Huang Renxun outlines the five-layer AI architecture: a multi-trillion-dollar investment roadmap

Jensen Huang has recently come out to “preach” again. He said that all those colorful apps on our phones and the software chugging along on our computers will all be “finished”! The next era is the era of “AI agents.” What is an AI agent? Imagine having a “Micheal” living in your phone — you tell it, “Help me book a high-speed train to Hangzhou for Friday, check if it’s raining by West Lake, and make a reservation at my wife’s favorite restaurant,” and it instantly gets everything done for you. That is an agent.

Huang lays out the paths to profit in the future AI world as five clear levels. Understanding these five levels means understanding the flow of wealth in the AI era.

01

First Level: Energy Layer — AI’s “Power Bank” and “Electricity Bill”

The first level Huang talks about, the foundational level, is energy. You might laugh when you hear it: “After all this, is AI’s end just paying the electricity bill?”

Don’t laugh just yet. This is fundamental. No matter how powerful your graphics card is or how big your data center, without electricity, aren’t they just a pile of high-end scrap metal? Today’s AI is a “power hog.” Training a large model consumes an alarming amount of electricity. Some say the electricity used for one training session of AI could power a household for hundreds of years.

It’s like wanting to drive a sports car; the first thing you check is not the engine but whether the gas station has fuel. So the investment logic at this level is quite simple: it’s a necessity in fierce demand. Whether it’s coal-fired power or environmentally friendly solar and wind energy, as long as the AI vehicle is still running, energy is the most reliable “water seller.”

It’s understandable like this: during the California Gold Rush, the big money was not made by the gold miners but by those selling shovels, water, and jeans. Here, energy is that least noticeable but indispensable “shovel.”

02

Second Level: Chip Layer — AI’s “Strongest Brain”

With electricity, who does the work? That brings us to Huang’s old trade — the chip layer.

If energy is “food,” then chips are AI’s “digestive system.” Why are GPUs (graphics processing units) so hot? Because they’re “specialized geniuses” that don’t do much else but focus on performing countless calculations, which is exactly what AI needs.

Look at it now; whoever has a high-performance chip walks with a sense of pride. This thing is now harder than “hard currency”; when Nvidia’s financial report comes out, the whole world trembles. Why? Because it is the “heart” of the AI world, and it’s the only one that can be exchanged.

The investment logic at this level is also straightforward: the highest technical barriers, winner takes all. Whoever masters the most advanced chip technology has the entire AI industry’s throat in their grasp. This is a king’s game at the top of the pyramid; ordinary players can only watch the excitement but must know who the real “decision-maker” is.

03

Third Level: Infrastructure Layer — AI’s “Super Factory”

Having a heart isn’t enough; you need to build it a house, with a climate-controlled environment, and have security guards on duty 24/7. This is the third level, the infrastructure layer.

In simple terms, it refers to those massive data centers, densely packed servers, and the buzzing cooling systems. Just think, with millions of chips working together, how frightening must the heat output be? It’s said that one of the biggest costs for data centers nowadays isn’t purchasing equipment but paying for air conditioning! So now even companies specializing in liquid cooling are taking off.

This layer is like real estate; it serves as AI’s “factories” and “offices.” Its investment logic is: huge economies of scale, vast growth potential. As AI applications proliferate, the demand for data centers is practically limitless. Those in the infrastructure game, get ready for orders.

04

Model Layer — AI’s “Soul Artist”

With the factory and equipment, it’s time to inject AI with a soul. This is the fourth level, the model layer.

You can think of large models as AI’s “brain” or “soul.” For example, the large model for autonomous driving needs to learn how to read the road, how to interpret traffic lights, and how to dodge suddenly appearing obstacles. It’s like a genius child that requires a massive amount of data for nourishment and education before it can stand on its own.

This level is where technical competition is most intense. Today OpenAI releases GPT-4, tomorrow Google might launch Gemini Ultra, and it’s a fierce battle. Why? Because whoever controls the smartest, most understanding “brain” will hold absolute leverage in the next layer of applications.

Investing at this level is like “angel investing,” betting on future geniuses. The potential is huge, but so are the risks because you never know which “prodigy” will ultimately succeed.

05

Fifth Level: Application Layer — AI’s “Money Printer”

Finally, we arrive at the most exciting top level — the application layer. This is also where Huang says the real “monetization” of AI happens.

The first four layers are all about building behind the scenes. This layer is where the average person can truly interact with and use these technologies. For instance, an autonomous vehicle, where you sit back comfortably and let it take you to your destination, then pay via a QR code. Or that assistant that helps you book tickets, which you enjoy so much that you’re willing to pay a monthly membership fee.

This is what Huang calls “AI agents.” They are no longer individual software programs that you need to learn and operate; they have transformed into services that understand you and help you work.

The investment logic at this layer is the most straightforward: the fastest to implement, the strongest monetization capability. Whoever can create an AI application that makes users go “WOW” can directly turn technology into cash. Just like in the early days of the internet, the ones who made the big money were Amazon, Taobao, and Tencent, as they effectively applied technology to everyday scenarios like shopping and chatting.

So, understanding Huang’s five levels gives you insight into where the money flows in the AI era.

1. The bottom layer (energy, chips) is the ballast, stable, suitable for long-term holding;

2. The middle layer (infrastructure, large models) is a rocket, fast, with huge growth potential;

3. The top layer (applications) is the money printer, precise, directly facing users, with the strongest monetization capability.

Previously, AI felt distant, like something out of a sci-fi movie. Now, AI feels close; it’s becoming that all-powerful “agent” right next to you, reshaping every software and service.

As for ordinary people? Either work hard to become a brick in one of these five levels; or be ready with your wallet to enjoy the unprecedented delightful services brought by AI agents. After all, even laziness may become a virtue in the future.