Jensen Huang: Compute Is the New Economy

At the Morgan Stanley Technology, Media & Telecom Conference in San Francisco, NVIDIA CEO Jensen Huang outlined what may be a consequential reframing of the technology industry. His argument can be distilled into a single equation: Compute equals tokens, tokens equal intelligence, and intelligence equals economic output at every level, from companies to countries.

Huang placed this moment within the broader arc of AI development, describing three inflection points: generative AI, reasoning AI and now agentic AI—with systems that can autonomously plan, research and execute. And that requires an extraordinary number of tokens—the units of data that models process and produce—dramatically increasing the computation required and economic implications for companies and countries. Huang spoke during a fireside chat with Morgan Stanley Co-President Dan Simkowitz, who underscored the firm’s long-standing relationship with Huang, as Morgan Stanley took NVIDIA public in 1999.

Huang’s comments focused on three themes shaping AI’s impact in this new paradigm:

1. How rising compute demand translates directly into higher token generation, revenue and GDP
2. Why efficiency, measured in tokens per watt, is becoming a critical CEO-level decision tied to topline company growth; and
3. How agentic AI is poised to expand and evolve the software industry.

1. More Compute, More Tokens, More Revenue and GDP

Agentic AI can consume 1 million times more tokens than a standard generative prompt, which is causing compute demand to skyrocket, Huang said. He redefined data centers as AI factories whose primary output is tokens, and connected compute directly to corporate revenues and national gross domestic product. For companies and countries, having access to higher compute and tokens enables AI agents that can drive economic output.

“I’m certain compute equals revenues,” Huang said. “I'm certain also that compute equals GDP. Therefore, every country will have it because not one country in the future will say, ‘Guess what? You know, we're going to opt out on intelligence.’”

2. Tokens Per Watt: The New CEO-Level Metric for Growth

With data centers as a direct driver of top-line growth for companies, hardware architecture is a CEO-level strategic decision. Huang had a message for CEOs and corporate leaders at the conference: Since AI factories are capped by number of gigawatts, the tokens per watt, or performance per unit of energy, becomes the single most important metric for a company’s revenue growth. Power and infrastructure limitations mean companies cannot afford to make inefficient choices, Huang said.

“AI is a factory because factories are power-limited. It doesn't matter how many plants you have. Each plant is still 100 megawatts or gigawatt, and therefore tokens per watt is the single most important thing for the top line of companies. And [CEOs] have to make those decisions very, very carefully.”

3. Expansion of Software Industry Through Agentic AI

The traditional software market business model, which is today largely licensing tools, may be headed toward renting specialized AI agents to customers. Following high-profile releases that demonstrated a step-change in agentic AI capability, the software market has reflected uncertainty about how AI will reshape the industry. Huang made a categorical prediction: There will be no software in the future that is not agentic. He argued it would be irrational to have software that is “dumb” when intelligence is available.

Thus, software companies will become massive token consumers—either by producing tokens themselves, reselling tokens from providers like large language models, or renting out specialized AI agents that are powered by tokens. They will need tokens as an input to then offer agents as an output, expanding addressable revenue beyond pure software licensing.

“The IT industry is a couple trillion dollars today. They’re tool renters. In the future, they'll rent agents that use those tools, which means that the software industry in the future will be much larger than the software industry of today,” Huang said.