Artificial intelligence is big. So big that we have to wonder some times if and when it will come crashing down on us.
Now, we recently asked if AI infrastructure and software could crash the markets due to structural issues, but with recent volatility across the technology sector, we’re returning today to discuss the possibility of an AI bubble and the likelihood that a bubble bursts—or is bursting.
This week’s financial market turmoil has led to a deluge of opinions on when, how and why an AI bubble has already burst or will burst in the near future. Before we proceed, we’ll go ahead and break one of our own rules by taking a few shots at the messengers of AI financial doom. When we peruse the discussion of an AI bubble across the internet, we find familiar voices. The usual suspects, if you will.
The most vociferous proponents of an AI bubble seem to be the ones who are hostile to the technology itself—to borrow a line from “The Dark Knight,” these are the people who just want to watch the world burn. The next most likely source of bubble-talk are the usual investing suspects who are always talking about the next market crash, often because the products they are hawking offer some sort of purported safety, derisking or protection for investment portfolios—we’re looking at you, fixed income managers. Finally, we have the chicken-littles, smaller professional and personal investors who are always looking over their shoulders for the next risk of a wipeout.
But there are definitely others who, like ourselves, feel like there is an AI bubble for reasons other than their personal ideology or other interests like their own business and investment revenues. Whether we speak positively or negatively of AI, you can be sure we’re not talking our own book when we render an opinion. Of course there’s a bubble, but we also argue that underneath a market bubble, there’s plenty of tangible gains to be made with artificial intelligence.
Artificial intelligence is real.
And So Is the AI Bubble
A bubble does not mean the underlying technology is fraudulent. Nor does it imply that the technology lacks transformative potential. Instead, financial bubbles occur when investor expectations become disconnected from realistic assumptions regarding future earnings and cash flows. Classic bubbles generally contain several recognizable characteristics:
- explosive capital investment
- rapidly rising valuations
- easy financing
- narratives that justify almost unlimited growth
- concentration of investment into a relatively small group of companies
- diminishing concern over traditional valuation metrics
Many of these characteristics undeniably exist within today’s AI market. Yet many economists argue that those same characteristics also accompany nearly every genuine general-purpose technology revolution. This explains why opinions remain sharply divided.
The Bullish Argument: This Isn’t 1999
Those rejecting bubble comparisons begin with a straightforward observation. Unlike many speculative episodes, many of today’s AI leaders are extraordinarily profitable companies.
- Microsoft.
- Alphabet.
- Amazon.
- Meta.
- NVIDIA.
- Broadcom.
- Oracle.
These are not venture-backed startups with no revenue. Collectively they generate hundreds of billions of dollars in annual profits while funding AI investments from enormous existing businesses. That distinction is critical. During the dot-com era many companies had little more than websites and ambitious business plans. Today’s AI leaders possess:
- global cloud infrastructure
- dominant software platforms
- enterprise customers
- recurring subscription revenue
- substantial free cash flow
- investment-grade balance sheets
Even companies spending extraordinary sums on AI infrastructure continue posting impressive revenue growth.
Recent earnings reinforced that argument. Amazon raised planned capital expenditures to roughly $220 billion after reporting accelerating AWS growth and strong AI demand extending well into future years. Microsoft likewise continued reporting rapid Azure expansion while investors rewarded what many viewed as disciplined capital allocation rather than unlimited spending. To AI bulls, these are not symptoms of irrational exuberance. They are evidence of a technology entering commercial maturity.
The Infrastructure Analogy
Supporters increasingly compare today’s AI spending not to cryptocurrency speculation but to infrastructure booms. Their argument resembles this:
- Before electricity transformed manufacturing, someone had to build power plants.
- Before automobiles dominated transportation, roads had to be constructed.
- Before the internet connected the world, fiber-optic networks had to be installed.
AI likewise requires enormous physical infrastructure:
- data centers
- electrical generation
- semiconductor fabrication
- networking equipment
- cooling systems
- optical interconnects
- cloud capacity
Much of today’s unprecedented capital expenditure therefore represents building the foundation upon which future AI applications will operate. Viewed through that lens, current spending appears less like speculation than industrial investment.
Why Investors Are Growing Nervous Anyway
The bullish story, however, has collided with an equally compelling financial reality. The infrastructure is becoming extraordinarily expensive. During just the past several weeks, analysts have highlighted projections that the largest technology companies may spend roughly three-quarters of a trillion dollars on AI infrastructure during 2026 alone, with cumulative investment well above one trillion dollars over a relatively short period.
That spending includes:
- GPUs
- custom AI accelerators
- networking equipment
- energy contracts
- construction
- cooling
- transmission infrastructure
- real estate
- specialized labor
Unlike software, these investments cannot easily be scaled back. Once billions have been committed to a data center, the spending largely becomes irreversible. Consequently, investors have begun asking a difficult question: When do these investments begin producing proportional profits?
The Great Monetization Debate
This question lies at the center of nearly every current discussion about an AI bubble. Enterprise AI adoption continues expanding. Consumers increasingly use AI assistants. Financial institutions deploy AI copilots. Software vendors integrate generative AI into nearly every product. Yet revenue growth has not always matched infrastructure spending.
Recent commentary from The Economist argued that AI revenues continue growing rapidly—but perhaps not rapidly enough relative to extraordinary capital expenditures. Investor concerns increasingly focus on whether future returns can justify today’s unprecedented spending levels.
In other words:
- Adoption is real.
- Demand is real.
- Revenue is real.
The uncertainty concerns whether revenue will become sufficiently large quickly enough. That distinction separates today’s debate from simplistic claims that AI lacks value. Few credible analysts make that argument. Instead, they question valuation timing.
Cash Burn Becomes the New Market Obsession
Wall Street has gradually shifted attention away from AI model quality toward balance sheets. Several major financial publications have noted that hyperscalers are assuming increasingly large long-term commitments for data centers, chips and infrastructure that may take years to generate proportional returns.
As one investment manager recently observed, the issue is no longer whether AI works. The issue is whether capital markets can comfortably finance this scale of investment before profits catch up. That concern intensified during July as technology earnings prompted greater scrutiny of free cash flow rather than merely AI announcements.
Markets increasingly rewarded evidence of spending discipline. Microsoft’s relatively restrained capital spending guidance was interpreted positively, suggesting investors may now prefer companies demonstrating financial discipline rather than simply spending the most on AI. That represents an important psychological shift. Earlier in the AI boom, higher spending implied stronger leadership. Increasingly, higher spending also implies greater financial risk.
The Market Has Become More Selective
Perhaps the clearest evidence that sentiment is evolving is that investors are no longer rewarding every AI announcement equally. Companies must increasingly demonstrate:
- measurable customer demand
- recurring revenue
- improving margins
- enterprise adoption
- pricing power
- operational efficiency
Simple references to AI during earnings calls no longer produce automatic enthusiasm. Markets have become discriminating.
Ironically, that development may actually represent a sign of improving market health rather than imminent collapse. If the optimistic view argues that artificial intelligence resembles the arrival of electricity or the internet, the skeptical view begins with a different historical comparison.
To many economists, investors and market historians, today’s environment increasingly resembles the late stages of the dot-com boom. That comparison does not suggest artificial intelligence lacks transformative potential. Rather, it suggests financial markets may once again be assigning unrealistic valuations to technologies that will ultimately succeed. This distinction is central to understanding today’s debate.
During the late 1990s, internet technology unquestionably changed commerce, communications and society. Amazon, Google and eBay emerged from that period to become some of the world’s most valuable companies. Yet thousands of internet firms disappeared. Investors who purchased technology stocks indiscriminately often suffered losses exceeding 80%. The internet wasn’t the bubble. Internet valuations were. Many observers believe artificial intelligence may eventually produce a similar outcome.
“Every Revolution Has a Bubble”
One of the most common arguments made by economists is surprisingly balanced. They acknowledge AI’s extraordinary potential while simultaneously expecting a speculative correction. Their reasoning follows a historical pattern. Nearly every transformational technology has produced excessive investment.
- Railroads.
- Automobiles.
- Electricity.
- Telecommunications.
- The internet.
- Fiber optics.
- Housing finance.
Each attracted legitimate enthusiasm. Each also encouraged excessive speculation. The resulting crashes destroyed capital without eliminating the underlying innovation.
This perspective suggests investors should distinguish between technological revolutions and investment cycles. History demonstrates they rarely unfold smoothly. Instead, they frequently overshoot before eventually settling into sustainable growth.
The Magnificent Seven Problem
Much of today’s concern centers on market concentration. Never before have so few companies represented such a large percentage of global equity valuations.The largest technology firms dominate:
- market capitalization
- AI infrastructure spending
- semiconductor demand
- cloud computing
- enterprise software
- venture investment
- AI model development
That concentration creates extraordinary opportunities. It also creates extraordinary risks. If investor confidence weakens toward only a handful of these companies, broad equity indexes could experience meaningful declines despite relatively healthy economic conditions elsewhere.
Unlike previous technology cycles, today’s AI investment boom has become deeply intertwined with retirement accounts, pension funds, ETFs and index investing. Consequently, an AI correction would likely affect investors far beyond Silicon Valley.
The Capital Expenditure Question
Perhaps no statistic receives more attention than capital expenditures. Technology companies have announced spending plans that would have seemed almost unimaginable only a few years ago. Collectively, hyperscalers now expect to invest hundreds of billions of dollars annually constructing AI infrastructure. Those expenditures include:
- semiconductor purchases
- networking hardware
- data-center construction
- electrical infrastructure
- cooling systems
- custom silicon
- land acquisition
- transmission capacity
Wall Street generally applauded these announcements initially. Increasingly, however, analysts are asking whether spending has become self-reinforcing. One concern raised in recent reporting is that technology companies may feel compelled to continue spending simply because competitors continue spending.
No company wants to appear underinvested if artificial general intelligence eventually arrives. Yet no company wants to be left holding hundreds of billions of dollars in underutilized infrastructure if demand develops more slowly than expected. This dynamic resembles what economists call an investment race. Investment decisions become driven partly by competitive pressure rather than immediate financial return.
Data Centers Have Become the New Dot-Com Fiber
Several commentators have drawn comparisons between today’s AI infrastructure boom and the telecommunications expansion of the late 1990s. Then, companies constructed enormous fiber-optic networks anticipating limitless internet traffic. Eventually, demand arrived. But not quickly enough. The result was years of excess capacity. Many firms failed before utilization justified their investments.
Some analysts wonder whether AI data centers could produce a similar imbalance. If enterprise adoption accelerates exactly as forecast, current construction may prove justified. If adoption slows—even modestly—markets could suddenly question enormous infrastructure commitments. That uncertainty explains why investors increasingly scrutinize utilization rates rather than simply celebrating construction announcements.
Hidden Debt and Financial Engineering
Another emerging concern involves how AI investment is being financed. Recent investigations have examined whether some technology firms increasingly rely upon long-term contractual obligations and financing arrangements that do not always appear prominently in conventional debt metrics. Analysts note that these obligations include:
- long-term leases
- power agreements
- infrastructure commitments
- cloud financing
- semiconductor purchase contracts
Although these arrangements differ from traditional borrowing, they nevertheless represent substantial future financial obligations.
Some market observers argue investors have underestimated the cumulative effect of these commitments. Others counter that companies generating tens of billions of dollars in annual free cash flow remain well positioned to absorb them. Regardless of which interpretation proves correct, balance-sheet analysis has become considerably more important than it was earlier in the AI boom.
Revenue Has Improved—But Expectations Have Improved Even Faster
Perhaps the greatest challenge facing AI companies is psychological rather than technological. Expectations have become enormous. Markets no longer reward merely good results. They increasingly demand extraordinary results. For example, a company might report:
- 30% revenue growth
- expanding enterprise adoption
- record profits
- accelerating AI subscriptions
Yet still disappoint investors if analysts expected even faster growth.
This phenomenon has become increasingly common throughout technology earnings. The higher valuations climb, the narrower the margin for disappointment becomes. Consequently, companies must continuously exceed already optimistic assumptions simply to maintain existing share prices. That dynamic often characterizes mature speculative markets.
Venture Capital Faces Its Own Reckoning
Public markets are not the only focus of bubble discussions. Private markets have experienced equally remarkable enthusiasm. Foundation-model developers have raised financing rounds measured in tens of billions of dollars. Infrastructure startups routinely command multi-billion-dollar valuations before reaching profitability. Agentic AI companies continue attracting unprecedented venture investment.
Many venture investors defend these valuations. They argue the addressable market for enterprise AI remains enormous. Critics, however, observe that startup valuations increasingly assume successful commercialization years before sustainable business models have fully emerged.
Historically, venture capital cycles have tended to amplify public market enthusiasm. When public valuations contract, private valuations often follow—sometimes with considerable delay.
Consumer AI Adoption Remains Uneven
Another factor tempering investor enthusiasm involves actual consumer behavior. Generative AI has undoubtedly achieved extraordinary adoption. Millions of consumers use AI assistants daily. Businesses increasingly integrate AI into workflows. Yet surveys continue showing uneven usage across demographics, professions and industries.
Some workers embrace AI enthusiastically. Others remain reluctant. Educational institutions continue debating appropriate usage. Creative professionals increasingly express concerns regarding intellectual property, employment and originality.
Recent research also highlights generational differences in AI acceptance, suggesting adoption may be substantial without becoming universal. Investors therefore face a complicated question: How quickly will widespread enthusiasm translate into sustainable revenue? That question remains surprisingly difficult to answer.
Enterprise Adoption Is Real—But Often Incremental
Corporate America likewise presents a nuanced picture. Most large organizations now deploy AI somewhere within their operations. Banks summarize meetings. Insurers automate claims documentation. Asset managers analyze research. Law firms draft contracts. Healthcare providers produce clinical documentation.
Yet these deployments frequently remain incremental. Companies often begin with:
- copilots
- summarization
- search
- customer support
- coding assistants
- workflow automation
These applications generate measurable productivity improvements. They do not necessarily justify unlimited infrastructure investment overnight. Many executives therefore describe AI as a long-term transformation rather than a short-term profit engine. Markets, however, often price stocks based upon expectations several years into the future. When expectations move faster than implementation, valuation questions naturally arise.
What Could Actually Pop an AI Bubble?
The most important point often overlooked is that bubbles rarely burst because observers identify them. Instead, they burst because some catalyst changes investor psychology. Analysts currently identify several possible catalysts.
- Earnings disappointments
If several major technology companies report slowing AI revenue simultaneously, investors may begin reassessing growth assumptions.
- Reduced capital expenditures
Should hyperscalers unexpectedly scale back infrastructure spending, markets might interpret the decision as weaker-than-expected AI demand.
- Slower enterprise adoption
If corporations continue experimenting but delay broad deployment, revenue growth could fall below expectations.
- Semiconductor oversupply
Should GPU demand weaken after years of extraordinary shortages, semiconductor valuations could come under pressure.
- Credit tightening
Large infrastructure projects depend upon favorable financing conditions. Higher interest rates or tighter credit could slow investment.
- Regulatory intervention
Governments worldwide continue developing AI regulation. Unexpected restrictions could affect profitability, deployment timelines or investor confidence.
- A Major AI Company Stumbles
Perhaps the greatest risk involves concentration. Should a major AI leader report unexpectedly poor financial performance—or encounter governance, cybersecurity or operational problems—investor sentiment could change remarkably quickly.
None of these scenarios necessarily imply catastrophe. Instead, they illustrate why investors increasingly distinguish between AI as a technology and AI as an investment theme. History suggests those two stories eventually diverge. The difficult question facing today’s markets is when.
Whether artificial intelligence is experiencing a speculative bubble ultimately may prove less important than a more practical question. How should investors behave if markets become significantly more volatile? That question is especially relevant because history suggests investors consistently overestimate both the permanence of booms and the permanence of crashes. Technology revolutions rarely proceed in straight lines. Markets almost never price them perfectly.
Instead, periods of extraordinary optimism are often followed by periods of excessive pessimism before eventually settling into a more rational equilibrium.
Artificial intelligence may well follow that same pattern.
For investors, financial advisors and wealth managers, the objective is therefore not predicting the exact day an AI bubble might burst. It is constructing portfolios capable of surviving whichever scenario ultimately unfolds.
Would an AI Crash Resemble 2000?
The comparison most frequently made is the dot-com collapse. Yet there are important differences. During 1999 and early 2000, many internet companies generated little revenue. Numerous businesses lacked viable business models. Easy capital financed speculative ventures whose primary asset was often a memorable web address.
Today’s AI leaders present a very different financial picture. The largest beneficiaries of AI spending are generally among the world’s strongest companies.
They possess:
- enormous cash reserves
- diversified revenue streams
- recurring enterprise customers
- mature management teams
- substantial operating profits
- global infrastructure
That reality leads many economists to believe any correction would likely differ from the collapse experienced twenty-five years ago. Rather than thousands of bankruptcies, the more probable outcome may involve:
- valuation compression
- slower earnings growth
- reduced venture funding
- consolidation
- selective bankruptcies among smaller firms
- lower—but healthier—expectations
Reuters recently published an opinion that even a meaningful AI market decline would be unlikely to resemble either the 2000 dot-com collapse or the 2008 financial crisis because today’s leading companies possess much stronger balance sheets and far more established businesses. Instead, any correction would more likely represent a repricing of expectations than a systemic financial breakdown. That distinction matters enormously for long-term investors.
The Most Likely Scenario: Consolidation Rather Than Collapse
We have no crystal ball, and if we did, we wouldn’t trust it any more than our kindergartener’s magic 8-ball. Many analysts increasingly describe the coming years not as the bursting of an AI bubble but as a period of consolidation. Under that scenario:
- AI adoption continues.
- Enterprise spending grows.
- Revenue expands.
- But stock prices stop rising faster than fundamentals.
In other words, businesses gradually “grow into” their valuations—or valuations gradually decline until they better reflect earnings. This process has occurred repeatedly throughout financial history. Markets become enthusiastic. Reality catches up. Eventually expectations become sustainable. Such an outcome would likely disappoint speculators while rewarding patient long-term investors.
Winners and Losers May Separate Dramatically
If history serves as a guide, the companies surviving any AI correction may become extraordinarily valuable. After the dot-com collapse:
- Amazon survived.
- Google emerged.
- Salesforce expanded.
- Netflix evolved.
- Thousands of competitors disappeared.
Artificial intelligence could produce a similar pattern.
Companies possessing genuine competitive advantages—proprietary data, strong enterprise relationships, differentiated models, semiconductor leadership, software ecosystems and recurring revenue—may strengthen further after speculative excesses disappear.
Meanwhile, firms relying primarily upon excitement, marketing or generous venture funding could struggle. This distinction is particularly important because today’s AI marketplace includes thousands of companies with dramatically different business models despite often being grouped together by investors under the single label of “AI.” History suggests markets eventually become much more selective.
What Financial Advisors Can Tell Clients
For financial advisors, an AI correction would present both challenges and opportunities. The immediate challenge would involve managing client emotions. Many investors who entered AI-related investments during periods of rapid appreciation may experience significant anxiety if prices decline sharply.
Behavioral finance teaches that losses generally feel substantially more painful than equivalent gains feel rewarding. Consequently, advisors may spend considerably more time counseling clients during market declines than during market advances. The most effective advisors are therefore unlikely to focus exclusively on predicting markets. Instead, they may emphasize disciplined investment processes.
Clients should understand that:
- volatility is normal
- technological revolutions are rarely smooth
- diversification remains valuable
- concentration increases risk
- long-term investing requires patience
These principles have remained remarkably consistent across decades of market history. Artificial intelligence does not eliminate them.
Diversification Still Matters
One surprising consequence of the AI boom has been increasing portfolio concentration. Many investors now possess substantial exposure to the same handful of technology companies through:
- index funds
- retirement plans
- growth mutual funds
- ETFs
- individual stock ownership
Some investors may therefore be considerably less diversified than they realize. Financial advisors increasingly emphasize examining actual economic exposure rather than simply counting the number of securities owned. Owning five different funds heavily concentrated in the same technology companies does not necessarily provide meaningful diversification.Should AI valuations compress simultaneously across the sector, seemingly diversified portfolios may perform more similarly than investors expect.
Financial Services May Be Among the Biggest Long-Term Winners
For you, dear readers, perhaps the most significant observation concerns financial services itself. As we wrote in last week’s “Looking Back,” unlike previous technology cycles focused primarily on consumer applications, AI is increasingly becoming embedded within institutional finance.
- Banks employ AI to detect fraud.
- Asset managers summarize research.
- Insurance companies automate underwriting.
- Wealth managers prepare client meetings.
- Broker-dealers monitor compliance.
- Investment firms increasingly develop AI agents capable of assisting analysts, operations teams and advisors.
Importantly, these applications generally improve efficiency rather than replacing financial professionals altogether. Recent research from the CFA Institute, Financial Planning Association and numerous industry surveys suggests AI adoption within finance continues emphasizing augmentation rather than wholesale replacement of expertise. Financial professionals increasingly supervise AI systems performing documentation, analysis and workflow automation while retaining responsibility for judgment, fiduciary obligations and client relationships.
That distinction may ultimately prove critical. Even if AI-related equity valuations fluctuate dramatically, AI deployment throughout financial services is likely to continue. The technology’s usefulness does not depend upon stock prices.
What Happens After the Bubble Bursts?
Suppose a significant AI correction eventually occurs. What follows? History provides several clues. Immediately afterward:
- Venture funding becomes more selective.
- Weak companies disappear.
- Hiring slows.
- Valuations decline.
- Media coverage turns pessimistic.
- Then something interesting usually happens.
- Engineers continue building.
- Researchers continue improving products.
- Customers continue adopting useful technology.
- The speculative cycle fades.
- The technological revolution continues.
The internet did not disappear after 2000. Cloud computing did not disappear after 2008. Mobile computing did not disappear after numerous smartphone corrections. Artificial intelligence would likely follow the same trajectory. The survivors may ultimately become stronger because speculative competition diminishes.
A Technology Revolution Can Still Produce an Investment Bubble
One misconception continues appearing throughout public discussions. People often frame the debate as a binary choice. Either AI changes everything, or AI is a bubble. History suggests both statements can simultaneously be true. Railroads changed everything. The railroad bubble still burst. The internet changed everything. The dot-com bubble still burst.
Artificial intelligence may likewise transform virtually every industry while simultaneously producing periods of excessive investor optimism. Understanding that possibility helps explain why respected economists disagree about timing while broadly agreeing about AI’s long-term significance.
The Long View
Evidence suggests the market has entered a more mature phase of the AI investment cycle. Enthusiasm remains high. Capital spending remains extraordinary. Enterprise adoption continues accelerating. Yet investors increasingly ask harder questions about valuation, cash flow, capital allocation and return on investment. That shift is healthy.
Markets ultimately reward businesses that create sustainable economic value rather than merely attracting attention. Whether today’s environment ultimately becomes known as the “AI Bubble” or the “AI Infrastructure Boom” may only become clear years from now.
What appears increasingly certain, however, is that artificial intelligence itself is unlikely to disappear. Some companies undoubtedly will. Some valuations almost certainly will. Some investment theses inevitably will. But the underlying technology continues advancing into nearly every corner of the global economy—including financial services, where AI is rapidly evolving from an experimental productivity tool into essential market infrastructure.
For investors and financial advisors, that distinction may be the most important conclusion of all.
The objective is not to predict every swing in market sentiment. It is to recognize that technological revolutions create both extraordinary opportunities and extraordinary risks—and that disciplined portfolio management, rigorous valuation analysis and diversified long-term investing remain just as valuable during an AI revolution as they were before one began.
Researched by DWN Staff
