Source: Gelong, North City's Mr. Xu
Data support: GogoBigData
The AI bubble is becoming the most divisive consensus in the global market. Dalio says the bubble is already high, while Huang Renxun says the opportunity is just beginning; one sees overheating in the capital market, the other sees the dawn of a productivity revolution.
The real issue is not whether there is an AI bubble, but what will be left behind after it bursts. The dot-com bubble of 2000 caused the Nasdaq to plummet, companies to collapse, and wealth to evaporate, but it also left behind the infrastructure of submarine cables, broadband networks, and cloud computing, which ultimately supported Amazon, Netflix, YouTube, and the mobile internet.
Today, AI stands at a similar juncture. On one side, hundreds of billions of dollars are pouring into data centers, electricity, liquid cooling, optical modules, and GPUs; on the other side, there is a huge gap yet to be filled by application revenues. A bubble clearly exists, but the underlying productive forces are not inflated. When token costs plummet and intelligence begins to be called upon like water and electricity, AI will no longer be just a chatting tool, but will enter real workflows in code, healthcare, finance, law, manufacturing, and scientific research. The market will wash away shell companies and PPT entrepreneurs, but it will not reverse the direction of AI+. The bubble will burst, but the industry will remain. Enjoy:
Recently, the market has experienced sharp fluctuations, and theories of an "AI bubble" are rampant.
-
Ray Dalio, founder of Bridgewater Associates, says: There is a bubble in the AI market, and it is "relatively high."
-
NVIDIA CEO Jensen Huang says: There is a huge opportunity in AI, and the demand for computing power is just beginning to explode.
Who should we believe?
They are both correct.
Is there a bubble in the AI industry? Absolutely.
However, bubbles in the tech sector are often the only tribute society can pay to disruptive advanced productive forces. They are not merely a derogatory term.
In the long run, this is an inevitable phenomenon at the dawn of advanced productive forces.
Many liken the current situation to the dot-com bubble of 2000, feeling deeply concerned. The dot-com bubble indeed caused the Nasdaq to plummet nearly 78%, evaporating over 5 trillion dollars in wealth.
But twenty years later, what industry can operate without the internet? Today, the value of the internet industry far exceeds that of the bubble era.
At least superficially, the AI bubble resembles that. The bubble in the capital market cannot stop nearly every sector in society from actively being empowered by AI.
AI+ is the inevitable trend. Just as all industries today cannot do without the internet, all industries in the future will not be able to do without AI.
01 The "IQ Tax" That Innovation Must Pay
In that era when companies could go public and raise money just by having .com in their name, the Nasdaq surged nearly 600% between 1995 and 2000. Then followed a two-and-a-half-year financial storm.
Iconic names of that time, like the software company MicroStrategy, plunged 62% in a single day due to accounting scandals and exaggerated claims; Pets.com (online dog food seller), Webvan (the pioneer of fresh food e-commerce) simply went bust... In the panic, almost everyone blamed the internet as a scam.
However, the physical infrastructure precipitated by the over-expenditure of speculative capital often nourishes the super giants of the next era at extremely low costs. The bubble burst not because of the internet technology itself, but because the physical construction speed of infrastructure could not keep pace with the market's rhythm.
For example, the global undersea cables and dense wavelength division multiplexing networks laid by those once-prominent telecom companies (like WorldCom, Global Crossing), which bankrupted them with heavy investment, became the perfect breeding ground for the later rise of Netflix, Zoom, and the mobile internet.
Without the frenzied and premature global investment in telecom infrastructure around 2000, the later video streaming explosion of YouTube and the cloud computing infrastructure would not have happened.
Amazon is the most typical example. Its stock price plummeted from a high of $107 in 1999 to $7 in 2001, a drop of over 90%. But it survived because its underlying business logic, "reconstructing retail with the network," aligned with the direction of advanced productive forces.
This is a classic case of Amara's Law: we tend to overestimate the effect of a new technology in the short run and underestimate its effect in the long run. In the early stages of a technological revolution, the fervor of speculative capital inevitably leads to overinvestment, forming a bubble. It's the IQ tax that innovation must pay. But when the bubble recedes, what remains will be more indestructible advanced productive forces.
02 Why Are Corporate AI Expenditures Rising Instead of Falling?
Back in 2026, the AI industry bubble appears even larger.
Just five major cloud service providers—Amazon, Google, Meta, Microsoft, and Oracle—are expected to have capital expenditures reaching $690 billion in 2026, with total AI infrastructure investment projected at $5.3 trillion by 2030. Of this, only about 25% is for GPUs; the remaining 75% is all poured into physical infrastructure: liquid cooling systems, power transmission, network switches, optical modules, and land.
On the revenue side, the combined total revenue in 2026 for all leading pure AI companies like OpenAI, Anthropic, Cohere, Mistral, Perplexity is projected not to exceed $40 billion.
The foundational layer invests nearly $700 billion, while the application layer recovers tens of billions. This severe asymmetry—isn't that a bubble?
Such a conclusion cannot be drawn so simply and crudely. One critical point must not 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 model architecture optimization and inference computing power improvements, the price per million tokens for models of comparable intelligence plummeted to $0.1–$0.15.
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.
Following traditional linear thinking, if 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 chatting companion; it has entered a new era of agents and multi-modal enhanced retrieval. Enterprises are starting to let AI agents automatically run thousands of cycles of tasks—writing code, scanning millions of legal contracts, simulating biological experiments.
Cheap tokens have unlocked a massive volume of long-tail demands that were previously not commercially viable due to cost constraints.
Comparing NVIDIA in 2026 to Cisco, the networking hardware titan of 2000, also reveals this trend. Their ecosystem positions are remarkably similar, but their underlying financial health is worlds apart.
(Hardcore Financial Comparison of NVIDIA and Cisco)
This precisely confirms the economic principle of the "Jevons Paradox": technological progress improves energy use efficiency, but instead of reducing energy consumption, it leads to greater demand due to lower costs.
Even after last year's so-called "DeepSeek moment," the market quickly sobered up in the following months: the more optimized the algorithms, the lower the threshold for enterprises to adopt AI, and ultimately, the total computing power consumption rises exponentially instead.
It is precisely for this reason that AI has the potential to gradually embed itself into almost all old industries. Just as all industries embraced Internet+ over the past two decades. From SaaS software to biomedicine, to advanced manufacturing robots driven by embodied intelligence, by 2026, almost every industry is embracing AI+. No one discusses "whether we should use AI," but instead worries: "Is our data properly cleaned? Do we have enough API call quotas? Is the RAG architecture optimal?"
Currently, there is indeed a bubble in the AI industry. But for enterprises, if you don't embrace the bubble, you will be crushed by the times. This has been proven over the past two decades of the internet era.
03 The Market's Deep Evolution: From Infrastructure to Application
Currently, we are undoubtedly at a critical node in the technology life cycle: just before the "Trough of Disillusionment" on Gartner's Hype Cycle, or at the turning point described by the theory of "Technological Revolutions and Financial Capital."
The AI bubble is already bursting, though many haven't realized it. In the past, any upstart could write a few dozen pages of PPT, wrap a layer of OpenAI's API, and secure funding. Now, as the tide recedes, these companies with no moat and only concepts are dying en masse.
This is the market purifying itself, a manifestation of the bubble bursting. But this is only the surface. The deep logic of the market is undergoing three profound evolutions:
First, value transfer from CapEx to OpEx
Currently, money is being made by those selling shovels—NVIDIA, TSMC, and companies selling optical modules and server liquid cooling equipment are reaping most of the benefits. However, as computing power gradually becomes "infrastructure-ized," like water and electricity, the real excess profits will gradually shift to the application layer. These are the AI-native enterprises that can use extremely low-cost tokens to genuinely solve pain points in vertical industries and reshape business processes (OpEx optimization).
Second, valuation multiple compression and earnings digestion
The high valuations assigned to AI infrastructure do not necessarily imply an impending crash. In many cases, the rapid growth in corporate earnings can digest the high valuations over time, trading "time for space." As long as the revenue growth rate of cloud computing giants keeps pace with the depreciation rate of their capital expenditures, this game of hot potato can evolve into an unprecedented industrial upgrade.
-
For example, leading global automakers and chip giants, by introducing end-to-end AI twin technology, have shortened the cycle from R&D to mass production of new products by 35% and improved overall equipment effectiveness by 18%.
-
In the financial industry, by 2026, quantitative trading, risk control, and credit assessment are already fully dominated by multi-modal Agents. AI not only processes macro expectations at microsecond-level timestamps but is deeply involved in every micro-level asset pricing.
-
In industries highly reliant on senior professional knowledge, such as law, healthcare, and auditing, AI has already completed its transformation from a "junior assistant" to a "partner-level expert."
Among the over 1 billion active users of ChatGPT, Gemini, and Claude, a significant portion uses them as substitutes for daily high-intensity intellectual labor. This includes you and me. All of the above are real things happening, visible to everyone.
04 Conclusion
Looking back at the magnificent history of technology, the "creative destruction" proposed by Schumpeter is always at play.
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 transformed into application-layer profits in the short term, the market is bound to face a brutal reshuffle. It will eliminate speculative shell companies that rely solely on flashy PPTs, leaving behind those with genuine technological depth and practical implementation scenarios.
After this reshuffle, those cheap yet vast computing centers and highly optimized model algorithms will serve thousands of industries at extremely low prices.
After 2000, humanity ushered in a digital era where no industry could do without the internet. Today, we are also irreversibly moving towards an intelligent golden age where all industries will be governed and empowered by AI.
Amidst the clamor of the bubble, the underlying productive potential carries not a single drop of moisture.
