^{Wealth & Poverty Review} Jensen Huang’s Information Theory of Economics

As the author of a book, “Life After Capitalism,” on the information theory of economics, I reveled last week in vindication by none other than Jensen Huang, CEO of Nvidia (NASDAQ: NVDA).

In the study of the economy today, experts everywhere uphold economics as “the science of scarcity,” governed by “incentives,” measured by money, minted and controlled by governments. Individuals, companies, and nations, driven by need and greed, are seen to struggle to amass scarce material resources and harvest their tokens of government money. Cramped by a limited earth, bathed in noxious life-giving CO2, afflicted with “climate change” catastrophe, Elon Musk even seeks to escape our world and harvest wealth across the universe in outer space.

As the founder and guide for 33 years of Nvidia, the world’s most valuable company here on earth, Jensen Huang surely grasps the “inner loops” and dynamics of wealth.

In a nearly two-hour interview at Morgan Stanley last week, the incandescent thinker in his black leather jacket made it explicit and undeniable. Wealth is not material but mental.

“You can’t build wealth by printing money, but you can by printing brains—intelligence,” Huang said.

Nvidia’s system has an “inner loop” which uses “tokens” to produce intelligence. Think of the role of human written language in life. Tokens are the written language of artificial intelligence.

The “inner loop” of token generation uses only 5% of the power to yield 95% of the intelligence. Human minds are the “inner loop” of the world economy and the ultimate source of all its value and wealth.

The word for the key form of industrial intelligence in the modern world is “compute.” It’s performed by machines governed by humans. Data centers, he said, merely store and retrieve data. No one can be sure that such data can be converted into value. Nvidia, by contrast, creates “AI factories” that manufacture “tokens,” which are units of value that enable AI output. Produced by intelligence, the tokens begin as components of text, numbers, or pixels. In what I call Huang’s “information theory of economics,” generative artificial intelligence (AI) models turn electricity into intelligence measured by tokens per watt per second.

What is scarce, according to Huang, is not materials, but human minds and human time. Tokens are forms of information that are transformed by AI models into intelligence, which produces wealth.

As I put it, wealth is knowledge, created by learning new information, defined by its surprise. As Information Theory inventor Claude Shannon put it, information is “unexpected bits,” gauged by their surprisal. Money is time, defining the scarcities of human hours and years. As Huang says, in human intelligence and labor, “we are always millions of people short.”

Performing physical simulation of images governed by the speed of light, the ultimate measure and boundary of time, Nvidia began by mastering the mathematical essence of 3D graphics. From the beginning, Nvidia’s success came from its mastery of time.

Now Huang has purchased Open Claw, which he calls “the most important software ever” because it provides humans with agents that they can talk to (agents with I-O) to perform computer tasks. In just three weeks, its sales exceeded the 30-year total of Linux, the previously dominant open-source tool.

Open Claw, says Jensen, is the new “Windows.” Just as Microsoft Windows provided the operating system for the PC, Open Claw supplies the operating system for agents that you can tell what to do. You don’t have to type and manipulate software, you merely talk and dictate tasks to agents. “Every SAS (software as a service) company will become a GAS company (generative AI services).”

Humans don’t want to work, he said; they want to have work done through the use of tools. Open Claw is a tool for human intelligence to get work done. Open Claw can affect a kind of renaissance of AI agents specialized in particular domains.

But as Huang explains, AI is what happens inside the building. Now he is moving outside the building, into the physical world, to provide the intelligence for self-driving cars and robotics. For these apps, Nvidia provides open-source programs such as Alpamayo for autonomous cars, GROOT for robots, and La Proteina for biology. Again, exploiting Nvidia’s full stack knowledge of the material world begun in its development of 3D gaming—with light-simulation in rendering 3D images—the system gives mastery of time and context. It provides for such crucial factors as gravity and object permanence in its best-selling physical model called COSMOS.

But wealth remains particular. As Huang insists, “There is no such thing as a universal accelerated computer system. Nothing about our system is compatible with anyone else’s system.” It’s just compatible with human minds and workloads. Like the human mind, it is “full-stack.” Nvidia builds full stack systems that can operate with human guidance in the physical world.

Huang’s scheme reduces to an information theory of economics, with tokens per watt per second the key to revenues for the top line of companies and the ultimate source of national and world gross domestic product (GDP). None the less, the ultimate source of wealth remains in the thrall of the inexorable scarcity of time and human genius amplified by Nvidia’s AI omniverse.

In the view of the information theory of economics, however, Huang may express excessive confidence in his multi-gigawatt full-stack “claws” as compared to the human model that runs on 12 to 14 watts. With its ever-growing complexity and energy voracity, the stack may become vulnerable and even obese.

After all, the key measure of information and intelligence is surprisal. Emerging in the semiconductor industry, advocated by Elon Musk with his coming “Terafab” as a new model of intelligence beyond chips, is “wafer-scale integration.” It’s a surprise that can vastly multiply tokens per watt per second.

The current AI world will have to deal with this 12-14 watt surprise and turn it into wealth.