What to Watch at Nvidia’s 2026 Shareholder Meeting? Jensen Huang: Every Token Is Profit, AI Monetization Already Has an Answer

Jensen Huang emphasized repeatedly in his speech that AI has completed the transition from technical experimentation to commercial production.

Using GitHub developer data as an example, he noted that the number of pull requests merged by global developers was 400 million in 2024, increased to 500 million in 2025, and grew nearly threefold in just the first few months of 2026. Behind this data is the reality that AI agents are replacing humans on a large scale to complete programming, design, and data analysis tasks, with every generated token becoming a monetizable unit of profit.

Huang proposed that in the past, the core function of data centers was to store and transmit files, whereas today, the core mission of new AI factories is to produce tokens—these monetizable units of intelligence that serve as the raw materials for code, answers, designs, actions, and services.

He described the AI industry ecosystem using a "five-layer cake" structure, consisting of energy, chips and systems, infrastructure, models, and applications from bottom to top. This framework implies that Nvidia's business scope extends far beyond chips themselves, spanning the entire AI production chain.

Under this new logic, customers purchasing Nvidia systems are not procuring computing tools, but are instead building AI factories capable of directly generating revenue.

Huang emphasized that the efficiency of the factory architecture—namely, how many tokens can be produced per watt and how low the cost per token is—has become the most critical dimension of competition.

Nvidia's full-year revenue for fiscal year 2026 reached $216 billion, up 65% year-on-year, with data center business revenue at $194 billion, up 68% year-on-year; operating cash flow reached $103 billion, with $41 billion returned to shareholders over the year.

Huang specifically pointed out that although the purchase price of Nvidia systems may not be the lowest, Nvidia can produce the most tokens at the lowest cost and achieve the highest throughput. This business model has been fully validated by the market, laying a solid foundation for Nvidia's continued leadership in the AI era.

With the rapid adoption of large model training, inference, and AI Agent applications, the strategic value of memory and storage in data centers is continuously rising, and Nvidia, with its full-stack technology advantages, is becoming the core driver of this AI industrial revolution.

Jensen Huang positioned the Vera Rubin platform as 'one of the most important products in the company's history.' Unlike the previous Hopper platform focused on training and the Blackwell platform designed for inference, Vera Rubin is a complete AI factory solution tailored for the era of AI agents, with the Vera CPU representing Nvidia's milestone first foray into the general-purpose CPU market.

Huang explained that AI agents operate in a completely different way from humans; they live in a nanosecond-level computing world, frequently calling tools, accessing databases, executing code, and iterating on tasks. In this scenario, if the CPU becomes a bottleneck, expensive GPUs will sit idle, with every second of idleness representing lost revenue for the AI factory. To address this, Nvidia built the Vera CPU from scratch specifically for agents—no longer marketing itself on core counts, but instead pursuing ultra-low latency response to meet the concurrent demands of billions of agents worldwide.

As the first CPU to feature LPDDR5 memory, Vera is 1.8 times faster than traditional x86 CPUs, delivers a 50% boost in single-core performance, and directly supports FP8 precision to handle AI inference and reinforcement learning tasks without requiring GPU transit.

Currently, Vera Rubin has entered full mass production, and major global model developers, public clouds, AI clouds, and hyperscalers have already begun deployment.

Jensen Huang stated: 'Vera Rubin is not a single chip, but an AI factory platform, and the ecosystem is already in motion. Every major model developer, public cloud, AI cloud, and hyperscaler is preparing to build on it.'

The division of labor for Nvidia's third-generation AI products is clear. Hopper focuses on pre-training; Blackwell scales inference to the rack level; and the Vera Rubin platform, consisting of the Vera CPU for scheduling and the Rubin GPU for computing, is designed specifically for the era of AI agents.

Notably, Nvidia is the only vendor in the industry to simultaneously possess three high-speed networking portfolios—NVLink, Spectrum-X Ethernet, and InfiniBand—providing a unique interconnect foundation for the Vera CPU to ensure low-latency response and highly efficient collaboration in agentic AI scenarios.

In addition to digital-world agents, Jensen Huang has also turned his attention to real-world physical AI. He pointed out that artificial intelligence is moving beyond the virtual world to be fully deployed in end-user devices such as autonomous driving, humanoid robots, and industrial smart equipment, equipping physical hardware with complete intelligent capabilities for perception, reasoning, planning, and action. The large-scale deployment in this field will catalyze a new, trillion-dollar wave of infrastructure investment.

Jensen Huang compared AI infrastructure to major landmark infrastructure projects in human history, such as power grids and the internet, believing that this will be a construction boom lasting for decades. He emphasized that Nvidia's technology will not only support digital-world agents but will also provide core computing power support for physical AI, which will become the company's next major growth engine.

Faced with his company's stock underperforming the broader US market so far this year, Jensen Huang sought to stabilize market confidence by reinforcing the view that the AI capital expenditure supercycle remains in its early stages.

He reiterated plans to return more than 50% of free cash flow to shareholders and declared several capital return policies as long-term commitments. During its most recent earnings call, Nvidia announced an authorized $80 billion share buyback program.

Looking ahead, Huang expressed full confidence: "AI infrastructure is no longer experimental; it has entered the production stage." He predicted that computing demand in the era of AI agents will grow exponentially, and Nvidia, with its full-stack technological advantages and ecosystem moat, will continue to lead what he calls the largest infrastructure buildout in human history.

This content was translated using AI and reviewed for clarity. It is for informational purposes only.