will undoubtedly reshape the efficiency of the entire industry.

Jensen Huang said, "Knowledge is a very large industry, but the truly huge industry is the one that combines information and the real world, with a value of $90 trillion."

In the eyes of technology believers, the arrival of physical AI may be as significant as when the Internet reduced the cost of information circulation to almost zero. If that revolution reshaped the production and distribution of information, this revolution will reshape the design and operation mode of the physical world itself.

Pascal Daloz, the CEO of Dassault Systèmes, said, "We are entering a new era where AI is no longer limited to predicting or generating content but begins to truly understand the physical world. When AI is rooted in science, physics, and verified industrial knowledge, it will become a multiplier of human wisdom."

In Daloz's view, when AI enters the physical world, its achievements do not lie in replacing designers and engineers. "Success does not lie in automation. Engineers don't want to automate past achievements; they want to create the future."

This article is from the WeChat official account "Economic Observer". Author: Song Di. Republished by 36Kr with authorization.