Instead of chasing trends within the AI supply chain yourself, it's better to directly hold the leading giant in AI infrastructure—NVIDIA.

Yesterday, NVIDIA announced a partnership with Corning, a leader in specialty glass and fiber optic connectivity products, to drive the expansion of fiber optic and optical connectivity capacity needed for U.S. AI data centers.

After the news came out, Corning's stock price surged by 12%.

This is very interesting.

Because it reminds me once again: the AI industry chain has never been just about GPUs. Truly large-scale AI infrastructure construction requires a complete set of systems behind it, including high-speed networks, optical communications, storage, power, liquid cooling, servers, software scheduling, and more.

But the question also arises:

As ordinary investors, do we really need to chase every link in the AI chain?

A while ago, NVIDIA invested in Intel at $23.28 per share, with a total of $5 billion. Based on a rough calculation of Intel's current stock price, the market value of this investment is already close to 4.9 times the cost, with a very substantial paper profit.

Looking further back, NVIDIA has long ceased to be a company that simply sells GPUs.

It acquired Mellanox back then, gaining high-speed networking and data center interconnect capabilities;

Later, it acquired Cumulus Networks, strengthening its data center network software capabilities;

Then, it acquired Run:ai, further enhancing its AI compute scheduling and infrastructure management capabilities.

A while ago, NVIDIA also reached an inference technology licensing agreement with Groq and absorbed core technical personnel like Groq founder Jonathan Ross. Groq itself focuses on AI inference chips, and Jonathan Ross was involved in the Google TPU project earlier in his career.

This move also speaks volumes:

When the market started shifting its focus from "training compute" to "inference compute," NVIDIA didn't stand still waiting for others to challenge it; instead, it actively worked to fill its capabilities on the inference side.

Looking at these things together, a very clear trend emerges:

NVIDIA is not just selling chips; it is continuously making systematic layouts around AI infrastructure.

What it buys, invests in, and partners with essentially answers one question:

What will future AI data centers still lack?

Previously, the market thought AI mainly lacked GPUs, so NVIDIA benefited the most.

Later, the market discovered that AI also lacks high-speed networks, optical communications, storage, power, liquid cooling, compute scheduling, and complete data center solutions.

But the key is, if these links are truly important enough, NVIDIA is unlikely to ignore them.

It will either actively participate, invest to secure ties, or incorporate these links into its ecosystem through partnerships.

So, I'm increasingly thinking:

Investing in NVIDIA, on one hand, captures the most critical link in the AI era's compute infrastructure;

On the other hand, it also delegates part of the judgment to the company that understands the industry best and has the strongest resource allocation capabilities within the AI infrastructure chain.

This point is very important.

The biggest problem for ordinary investors chasing AI sidelines is not a lack of logic, but too much logic.

Today storage is up, feeling storage can't be missed;

Tomorrow optical modules are up, thinking optical modules are the core;

The day after, power, liquid cooling, and servers start telling stories again, wondering if their portfolio is incomplete.

Chasing around like this can easily mess up one's core positions and investment rhythm.

But NVIDIA is different.

It is itself a core node in the AI infrastructure chain. It controls GPUs, the CUDA ecosystem, high-speed interconnects, complete system solutions, and resources for collaborating with cloud providers, large model companies, and the upstream and downstream of the industry chain.

More importantly, it has a clearer understanding than most ordinary investors of what the next step in AI infrastructure truly lacks.

Therefore, rather than chasing around the AI chain myself, it's better to directly hold the core company in the AI infrastructure chain.

This isn't to say that sideline companies have no opportunities.

In directions like storage, optical communications, power, and liquid cooling, if one gets the timing right, the short-term elasticity might be even greater than NVIDIA's.

But high elasticity does not equal high certainty.

For ordinary investors, the real difficulty is not finding a hot spot, but not losing sight of the main line they truly understand when hot spots keep changing.

My thinking is simple:

Hold core AI infrastructure companies like NVIDIA in the main portfolio;

Continue to hold platform giants like Microsoft and Google as core allocations for AI applications and cloud computing;

Small positions can be used to observe sideline directions, but they must not overshadow the main focus.

The AI industry chain is very long. I can't buy every link, nor is it necessary.

If the AI era continues to develop, leading companies will naturally lay out the truly valuable links.

What I need to do is not chase the entire industry chain, but hold the core nodes within the chain that have the most pricing power, ecosystem control, and resource allocation capabilities.

This is also my increasingly firm belief:

Rather than chasing around the AI chain myself,

It's better to directly hold the leading big brother in the AI infrastructure chain—NVIDIA.

Personal thoughts are not investment advice.