If the AI bubble has already burst, who will truly remain?

This article is sourced from: GeLong Chengbei Xu Gong

Data support: Gougu Big Data

The AI bubble is becoming the most divisive consensus in the global market. Dalio says the bubble is already high, while Jensen Huang says the opportunity has just begun; one sees the overheating of the capital market, while the other sees the beginning of a productivity revolution.

The real question is not whether there is a bubble in AI, but what will remain after the bubble bursts. The internet bubble of 2000 caused the Nasdaq to plummet, companies to go bankrupt, and wealth to evaporate, but it also left behind the infrastructure of submarine cables, broadband networks, and cloud computing, ultimately supporting Amazon, Netflix, YouTube, and the mobile internet.

Today's AI is also in a similar position. On one side, hundreds of billions of dollars are being poured into data centers, electricity, liquid cooling, optical modules, and GPUs; on the other side, there is a huge gap where application revenue has not yet fully materialized. The bubble clearly exists, but the underlying productivity is not inflated. When the cost of tokens plummets and intelligence begins to be called upon like water and electricity, AI will no longer just be a chat tool but will enter the real workflows of coding, 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. Below, enjoy:

In recent days, the market has experienced severe fluctuations, and the "AI bubble theory" has become rampant.

• Bridgewater founder Dalio 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 certainly is.

However, bubbles in the tech sector are often the only way society can pay tribute to disruptive advanced productivity. It is not merely a pejorative term.

In the long run, this is a phenomenon that will inevitably occur at the beginning of advanced productivity.

Many people are comparing the current situation to the internet bubble of 2000, worrying greatly. The internet bubble of that year did indeed lead to a nearly 78% drop in the Nasdaq and the evaporation of over $5 trillion in wealth.

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 bubble 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 The "IQ Tax" That Innovation Must Pay

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. 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 (which sold 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 the bubble bursts is not due to the internet technology itself, but because the physical construction speed of the infrastructure cannot keep up with the market's pace.

For example, those once-thriving telecom companies (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 the mobile internet.

Without the frenzied pre-investment in telecom infrastructure around the year 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 $7 in 2001, a drop of over 90%. But it survived because its underlying business logic, "reconstructing retail with the internet," aligned with the direction of advanced productivity.

This is the classic Amara's Law: underestimating the long-term impact of a new technology while overestimating its short-term effects. In the early stages of a technological revolution, the frenzy of speculative capital inevitably leads to over-investment, creating a bubble. This is the "IQ tax" that innovation must pay. But when the bubble dissipates, what remains will be a more resilient advanced productivity.

02 Why Are Corporate AI Expenditures Rising Instead of Falling?

Looking ahead 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 projected 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, Mistral, and Perplexity are expected to have a combined total revenue of no more than $40 billion in 2026.

Investing nearly $700 billion in the foundational layer while only receiving a few hundred billion from the application layer. This severe asymmetry—what else could it be but a bubble?

We cannot simply and bluntly conclude this. There is a key point that cannot be overlooked:

• In March 2023, when OpenAI released GPT-4, the mixed cost for every 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 dropped 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, if costs plummet, corporate AI expenditures 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, write code, scan millions of legal contracts, and simulate 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 can be compared to Nvidia in 2026 and Cisco, the network hardware giant of 2000. Both have extremely similar ecological niches, but their underlying financial health is worlds apart.

(Nvidia vs. Cisco's hardcore financial comparison)

This precisely confirms the economic "Jevons Paradox": technological progress improves energy efficiency, and 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 regained its composure in the following months: the more optimized the algorithms, the lower the threshold for companies to adopt AI, ultimately leading to an exponential increase in total computing power consumption.

It is precisely for this reason that AI is likely to gradually embed itself into almost all old industries. Just as every industry has been embracing internet+ over the past twenty years. From SaaS software to biomedicine, to advanced manufacturing robots driven by embodied intelligence, in 2026, nearly all industries are embracing AI+. No one is discussing "should we use AI," but rather worrying "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 has already been proven in the nearly twenty years of the internet era.

03 The Deep Evolution of the Market: From Infrastructure to Applications

We are undoubtedly at a very critical node in the technology lifecycle: on the eve of the "trough 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. A few newcomers can write dozens of pages of PPT and package an OpenAI API layer to raise money. 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 is also a manifestation of the bubble bursting. But this is just the surface. The deep 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. But 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 valuation of AI infrastructure is relatively high, but that does not mean it will necessarily collapse. In many cases, the rapid growth of corporate profits can gradually digest the high valuations by "exchanging time for space." 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% through the introduction of end-to-end AI twin technology, and improved the overall efficiency of the production line by 18%.

• 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 like law, healthcare, and auditing, which heavily rely on senior expertise, 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 it as a substitute for daily high-intensity cognitive labor. Including you and me. All of the above are real occurrences that everyone can see.

04 Conclusion

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 reshuffling. It will eliminate those speculative shell companies that rely solely on presenting PPTs and leave behind those with real technological foundations and practical scenarios.

After the reshuffling, those cheap and massive computing centers, along with 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 era where all industries are vertically integrated and empowered by AI.

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