Summary of Key Points
At NVIDIA's 2026 annual shareholders' meeting, all 10 board members were approved by the shareholders. CEO Jensen Huang emphasized that the "era of useful AI" has arrived—AI has moved from the experimental phase to the production stage, and the demand for computing power will continue for decades, involving the construction of large-scale infrastructure such as power grids and the internet. NVIDIA's product portfolio (Hopper/Blackwell/Vera Rubin) covers the entire process of AI training, inference, and agents. Vera Rubin has been put into production and has already received orders from major companies. Physical AI (agents in the real world, such as robots and autonomous vehicles) is seen as the next growth engine. NVIDIA also announced a significant increase in dividends and a $80 billion stock repurchase plan, committing to returning more than 50% of its free cash flow to shareholders. The company's stock price fell slightly by 0.56% on the day but rose by 0.82% after the market closed.
Detailed Analysis
1. "The Era of Useful AI": AI Finally Becomes a Profit-Making Tool
By "useful AI," Jensen Huang means AI that can help businesses make money. Previously, AI might have been just a toy in laboratories, but now it's different:
- 30 million software developers worldwide support an economy worth trillions of dollars, and AI is making them more efficient and creating greater value.
- Various industries are racing to use "agent-based AI" (AI that can complete tasks independently without human instruction). For example, companies use AI to automatically handle customer inquiries and write reports, which directly reduces costs or increases revenue.
- Investing in AI is no longer a bottomless pit; returns can now be calculated clearly. For instance, knowing how much profit can be generated from investing in computing power (by processing "tokens," which will be explained later).
2. AI Infrastructure Enters the "Production Stage": Building AI Like Factories
Jensen Huang compared the AI industry to a five-layer cake, with each layer being essential:
- Base Layer (Energy): Electricity is fundamental. Without a stable power grid, AI computing power is meaningless.
- Second Layer (Chips and Systems): NVIDIA's GPUs are like the "flour" used in baking the cake.
- Third Layer (Infrastructure): Traditional data centers for storing data have evolved into "AI factories" that produce tokens (the smallest units processed by AI, such as words or characters in a sentence).
- Fourth Layer (Models): Large models like GPT and Wenxin Yiyuan are the "recipes" for the cake.
- Fifth Layer (Applications: Tools like ChatGPT and AI-powered drawing tools are the "finished cakes" used by users.
Why is there such high demand for computing power? Because each token represents a unit of profit. The more tokens an AI factory produces, the more money companies can earn. NVIDIA's advantage lies in its ability to produce tokens at the lowest cost and fastest speed (using the same amount of electricity to process more tokens), making it the preferred choice for customers.
3. NVIDIA's Product Portfolio: Covering the Entire AI Process
Jensen Huang revealed the positioning of three core products that cover almost every aspect of AI development:
- Hopper: Used specifically for training large models (teaching AI knowledge, such as having AI read all the books on the internet).
- Blackwell: Responsible for inference (AI answering questions and making decisions). The amount of computing required for inference has now surpassed that of training, and Blackwell can process more tokens at a lower cost, giving NVIDIA a leading position in this area.
- Vera Rubin: Used for agents (such as robots and autonomous vehicles) and has been fully deployed. All major model companies and cloud providers are preparing to use it to build their infrastructure. Vera CPU is also considered one of the most important products in the company's history and has already started receiving orders.
These products have transformed NVIDIA from a chip manufacturer into a provider of comprehensive AI solutions, making it difficult for competitors to replace.
4. Physical AI: The Next Profit-Generating Area
Jensen Huang called physical AI the next growth opportunity. Physical AI refers to AI that can act in the real world, such as:
- Robots working in factories, perceiving their environment and avoiding obstacles.
- Autonomous vehicles on the road, recognizing traffic lights and avoiding pedestrians.
- Smart factories that can automatically adjust production lines.
All these applications require AI to respond in real-time to dynamic environments, so new infrastructure (such as more advanced sensors and computing networks) is needed. NVIDIA aims to capitalize on this opportunity to expand AI from the digital world to the physical one.
5. Shareholder Returns: Money Back for Investors
How will NVIDIA distribute its profits to shareholders? Jensen Huang outlined the following plans:
- Increase Dividends: Distributing more cash directly to shareholders.
- Expand Stock Repurchases: Recently authorized $80 billion in stock repurchases (fewer shares on the market mean higher potential stock prices).
- Long-Term Commitment: Returning more than 50% of free cash flow to shareholders in the future.
This shows that NVIDIA is not only profitable now but also willing to share its profits with investors, which is a positive signal for them.
Final Note: Why the Stock Price Fell Slightly on the Day?
Despite all the good news, the stock price fell by 0.56% on the day but rose by 0.82% after the market closed. This may be due to concerns about AI becoming too overhyped in the short term, but the subsequent increase indicates that investors still recognize NVIDIA's long-term potential. Overall, the shareholders' meeting sent a strong signal that NVIDIA's advantages in the AI field will continue, and there are many opportunities for growth in the coming decades.