Original Title: The AI Bubble is Already Bursting
Original Author: Chengbei Xu Gong, Gelong
Recently, the market has experienced significant volatility, with "AI bubble theory" gaining widespread attention.
Ray Dalio, founder of Bridgewater Associates, said: There is a bubble in the AI market, and the level is "relatively high."
Jensen Huang, CEO of NVIDIA, said: AI presents a huge opportunity, and the demand for computing power has just begun to explode.
Who should we believe?
Both of them are correct.
Does the AI industry have bubbles? It certainly does.
However, bubbles in the tech sector are often the only way for society to pay tribute to disruptive advanced productive forces.
It is not merely a pejorative term.
In the long run, this is an inevitable phenomenon when advanced productive forces first emerge.
Many people compare the current situation to the 2000 dot-com bubble, feeling deeply anxious.
The dot-com bubble indeed caused the Nasdaq to plummet by nearly 78%, evaporating over $5 trillion in wealth.
But twenty years later, which industry can function without the internet?
Today, the value of the internet industry far exceeds that of the bubble period.
The AI bubble, at least superficially, appears to be a similar situation.
The bubbles present in the capital market cannot stop almost every industry in society from actively being empowered by AI.
AI+ is an unstoppable trend.
Just as no industry today can function without the internet, in the future, no industry will be able to function without AI.
01
In that era when any company with a .com in its name could go public and raise money, the Nasdaq surged nearly 600% from 1995 to 2000. Subsequently, a financial storm lasting two and a half years ensued.
Well-known names from that time: software company MicroStrategy, due to accounting scandals and exaggerated claims, plummeted 62% in a single day; Pets.com (selling pet food online), Webvan (pioneer of fresh food e-commerce) directly went bankrupt.
......
In the panic, almost everyone accused the internet of being a scam.
However, the physical infrastructure deposited by the excessive spending of speculative capital often nurtures the super giants of the next era at extremely low costs.
The reason the bubble burst was not due to problems with internet technology itself, but because the pace of physical infrastructure construction could not keep up with the market's rhythm.
For example, the once-dominant telecom companies (like WorldCom, Global Crossing) poured massive sums into laying global submarine cables and dense wavelength division multiplexing (DWDM) networks. While these companies themselves went bankrupt, these cheap "information superhighways" became the perfect breeding ground for the later rise of Netflix, Zoom, and the mobile internet.
Without the crazy, preemptive global investment in telecom infrastructure around 2000, there would have been no subsequent explosion of YouTube's video streaming, let alone later cloud computing infrastructure.
The most typical example is Amazon.
Its stock price fell from a high of $107 in 1999 to a mere $7 in 2001, a drop of over 90%.
But it survived because its underlying business logic, "reconstructing retail with networks", aligned with the direction of advanced productive forces.
This is a classic case of Amara's Law: overestimating the short-term impact of a new technology while severely underestimating its long-term impact.
At the beginning of a technological revolution, the狂热 of speculative capital inevitably leads to overinvestment, forming bubbles.
This is the intelligence tax that innovation must pay.
But when the bubble subsides, what remains will be even more unshakeable advanced productive forces.
02
Returning to 2026, the bubble in the AI industry appears even larger.
Just the five major cloud service providers—Amazon, Google, Meta, Microsoft, and Oracle—are projected to have capital expenditures of $690 billion in 2026, with total AI infrastructure investment expected to reach $5.3 trillion by 2030.
Of this, only about 25% is spent on GPUs; the remaining 75% is entirely invested in physical infrastructure: liquid cooling systems, power transmission, network switches, optical modules, and land.
In terms of revenue, all leading pure-play AI companies, including OpenAI, Anthropic, Cohere, Mistral, and Perplexity, are expected to have a combined total revenue of no more than $40 billion in 2026.
Nearly $700 billion invested in the foundational layer, while the application layer generates hundreds of billions in return.
Such severe asymmetry, is this not a bubble?
We cannot jump to such a simple and crude conclusion.
There is a crucial point that cannot be overlooked.
In March 2023, when OpenAI released GPT-4, the mixed cost per million tokens of input was about $30.
By April 2025, with the optimization of model architectures and improvements in inference computing power, the price for models of comparable intelligence level had plummeted to $0.1-0.15 per million tokens.
According to Stanford University's "AI Index Report" and TokenCost data: AI inference costs have fallen by over 99.7% in the past two years.
According to traditional linear thinking, with costs plummeting, 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 chatbot; it has entered a new era of agents and multimodal augmented retrieval.
Companies are now having AI agents automatically run thousands of task loops: writing code, scanning millions of legal contracts, simulating biological experiments.
Cheap tokens have unlocked vast amounts of long-tail demand that were previously uncommercializable due to cost constraints.
We can also see this by comparing NVIDIA in 2026 with Cisco, the network hardware giant in 2000.
Their ecological positions are extremely similar, but their underlying financial health is vastly different.
This precisely validates the economic concept of "Jevons paradox": technological progress improves energy efficiency, but instead of reducing energy consumption, it leads to greater demand due to lower costs.
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 barrier for enterprise AI adoption, ultimately causing total computing power consumption to rise exponentially.
It is precisely because of this that AI has the potential to gradually embed itself into almost every traditional industry.
Just as all industries have embraced internet+ over the past two decades.
From SaaS software to biomedicine, to advanced manufacturing robotics driven by embodied intelligence, in 2026, almost every industry is embracing AI+.
No one is discussing "should we use AI?", but rather worrying "is our data cleaned? Do we have enough API call quotas? Is our RAG architecture optimal?"
Currently, there is indeed a bubble in the AI industry.
But for businesses, if you don't embrace the bubble, you will be crushed by the times.
This has been proven by the internet era over the past two decades.
03
Currently, we are undoubtedly at a critical node in the technology lifecycle: on the eve of the "Trough of Disillusionment" on the Gartner Hype Cycle, or at a turning point in the "Technological Revolutions and Financial Capital" theory.
The AI bubble is already bursting, but many people haven't realized it.
Over the past few years, many venture capitalists (VCs) developed a fear of missing out (FOMO).
A few rookies could raise money with just a few dozen pages of PowerPoint, wrapping an API layer over OpenAI. Now, as the tide recedes, these companies without moats, only concepts, are dying in droves.
This is the market's self-purification, and it is also a manifestation of the bubble bursting.
But this is only the surface.
Three profound evolutions are occurring in the market's underlying logic:
First, the shift in value from CapEx to OpEx
Currently, the money is mostly being made by the shovel sellers: NVIDIA, TSMC, and those selling optical modules and server liquid cooling equipment have reaped most of the dividends.
However, as computing power gradually becomes "infrastructure-ized," like water and electricity, the true excess profits will gradually shift to the application layer.
That is, to those native AI companies that can truly solve vertical industry pain points and reshape business processes (OpEx optimization) using extremely low-cost tokens.
Second, valuation multiple compression and performance digestion
The market's high valuation for AI infrastructure does not necessarily mean an imminent crash.
In many cases, the high-speed growth of corporate profits can gradually "exchange time for space," digesting lofty valuations over time.
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 and chip giants, by introducing end-to-end AI twin technology, have reduced the new product development-to-mass-production cycle by 35% and increased overall equipment effectiveness by 18%.
Also, in the financial industry, by 2026, quantitative trading, risk control, and credit assessment are fully dominated by multimodal Agents. AI is not only processing macroeconomic expectations with microsecond-level timestamps but also deeply involved in every micro-level asset pricing.
In highly knowledge-intensive industries such as law, healthcare, and auditing, AI has also completed its evolution from "junior assistant" to "partner-level expert."
Among the over 1 billion active users of ChatGPT, Gemini, and Claude, a significant portion use them as daily substitutes for high-intensity mental labor.
Including you and me.
All of the above are real, tangible events that everyone can see.
04
Looking back at the magnificent history of technology, the "creative destruction" proposed by Schumpeter is always playing out.
The capital market is always impatient, hoping that $1 invested today will yield $10 tomorrow.
When nearly $700 billion in infrastructure investment cannot be fully translated into profits at the application end in the short term, the market is bound to experience a brutal reshuffling.
It will eliminate those speculative shell companies that survive only on PowerPoint presentations and leave behind those with genuine technical substance and landing scenarios.
After the reshuffle, those cheap and massive computing centers and highly optimized model algorithms will serve all industries at extremely low prices.
After 2000, humanity entered a digital era where no industry could function without the internet.
Today, we are also irreversibly heading towards an era of intelligent supremacy where all industries are governed and empowered by AI.
Amidst the noise of bubbles, the underlying productive force potential has not a single drop of moisture.
