AI不是泡沫那麼簡單：資本市場正在重新排座次｜AI Is More Than a Bubble: The Market Is Reordering Itself

Looking at several recent signals together, I think it is worth paying attention to.

First, in the S and P five hundred, the technology sector's market value has exceeded thirty-nine percent.

This proportion which is higher than the dot-com bubble in two thousand.

Second, Nvidia and Microsoft have started to cooperate to directly integrate AI capabilities into PCs.

That is to say, AI is not just cloud-based large models, or ChatGPT.

It's not just about opening a web page and asking it questions.

It's starting to enter your computer.

and your local devices.

It enters your daily work scenario.

Third, SoftBank announced to invest in building an AI data center in France.

This is not a small project.

with a maximum capacity of five gigawatts.

with a maximum investment of seventy-five billion euros.

These things don't seem to be on the same line.

One is the structure of the US stock market, One is the AI PC.

and the other is data centers.

But I think behind them is the same thing.

AI is no longer a simple technology issue.

It is becoming the main line and redistribute weights.

Many people are now looking at AI are still asking if Nvidia is expensive, whether semiconductors can be chased, whether AI is overvalued again.

Of course, these questions are important.

But I think the bigger change is not how much a stock has risen, but the capital market is rearranging its priorities.

What did the market value most in the past?

Oil, banking, automobiles, real estate, insurance, consumption These things represented the order of the previous economy.

Oil represents the energy of the industrial age, Banks represent credit creation.

cars represent the crown of manufacturing.

Consumer giants represent channels and brands in the era of globalization.

But now the market is most willing to give high valuations is increasingly concentrated on another asset line.

Chip, computing power, cloud computing, storage, advanced packaging Data center, power, infrastructure This is no longer just a simple rise in tech stocks.

This is the capital market telling everyone the scarcest thing in the future has changed.

Today, I want to break this down into three points.

First, AI is moving from application explosion to a computing power explosion.

Second, AI is moving from a software story to a story of heavy assets.

Third, ordinary people can't all in.

but they can't completely ignore this main line.

The earliest time people saw AI should start with applications.

When ChatGPT comes out many people realized for the first time machines could really write like humans.

answering questions, and making summaries.

Later, AI painting, AI Video AI Office AI Search AI Programming Many people believe that AI is just a software.

It's about whose model is smarter.

whose application is more useful, who can make a hit product first.

But this is only the first layer.

The more applications, the greater the computing power demand behind it.

Every time you ask AI a question, it doesn't answer out of thin air.

It requires model inference behind it.

Model inference requires chips.

Chips require servers, The server requires a data center Data centers need electricity.

So the more prosperous AI applications are, the greater the hardware pressure behind it.

That's why in the past year or so, the market is not only buying AI application companies, but also started to revalue NVIDIA TSMC Hynix Micron Broadcom data centers, and power equipment.

Because people gradually realize that AI is not just run by a beautiful APP.

but it is a set of computing infrastructure.

In the previous Internet era, many companies were light assets.

You make a platform, as users grow, and marginal costs can quickly decrease.

But AI is different.

Every time AI upgrades its model, every time user growth, every increase in inference demand, it requires real computing power.

So as AI goes on, the less it resembles the traditional Internet.

It's not just about traffic.

It's all about computing power.

chips, and storage.

and power.

It's about who can continue to burn money and build infrastructure.

This is also why the capital expenditure of global cloud vendors is increasing.

Microsoft, Google, Amazon Meta including Oracle.

and Chinese tech giants are all investing in AI infrastructure.

You can say there is a bubble in it, but you have to admit that capital expenditure is real.

data centers are being built, and chip orders are also real.

and the demand for electricity is also real.

So I don't think AI is just a short-term topic More accurately, AI is a real trend.

But the market will continue to trade it in bubbles This sentence is very important, because the trend is real, but the price will fluctuate.

The industry is real, but the stock price will not always rise.

The general direction is fine, But for ordinary people, if they chase the hottest time of emotions, they will probably lose money.

The second point I think is more critical.

AI is changing from a software story to a story of heavy assets.

When people hear AI, they still have a few key words in mind.

Large models, algorithms, agents, applications, software But if you follow this chain down, you will find that it is getting harder.

Behind AI applications are models, Behind the model is computing power.

computing power is backed by chips, Behind the chip is advanced manufacturing Behind the data center is electricity Electricity, land, heat dissipation, and capital expenditures You see, the further you go, the less it looks like the Internet, the more like industry.

This is the difference between and the previous mobile Internet, is very different.

In the era of mobile Internet, the most important entrance was mobile phones, The APP is the traffic.

As long as you capture user growth, you can create a super platform.

But the AI era is different.

Of course, AI also needs entrances, But the more fundamental is infrastructure.

Without chips, The model doesn't work.

Without storage, Data throughput can't keep up No advanced packaging The performance of the chip will not go up.

There is no data center.

Large-scale reasoning cannot land No electricity.

everything can't work.

So AI is ostensibly the most cutting-edge technology, Behind it is the oldest constraint Electricity land manufacturing cost That's why I've always said AI cannot be treated as software.

It's not light at all.

The more AI develops, it will bring the capital market back to Hard assets, hard manufacturing, and hard infrastructure Look at SoftBank building an AI data center in France.

This is a typical example.

If AI is just a software application, why does it need to build data centers?

Why does it care about electricity?

why does it care about location?

why consider several gigawatts of capacity?

Because at this stage, AI is no longer just a matter of writing a few lines of code.

to solve the problem.

It has become an infrastructure competition.

who has enough cheap enough stable and clean electricity, who has strong data center construction capabilities, who has stable chip and storage supplies, Only those who have it, to stand firm.

This is also very important for Chinese investors.

Because many people only look at the most dazzling company in AI, such as Nvidia.

Nvidia is certainly very important.

It is one of the most core companies one of the core companies.

But the AI industry chain is not just NVIDIA.

Chip design is one layer.

chip manufacturing is one layer, advanced packaging is another layer.

HBM high bandwidth memory and storage are one layer servers are another layer, data centers are another layer, Power equipment, liquid cooling, optical modules, copper transformers is another layer.

So the really big change is not that one company goes up.

but the entire AI infrastructure chain has been re-priced.

In the past, many people thought that Power equipment, transformers, copper data centers were not sexy enough.

They are not as cool as large models, or as imaginative as robots, or as easy to spread as AI videos.

But capital markets will now see these things again Because AI will eventually encounter hard constraints.

Insufficient computing power is a hard constraint.

Insufficient storage is also a hard constraint.

Not enough electricity is also a hard constraint.

AI hasn't made the world lighter.

To some extent, it has made the world heavier.

It makes us see again those that seem traditional, very basic and bulky things may be the real foundation of technological revolution.

The third point is back to the capital market.

Why should we pay so much attention to this?

Because the capital market is the most sensitive, it will vote for the future in advance.

In the past few decades, who has been highly valued by the capital market.

Energy giants, banking giants, consumer giants automobile giants, internet platforms. The ranking of market value of each generation ultimately reflects the economic order of that era.

Oil companies stood at the forefront because the industrial era could not do without energy.

The bank is standing in front because Credit expansion supports economic growth Consumer companies stand in front because globalization and middle-class consumption are the main themes.

Internet companies stand in front because traffic and data have become core assets.

Now the market is starting to give chips, cloud computing AI platform, higher weight of computing power infrastructure This shows that the capital market believes that the most valuable thing in the next stage may no longer be energy in the traditional sense, channels and credit, It's about computing power, data, energy, advanced manufacturing, infrastructure.

Therefore, this round of AI market cannot be simply understood as a hype of tech stocks.

It's more like the capital market is changing the map.

On the old map, The most expensive are oil, banks, cars, consumption On the new map, The most expensive are chips, computing power, and cloud storage.

power and data centers.

Of course, this does not mean that traditional industries are worthless.

cars still have value, banks also have value.

Energy consumption still has value.

But the market will constantly ask a question.

who will represent the next round of growth, who controls the next round of scarce resources, and who can get the next round of capital expenditure.

The answer to this question is changing.

So you will see that some traditional giants are being surpassed by tech companies.

This matter can't just be seen as stock price news.

It shows that the market's imagination of the future has changed.

People used to believe in industrialization, globalization Real estate, consumption upgrade Now the market is starting to believe AI data center and automation This is the change in asset pricing.

What are the implications for ordinary people?

I think there are three points.

The first is not to treat AI as a short-term theme AI will have bubbles.

and AI will have corrections.

Related stocks will rise and fall sharply.

But we can't just say it's just hype and say it's just hype.

Many real trends are all traded in a bubble by the market.

The Internet era is like this.

and the new energy era is like this.

AI is likely to be the same.

The industrial trend is real.

but the market price will be overheated in stages.

What ordinary people really need to do is not to ask whether AI can be chased every day, but to first distinguish whether you are doing short-term trading or long-term asset allocation.

If you are in short-term trading, you have to endure high volatility.

It rises today and falls tomorrow.

A financial report A policy A valuation switch It can make you feel bad.

But if you are doing long-term allocation, you can't just focus on the three-day price changes.

What you need to see is whether this main line will still exist in three or five years.

My own judgment is that AI is not a short-term theme.

But the price of AI assets will definitely experience overheating and cooling repeatedly.

Both of these statements are true.

Second, don't focus on a single company.

Many ordinary people participating in AI the most common mistake is is to simplify a big industry into a single stock.

When it comes to AI, it's Nvidia.

When it comes to domestic substitution, it's just a few popular stocks.

When it comes to storage, it's the companies that have risen the most recently.

This is very dangerous.

Because the AI industry chain is too long, It cannot be completed by one company.

It has an application layer, a model layer, a chip layer, and a server layer.

There is a storage layer data center layer, power equipment layer, and terminal devices.

For example, AI PCs, AI phones, robots Each layer has different business models and the fluctuations are different.

Some companies grow rapidly but have high valuations Some companies look traditional but have more solid orders.

Some companies benefit significantly in the short term, but the long-term competition is fierce.

Some companies have good stories, But it is still far from true profitability So if ordinary people don't have the ability to research, Don't bet on a single stock easily and don't buy stocks at the peak of a market.

and hear a story, and put a large position in it.

If the trend is right, doesn't mean you buy the right company.

Even if the company is right, nor the price you buy.

Right price doesn't mean you can hold it.

The hardest part of investing is here.

I know many people will ask now, since AI is so important, should ordinary people also invest in AI?

My answer is very clear.

No.

Masayoshi Son Can All in because he is originally a venture capitalist.

Nvidia, Microsoft, SoftBank, these companies can invest heavily.

because They are already at the heart of the industrial chain But ordinary families are not venture capital funds.

The money of ordinary families first need to solve the safety cushion, cash flow, It's pension, education, health care And the uncertainties in life So ordinary people can't learn from the positions of big capitalists.

but ordinary people can learn their sense of direction.

We Can't All in AI However, we can't pretend that AI doesn't exist.

A more realistic approach is Put AI-related assets in the upper equity part of the family's asset allocation.

That is to say, First, there is a bottom layer of safety cushion, Such as cash, monetary funds, short-term debt, emergency funds followed by the middle layer of stable assets, So bonds, dividends, cash flow assets.

Finally, in the equity assets at the upper level for AI technology, Global innovation, computing power infrastructure, these directions reserve a place.

This position can be filled through ETFs, rather than trying to guess which stock will finally win.

In the next few years, the market will repeatedly teach us one thing.

Not everything cheap comes back.

Not all expensive things are bubbles More importantly, whether the asset you buy is on the tail of the old cycle or at the entrance of a new cycle.

This is the biggest reminder that AI gives to ordinary investors.

I'm Su Han.

See you next time.