Nvidia's Jensen Huang predicts: "Agent-based AI" will reshape the $50 trillion real economy, with computing power demand soaring by 1000%

The AI revolution enters its second half! Nvidia CEO Jensen Huang boldly announced at the 2026 ServiceNow Conference, “Agentic AI” will reshape the physical economy sector worth up to $50 trillion. He straightforwardly stated that future factories will be entirely operated and managed by robots, and this autonomous AI’s computing power demand has skyrocketed by 1,000% compared to two years ago. To prevent AI agents from “running amok” and destroying corporate data, Nvidia has partnered with ServiceNow to launch a strict AI governance sandbox, heralding a new era of “digital workforce” management.
(Background summary: Jensen Huang: Blackwell and Rubin should not be allowed to acquire China, NVIDIA’s market share in China remains zero but he refuses to back down)
(Additional background: Jensen Huang sends company-wide email embracing OpenAI Codex: Over 10k NVIDIA employees are already onboard, GPT-5.5 runs on GB200)

Table of Contents

Toggle

• Consuming the $50 trillion physical economy, computing power demand soars 1,000%
• Preventing AI from running wild, jointly launching “Digital Employee” governance sandbox
• FedEx’s real-world validation

The leader of the world’s most valuable company, Nvidia CEO Jensen Huang, doesn’t need to fly to Las Vegas himself, but he chose to stand on the main stage of ServiceNow Knowledge 2026 for the third consecutive year, sending a clear message to global enterprises: A revolution in the economy, 100 times larger than the software industry, has already begun.

“I’m here because I need my ServiceNow,” Huang Huang told the audience with a smile. This is not just polite talk; internally at Nvidia, workflows are built on ServiceNow; the supercomputer complex quotes that once took 5 days now only take 5 minutes to generate.

Consuming the $50 trillion physical economy, computing power demand soars 1,000%

In a live interview with CNBC, Jensen Huang proposed a bold thesis: the software industry is about to swallow an economy 100 times larger than itself. He pointed out that traditional industries like manufacturing plants, warehouses, and logistics networks, worth up to $50 trillion, have largely not been deeply touched by IT and software; but the emergence of Agentic AI will fundamentally overturn this equation:

“The entire production line will be operated by robots, managed by more robots, and the entire factory itself will be a robot.”

“They will run on ServiceNow, and you will manage them just like human employees.”

To support these AI agents capable of reading, reasoning, using tools, and generating massive tokens in real-time, Huang Huang revealed an astonishing figure: the computational power required for Agentic AI has surged 1,000% compared to generative AI two years ago. This will undoubtedly be the strongest tailwind for Nvidia’s future chips.

Preventing AI from running wild, jointly launching “Digital Employee” governance sandbox

However, empowering AI with autonomy also comes with huge risks. ServiceNow CEO Bill McDermott issued a stern warning at the outset, citing an example where an AI agent of a company gained high-level permissions and deleted the entire production database—including customer data and backups—in just 9 seconds. McDermott said:

“Governance is no longer just a feature; it is the core of the entire game. Without it, the whole company could collapse.”

To fill this critical vulnerability, ServiceNow announced a series of major updates for AI agents:

• AI Control Tower: A unified governance layer that allows companies to monitor all active AI agents, models, and workflows in real time.
• Project Arc and Open Shell: Enterprise-grade security “sandbox” developed in partnership with Nvidia. Huang Huang explained that this will ensure each AI agent has the same identity management, access control, and governance policies as human employees.
• Action Fabric: Allows any external AI agent (such as Claude, Gemini) to call regulated engines to “actually execute” enterprise tasks, not just provide suggestions.

FedEx’s real-world validation

The logistics giant FedEx’s testimony becomes the best footnote for this AI revolution. FedEx CEO Raj Subramaniam revealed that since 2020, the company has been building a digital twin of its supply chain, generating up to 2 PB of data daily, and executing 5 million workflows per month via ServiceNow.

FedEx’s CIO Vishal Talwar summarized the core attitude of enterprises facing AI agents:

“We see them as a ‘digital workforce’ that must be governed with the same rigor and policies as managing human teams.”

As AI moves from “thinking” to “acting,” a fundamental paradigm shift in enterprise management is unstoppable.