The AI bubble is bursting

Original Title: The AI Bubble is Already Bursting

Original Author: Chengbei Xugong, Gelong

In recent days, the market has been highly volatile, and the "AI bubble theory" is rampant.

Ray Dalio, founder of Bridgewater Associates, said: There is a bubble in the AI market, and the level is "relatively high."

NVIDIA CEO Jensen Huang said: There are huge opportunities in AI, and the demand for computing power has just begun to explode.

Who should we believe?

Both of them are correct.

Is there a bubble in the AI industry? There inevitably is.

However, bubbles in the tech sector are often the only way society can pay tribute when faced with disruptive advanced productive forces.

It is not merely a pejorative term.

In the long run, this is a phenomenon that will inevitably occur at the dawn of advanced productive forces.

Many people are comparing the current situation to the internet bubble of 2000, filled with anxiety.

The internet bubble of that year did indeed lead to a nearly 78% crash in the Nasdaq, with over $5 trillion in wealth evaporated.

But twenty years later, which industry can do without the internet?

Today, the value of the internet industry has long surpassed that of the bubble period.

The AI bubble, at least on the surface, appears to be a similar situation.

The bubbles existing in the capital market cannot stop almost all industries in society from actively being empowered by AI.

AI+ is an unstoppable trend.

Just as all industries cannot do without the internet today, all industries in the future will also not be able to do without AI.

01

In the era when a company could go public just by having a .com in its name, the Nasdaq surged nearly 600% from 1995 to 2000. Then came a financial storm that lasted two and a half years.

Back then, those well-known names, like the software company MicroStrategy, plummeted 62% in a single day due to accounting scandals and overblown claims; Pets.com (selling dog food online) and Webvan (the pioneer of fresh e-commerce) went bankrupt on the spot.

......

In the panic, almost everyone blamed the internet as a scam.

However, the physical infrastructure left behind by the excessive squandering of speculative capital often nurtures the next era's super giants at extremely low costs.

The reason for the bubble's burst is not the problem of internet technology itself, but rather that the pace of physical infrastructure construction cannot keep up with the market's rhythm.

For example, those telecom companies that were once thriving (like WorldCom and Global Crossing) invested heavily in laying global submarine cables and dense wavelength division multiplexing networks. Although this led to their own bankruptcies, these cheap "information highways" became the perfect breeding ground for the later rise of Netflix, Zoom, and mobile internet.

Without the frenzied pre-investment in telecom infrastructure around 2000, there would be no subsequent explosion of video streaming on YouTube, nor would there be the later cloud computing infrastructure.

The most typical example is Amazon.

Its stock price fell from a peak of $107 in 1999 to just $7 in 2001, a drop of over 90%.

But it survived because its underlying business logic, "restructuring retail through the internet," aligned with the direction of advanced productive forces.

This is the classic Amara's Law: to overestimate the short-term impact of a new technology while severely underestimating its long-term impact.

In the early stages of a technological revolution, the frenzy of speculative capital inevitably leads to over-investment, creating a bubble.

This is the intelligence tax that innovation must pay.

But when the bubble dissipates, what remains will be an even more indestructible advanced productive force.

02

Returning to 2026, the AI industry's bubble appears even larger.

Just the five major cloud service providers—Amazon, Google, Meta, Microsoft, and Oracle—are expected to reach $690 billion in capital expenditures by 2026, with total AI infrastructure investment expected to reach $5.3 trillion by 2030.

Of this, only about 25% is spent on GPUs, while the remaining 75% is all invested in physical infrastructure: liquid cooling systems, power transmission, network switches, optical modules, and land.

In terms of revenue, all leading pure AI companies like OpenAI, Anthropic, Cohere, and Mistral are expected to have a total revenue of no more than $40 billion by 2026.

Investing nearly $700 billion in the foundational layer and only receiving a few hundred billion back from the application layer.

Isn't this severe asymmetry a bubble?

One cannot simply and crudely conclude this.

There is a key point that cannot be overlooked.

In March 2023, when OpenAI released GPT-4, the mixed cost per million tokens input was about $30.

By April 2025, with the optimization of model architecture and the improvement of inference computing power, the price for models of equivalent intelligence level plummeted to $0.1-$0.15 per million tokens.

According to Stanford University's "AI Index Report" and TokenCost data: AI inference costs have dropped by over 99.7% in the past two years.

According to traditional linear thinking, when costs plummet, corporate AI spending should decrease.

But the reality is that corporate AI cloud spending tripled between 2024 and 2025.

Why?

Because when the marginal cost of "intelligence" approaches zero, AI is no longer just a simple text summarizer or chat machine, but has entered a new era of intelligent agents and multimodal enhanced retrieval.

Companies are beginning to let AI agents automatically run thousands of tasks, writing code, scanning millions of legal contracts, and simulating biological experiments.

Cheap tokens have unlocked a vast amount of long-tail demand that was previously constrained by costs and could not be commercialized.

This point can be compared between NVIDIA in 2026 and Cisco, the network hardware giant of 2000, to see the clues.

The ecological niches of both are extremely similar, but their underlying financial health is worlds apart.

This precisely confirms the economic "Jevons Paradox": technological progress improves energy efficiency, and not only does it not reduce energy consumption, but rather, due to lower costs, demand increases.

Even after experiencing the so-called "DeepSeek moment" early last year, the market quickly sobered up in the following months: the more optimized the algorithms, the lower the threshold for companies to adopt AI, and ultimately, the total consumption of computing power increased exponentially.

It is precisely for this reason that AI is likely to gradually embed itself into almost all old industries.

Just as all industries have been engaging in Internet+ over the past twenty years.

From SaaS software to biomedicine, to advanced manufacturing robots driven by embodied intelligence, by 2026, almost all industries are embracing AI+.

No one will discuss "Should we use AI?" but rather worry about "Is our data well cleaned? Is the API call quota sufficient? Is the RAG architecture optimal?"

Currently, there is indeed a bubble in the AI industry.

But for companies, if you do not embrace the bubble, you will be crushed by the times.

This point has already been validated by nearly twenty years of the internet era.

03

At present, we are undoubtedly at a very critical node in the technology lifecycle: on the eve of the "valley of disillusionment" on the Gartner technology maturity curve, or the turning point in the theory of "Technological Revolution and Financial Capital."

The AI bubble is actually already bursting; many people just haven't realized it.

In the past few years, a large number of venture capitalists have developed a fear of measures.

A few newcomers can write dozens of pages of PPT and wrap it in OpenAI's API to raise funds. Now, as the tide recedes, these companies without a moat and only concepts are dying in large numbers.

This is the market undergoing self-purification and also a manifestation of the bubble bursting.

But this is just the surface.

The deeper logic of the market is undergoing three profound evolutions:

First, the value shift from CapEx to OpEx

Currently, the money is being made by those selling shovels—NVIDIA, TSMC, and those selling optical modules and server liquid cooling equipment are reaping most of the benefits.

However, as computing power gradually becomes "infrastructural," like water and electricity, the real excess profits will gradually shift to the application layer.

That is, those AI-native companies that can use extremely low-cost tokens to truly solve vertical industry pain points and reshape business processes (OpEx optimization).

Second, valuation multiple compression and performance digestion

The market's high valuation of AI infrastructure does not necessarily mean it will collapse.

In many cases, the rapid growth of corporate profits can gradually digest the high valuations through a "time-for-space" approach.

As long as the revenue growth of cloud computing giants keeps pace with the depreciation rate of capital expenditures, this game of hot potato can evolve into an unprecedented industrial upgrade.

For example, global automotive manufacturing giants and chip giants have shortened the R&D to mass production cycle of new products by 35% and improved overall line equipment efficiency by 18% through the introduction of end-to-end AI twin technology.

In the financial industry, by 2026, quantitative trading, risk control, and credit assessment are fully dominated by multimodal agents. AI is not only processing macro expectations with microsecond timestamps but is also deeply involved in every micro-level asset pricing.

In industries that heavily rely on senior expertise, such as law, medicine, and auditing, AI has also completed its transformation from "junior assistant" to "partner-level expert."

ChatGPT, Gemini, and Claude have over 1 billion active users, a significant portion of whom use them as substitutes for daily high-intensity intellectual labor.

Including you and me.

All of the above are real occurrences that everyone can see.

04

Looking back at the magnificent history of technology, Schumpeter's concept of "creative destruction" is always in play.

The capital market is always impatient, hoping to invest $1 today and earn back $10 tomorrow.

When nearly $700 billion in infrastructure investment cannot be fully converted into application-side profits in the short term, the market will inevitably face a brutal reshuffle.

Eliminating those speculative shell companies that rely solely on PPT presentations to get by, leaving behind those with real technological foundations and practical scenarios.

After the reshuffle, those cheap and massive computing centers and highly optimized model algorithms will serve various industries at extremely low prices.

After 2000, humanity entered a digital age where all industries could not do without the internet.

Today, we are also irreversibly heading towards an intelligent and prosperous era where all industries are vertically integrated and empowered by AI.

Amidst the clamor of the bubble, the underlying productive potential is not inflated at all.