NVIDIA CEO Huang: AI Has Officially Entered Profit Phase

SANTA CLARA, Calif. — NVIDIA CEO Jensen Huang declared on Tuesday that artificial intelligence has officially entered a profitable phase, asserting that the long-debated question of AI investment returns now has a definitive answer. Speaking at NVIDIA’s 2026 Annual Shareholder Meeting, Huang framed the current AI infrastructure buildout as a multi-decade cycle that could become one of the largest infrastructure projects in human history.

“Useful AI has arrived, and it is profitable,” Huang said during his business update, according to 澎湃新闻 (The Paper). “The question of AI investment ROI has an answer.” He described modern AI data centers as “modern factories that produce tokens,” where every token can be converted into code, answers, designs, automated decisions, or services — directly translating into commercial profit.

Record Financial Performance

Huang’s declaration was backed by NVIDIA’s extraordinary financial results for fiscal year 2026, which ended January 25. The company reported revenue of $216 billion, up 65% year-over-year, with operating income reaching $130 billion — a 60% increase. Data center revenue alone surged 68% to $194 billion, underscoring the insatiable demand for AI computing hardware.

NVIDIA generated $103 billion in operating cash flow and returned $41 billion to shareholders. Diluted earnings per share rose 67% to $4.90. International revenue grew more than threefold to exceed $30 billion, with nearly 40 countries now deploying NVIDIA-powered AI factories, according to the company’s shareholder materials.

Huang also committed to returning 50% or more of free cash flow to shareholders this year, next year, and beyond, signaling strong confidence in sustained profitability.

Vera Rubin Enters Mass Production

A centerpiece of Huang’s address was the announcement that NVIDIA’s next-generation Vera Rubin architecture has entered full mass production. The platform represents a fundamental shift in chip design, purpose-built for the emerging AI agent era.

“Hopper was built for pre-training. Blackwell brought inference to rack scale. Vera Rubin is built for agents,” Huang explained, as reported by 财联社 (Cailianshe).

The architecture features two key components: Vera, a custom-designed “agent-oriented CPU” built from scratch for AI agents operating at nanosecond timescales, and Rubin, a “thinking GPU” optimized for large language model inference. Huang emphasized that in an AI factory, idle GPUs mean direct profit loss — making CPU performance critical.

“Vera is the CPU for agents, Rubin is the GPU for thinking — that’s why Vera is important,” Huang said. He predicted Vera could become one of the most important product launches in NVIDIA’s history, noting that orders have already begun arriving.

The Token Economics Revolution

Huang introduced a new framework for understanding AI’s economic impact, describing tokens as the fundamental unit of profit in the AI era. He cited concrete evidence of AI-driven productivity gains: GitHub developers merged 400 million pull requests in 2024, 500 million in 2025, and in the first months of 2026, the rate nearly tripled.

With approximately 30 million software developers worldwide earning roughly $3 trillion annually — supporting $100 trillion in economic activity — Huang argued that AI agents are amplifying human productivity. Agent-assisted output, he said, is approaching $9 trillion, representing a $6 trillion increase.

“Useful AI is profitable,” Huang reiterated. “Every token is a profit unit — that’s exactly why computing demand is so strong. Customers aren’t buying computers; they are building AI factories that generate revenue.”

Navigating Geopolitical Challenges

Huang devoted significant attention to the complex geopolitical landscape surrounding chip exports, particularly regarding China. While the US government has approved H200 chip exports to China under certain conditions — including a 25% revenue share — Huang disclosed that NVIDIA has not yet received any material revenue from these licenses. He also acknowledged uncertainty about whether the Chinese government will ultimately allow these chips to enter the market.

Taking a firm stance on chip smuggling, Huang stated: “National security is above all else. Attempting to piece together computing clusters in restricted markets through ‘smuggled’ chips is a complete ‘dead end.’” He warned that NVIDIA would not provide official support, software updates, or repair services for illegally diverted chips, rendering such efforts unsustainable.

Physical AI: The Next Growth Wave

Looking ahead, Huang identified “Physical AI” as NVIDIA’s next major growth phase — encompassing robots, autonomous vehicles, and factories that can perceive, reason, plan, and act autonomously in the real world. NVIDIA’s full-stack approach includes AI factories for training, Omniverse for simulation, Jetson for edge deployment, and Cosmos as the world foundation model.

A Multi-Decade Infrastructure Cycle

Huang drew historical parallels to describe the scale of the transformation underway. “This buildout will be measured in decades, similar to other critical infrastructure builds like electrical grids, transportation systems, and the internet,” he said. “We believe this will become the largest infrastructure build in human history.”

With NVIDIA’s stock closing at $199.00 on June 24 — a market capitalization of $4.82 trillion — and the company positioning itself as an AI industrial infrastructure provider rather than merely a chip company, Huang’s message was clear: the AI investment thesis has moved beyond speculation into demonstrated profitability.

As AI agents begin to perform real work and generate measurable economic value, the demand for NVIDIA’s infrastructure shows no signs of slowing. The question for investors and competitors alike is no longer whether AI will pay off — but how large the returns will ultimately be.