If you doubt the monetization capability of AI, then please try Codex

Today, AI is far more than just a "chatbot" in many people's minds.

In the past, when people mentioned AI, their first reaction might still be: chatting with people, writing articles, summarizing information, doing translations, and then making money through subscription models.

These functions certainly have value, but to be honest, customer stickiness may not be strong enough.

Because for many people, these types of functions are more like "icing on the cake": having it is more convenient, but life seems possible without it.

However, the emergence of Coding Agents has completely broken this simple scenario.

It doesn't just answer questions, nor does it just give you a suggestion; it can directly help you complete specific tasks: writing code, modifying code, fixing bugs, creating web pages, making small tools, and even iterating based on your feedback.

This is a qualitative change.

In the past, a person who knew nothing about programming, even with many ideas in mind, found it very difficult to actually create something.

But now the path has changed:

You propose a requirement, AI generates code;

You try the product, AI continues to modify it;

You find an issue, AI helps you fix it;

You want to add a feature, AI continues to iterate.

This has a very strong impact on ordinary people.

Because it's not giving you "emotional value," but rather "productivity."

It's not about making you feel entertained, but about truly letting you spend less time, less money, and ask for less help, turning things you couldn't do before into reality.

This is also why I think AI monetization has already proven viable in some high-value scenarios.

Many people doubt AI commercialization because they are still stuck in the stage of "Can AI write articles?" "Can AI chat?"

But after truly experiencing Coding Agents, you'll find that AI's value has already started shifting from content generation to task execution.

And task execution is a much harder business model.

For individuals, it lowers the barrier to creating tools.

For programmers, it improves development efficiency.

For enterprises, it has the potential to reduce costs in R&D, testing, maintenance, and internal automation.

This kind of value is easy to verify:

Whether the code runs, whether the tool works, whether efficiency has improved—it's immediately apparent.

Therefore, AI programming is not an ordinary feature, but a very clear monetization scenario.

More importantly, once this logic holds, what follows is not "too little AI demand," but "explosive AI demand"; not "too much computing power," but "computing power shortage."

Because a Coding Agent doesn't just generate a piece of text once and end.

It needs to understand requirements, read code, modify files, run tests, reason repeatedly, and iterate continuously.

Every task execution consumes model capabilities and computing power in the background.

This is also why I increasingly feel that the competition among AI companies in the future will, to a large extent, not only be about model capabilities, but also about computing power reserves, engineering systems, and ecosystem access.

For example, the competition between OpenAI and Anthropic.

Looking at it in stages, Claude Code's reputation is indeed strong, with many developers believing it offers a great experience in code understanding and engineering collaboration.

However, OpenAI's early accumulation of a massive user base, product matrix, and computing power resources may also become a very important advantage in the long-term competition.

Short-term product experience can lead back and forth, and model capabilities will also chase each other.

But if AI applications truly enter the stage of high-frequency task execution, whoever has stronger computing power supply, lower marginal costs, and a larger user base will have a better chance of succeeding.

This is also why I don't think the OpenAI-related ecosystem should be simply bearish now.

Currently, some OpenAI-related companies, such as Microsoft and Oracle, have not shown particularly exciting stock price performance in stages, and the market is still watching.

On one hand, the market worries about excessive capital expenditure;

On the other hand, the market worries that AI monetization isn't fast enough.

But if scenarios like AI programming, enterprise Copilot, cloud inference, and Agent workflows continue to be validated, the market will eventually re-understand the significance of these investments.

In the past, what everyone worried about was:

You're spending so much money on computing power, can you actually earn it back?

But if AI applications truly start moving from "chatting" to "working," this question will become:

Do you have enough computing power?

Do you have enough data centers?

Do you have enough chip supply?

Can you lower your inference costs?

At that point, computing power is no longer just a cost item, but the infrastructure of the AI era.

Just like electricity for industry, railways for trade, and cloud computing for the internet.

If AI continues to develop, computing power is an unavoidable underlying resource.

Of course, it's too early to conclude who will ultimately win between OpenAI and Anthropic.

Who among Microsoft, Google, Amazon, Oracle, and Meta has the highest return on investment also needs continuous observation.

But one thing I think is relatively certain:

As long as this AI arms race continues, as long as models keep upgrading, as long as applications keep expanding, as long as enterprises and individuals continue to treat AI as a productivity tool, then the happiest one might still be the shovel-seller $NVIDIA(NVDA.US).

The application layer can compete, model companies can take turns leading, and cloud providers can have their own wins.

But what they all need in common is computing power.

So, if you still doubt AI's monetization ability, you should really try using Codex.

When you see an idea turn into a working small tool within minutes with your own hands, you'll intuitively understand:

AI's commercialization is not just about storytelling.

It has already started making money by increasing productivity.

And as long as AI starts making money by increasing productivity, computing power will become a long-term necessity.

In this chain, the most certain beneficiary is still most likely the shovel-seller $NVIDIA(NVDA.US).