The belief in AI has turned into FOMO, and the patience of capital has started to be counted.

One thing to say in advance:

This article is not easy to read and is even a bit "mind - boggling". But it doesn't matter. As you read on, you'll find that I'm talking about the capital, human nature, and cycles behind AI.

Right after the National Day, there are new developments in the global capital market.

This time, it's on a larger capital level. On October 15th, an acquisition worth up to $40 billion was finalized: several top - tier US investment institutions jointly bought one of the world's largest data centers.

The day before, Google also announced in India that it would invest $15 billion to build a new AI hub.

It may seem far from us, but these two things are quite interesting: While everyone is still talking about models and agents, foreign capital has quietly started "laying the bricks"; they are no longer hyping up algorithms and are less concerned about the application layer. Instead, they are collectively rushing towards one thing: the foundation.

So, I've been thinking these days: Why are all people looking up to the sky while these companies are starting to invest in the ground?

01

To understand this wave of AI fever, we need to first see where the money is going.

I looked through a bunch of data, and it's quite astonishing.

As of June 2025, the annualized construction expenditure on data centers in the US has reached $40 billion, about 30% more than last year. This is a statistic by the US Bank Research Institute based on Commerce Department data, and Reuters has reported with the same caliber.

It's even more outrageous on a global scale.

McKinsey estimates that by 2030, the world will have to invest nearly $6.7 trillion in data centers to meet the computing power demand. Among them, the part for AI workloads will account for more than $50 trillion.

Even the IMF has pointed out in a report that the US has withstood the downward pressure this year thanks to this wave of AI investment fever.

What does it mean?

Everyone is talking about "models" on the lips, but the money is all being poured into physical foundations such as electricity, land, and "cooling systems".

I think this is a "return of gravity". In the past decade, the Internet has made capital accustomed to the light - logic model - fast replication, low cost, and high marginal returns. But in the AI era, the logic has completely reversed.

The smarter the AI, the heavier the computing power behind it; the larger the model, the higher the energy consumption. So, those who used to play with algorithms are now buying power plants, building computer rooms, and negotiating land rents.

This is the most overlooked turning point in the AI industry. Its essence is changing from an "intellectual game" to a "physical war".

To run a large model with 10 trillion parameters, you need not only GPUs but also electricity, cooling systems, and a stable network. Just like when humans first generated electricity and built airplanes, AI has officially entered its "heavy - industry era".

So, AI is a business that requires "steel, land, and hydropower".

Currently, several of the world's most profitable companies have almost all become computing power providers. NVIDIA is making huge profits from chips, and Microsoft is desperately expanding its data centers. What is this? A new round of "digital gold rush". Only this time, it's the capital that is really "mining".

I think there is a deeper - level logic behind this trend: AI is "re - centralizing".

In the future, each data center will be a new "urban node", and each infrastructure investment will be a new "energy dispatch". When data, computing power, electricity, and land are tied together, the global resource chessboard is reshuffled.

So, this is the so - called "physical turning point of AI": In the past two years, AI has been telling stories with language models. In the next few years, it will rely on power grids and land to fulfill its future.

However, when all things start to materialize, new problems also arise. For example, heavier capital means a longer capital cycle; the more you invest, the slower the return on investment. So, how long will it take to get the money back?

02

Investment in AI infrastructure has a long cycle, heavy depreciation, and high costs. What it fears most is "slow realization". Unfortunately, this track is just slow.

In the past two years, investors' emotions have been like a roller - coaster.

In 2023, there was excitement. ChatGPT brought a surge of faith, and everyone rushed in. In 2024, there was expectation. Everyone was waiting for when AI would turn into profits. By 2025, investors started to get anxious because they couldn't figure out the accounts.

I asked an agent to run several reports. The numbers are quite eye - catching:

The price - to - earnings ratio of the US AI sector has dropped from 58 times in 2023 to about 35 times this year. Almost all large - model companies are losing money, and the cost of AI cloud services is still rising.

This is a signal that the market has officially entered the "report cycle". In the past, capital was fed by stories. Now, it depends on tables: revenue curves, cash flows, and return periods. So, patience has become a luxury.

I have several friends in investment research. At the beginning of the year, they were full of confidence. Now, they are starting to calculate. They said that the "realization logic" of AI is too vague:

Although technology is advancing, productivity has not significantly improved. The larger the model, the more expensive the computing power. Enterprises buy AI services but don't know what they are really buying. The result is that capital is in a very delicate state, unable to withdraw and afraid to increase investment.

I think this is a kind of "cognitive fatigue". AI is too fast in narrative and too slow in realization. In 2023, it was a narrative bubble. In 2025, it has become a cash - flow dilemma.

The market has also gradually realized that this revolution is neither that fast nor that cheap.

Psychologically, this is a problem of the "price of patience". Interest rates haven't dropped, and inflation is high. Waiting itself has a cost. Every quarter is an agony for investors because capital also accrues interest over time.

The real contradiction of AI lies in the rhythm. Technology is exponential, while business realization is linear. When these two lines are put together, the gap is getting larger and larger, and market sentiment naturally starts to fluctuate. So, now investors are most worried about "the future being too slow".

This is why many recent AI investment reports are writing about "differentiation".

Because large - model manufacturers are still burning money, small - model companies are actually doing quite well. Hardware manufacturers' profits are soaring, while software growth is slowing down. The wealth effect of AI is starting to reverse.

In this rhythm, investors' patience is like a rubber band, getting tighter and tighter and will eventually break. When capital starts to get impatient, the reaction in the real world also starts to distort.

Masayoshi Son also made a move this year.

In March 2025, SoftBank bought the former Sharp LCD factory in Osaka, Japan, and plans to transform it into an AI data center. Three months later, he threw out an even bigger plan in Arizona: to build an "AI infrastructure park" worth $1 trillion.

This is one of his few public statements. This old gambler who bet on Alibaba and Arm is starting a new round of all - in.

I think this kind of abnormal courage is a kind of "counter - cyclical patience". While others are hesitating, he is increasing his investment. While others are worried about the return on investment, he is betting on the long - term. This is the most real footnote to this wave: Technology is rising, while capital is diving.

03

You know, the biggest problem with AI now is not the algorithm but the energy.

I saw a set of data - it's quite scary.

The annual power consumption of an ultra - large - scale data center in the US is equivalent to the total power consumption of a city with a population of 100,000. Another report says that a cloud - computing giant uses more than 1 billion liters of water a year to cool its servers.

What's the concept? In a drought - prone area, it's almost like draining a small river.

The appetite of AI is being magnified globally. For every bit of algorithm evolution, the energy consumption behind it has to increase. From electricity to cooling systems to land supply, it has become a "big - appetite revolution" in the real world.

This is the other side of the AI fever that is most easily overlooked. It swallows not only computing power resources but also the public resources of society.

I checked. In Arizona, local residents protested, saying that the data center "drank up all the city's water". In Mexico, some people held signs in opposition because the computer rooms caused power outages for them.

In India, as soon as Google's new base was approved, local media were asking: Is there enough electricity? Where will the water come from?

These news stories may seem trivial, but they actually show one thing: Technology is hitting the boundary of land. I think this is the real "implementation of AI"; it has crashed into reality.

From a geopolitical perspective, this "energy war" has started to spill over. The US is scrambling for electricity. Middle - Eastern countries are building "data deserts". India, Indonesia, and Vietnam are competing for the "right to implement computing power". In the past decade, we were competing for "data sovereignty". Now, we are competing for "computing power sovereignty".

Whoever controls energy can feed the algorithm. Technology needs speed, and society needs order, and energy happens to be stuck between the two.

I think this infrastructure revolution is, after all, a game between capital and nature. From coal and oil to computing power and electricity - every technological revolution is, in essence, a "re - distribution of energy". It's just that this time, we've given it a new name, AI.

But revolutions never only belong to the winners. When energy is in short supply, communities rebel, and the environment is damaged, the light of technology will also cast a shadow.

So, now I want to ask:

When resources are scarce and the environment rebounds, will humans continue to bet on AI? I think so. Because what we are betting on is never just technology but more like a cycle of "faith".

04

The steam engine, electricity, and the Internet - each has been regarded as a "salvation". When humans are anxious, they always need something to believe in. It can be God or code.

AI is the latest form.

It makes us feel that the complex world can finally be calculated, chaotic emotions can finally be understood, and even life choices can be "handed over to the algorithm". The more powerful the technology, the more people want to get close to it. That power is very much like a sense of certainty.

But the problem is: No matter how perfect the technology is, it can't live our lives for us.

I've seen too many such cycles.

When the wave of enthusiasm comes, people talk about the future. After the wave recedes, people talk about disillusionment. But as long as the next wave of technology emerges, people will believe again: This time is different. So, there is no right or wrong about bubbles; they just advance the fulfillment of our desires.

You see, the railway bubble left railways, the Internet bubble left networks, and the AI bubble will also leave computing power, models, data, and infrastructure.

This is human nature: We use hope to drive the world forward and then use disillusionment to correct the direction. Faith is born out of anxiety, bubbles stem from faith, and value always appears after the bubbles.

So, when the bubbles fade away, do we still remember why we believed in the first place?

AI has brought efficiency and also illusions. It has made humans face a more "intelligent" existence for the first time - but intelligence doesn't equal understanding. When machines start to answer everything, we may be more likely to forget to ask "why".

I think we are now in a delicate stage: The more powerful AI becomes, the weaker humans become because we start to outsource choices, entrust thinking, and transfer trust.

But if you think about it carefully, the capabilities of AI actually come from humans. It is humans who train it and teach it to think. AI is just a mirror that reflects our own imagination of the future and also our fears.

So, the core of this wave of enthusiasm is "what people still believe in".

Many people believe in technology and also in cycles. However, every time a new trend emerges, I want to see who can stay. Because in the end, it's not the models or algorithms that can fulfill the future but humans.

Perhaps, looking back ten years later:

New urban nodes, energy networks, and algorithm languages are all traces of the progress of civilization. However, the narrative of AI infrastructure has not yet reached China.

And foreign investors have already started to fall into another cycle: from expectation to anxiety and then to doubt. After all, isn't this just another form of FOMO?

This article is from the WeChat public account "Wang Zhiyuan" (ID: Z201440). Author: Wang Zhiyuan. Published by 36Kr with authorization.