There has been a lot of talk lately about whether the AI boom is actually a bubble. Most of the pieces focus on analyzing if AI is similar or dissimilar to the dot-com bubble. That is all Ok, but we believe that is not the main issue. The question is not if this boom is more or less parallel to the dot-com bubble, but the BIG picture: How is the AI industry? How is the market at large? How is the economy? And only then: Are we in an AI bubble?

In order to give a comprehensive answer to that question, we will consider what things are really different from 1999, what things are materially worse, and what is the same.

After reading this piece, we hope you will be able to judge for yourself if the full picture, of both the good and the bad, accounts for a positive or not-so-great perspective on the market in the short term.

I. Why Some Things Really Are Better Than in the Pre–Dot-Com Era

A fair assessment of today’s AI boom has to start with what is genuinely different from the late 1990s. Contrary to the popular narrative that we’re simply replaying the dot-com script, several structural features of the current cycle are fundamentally stronger. These strengths don’t eliminate bubble dynamics—those come later—but they do explain why the early innings of the AI surge feel more grounded than the speculative internet frenzy of 1999.

1. Stronger, Cash-Generating Companies—Not Unproven Startups

In the late 1990s, capital flooded into firms with little revenue, untested products, and limited operating history. Today’s AI investment wave is being led by the largest, most profitable companies in the world. J.P. Morgan notes that corporate investment in AI infrastructure has been “torrential,” driven overwhelmingly by hyperscalers like Microsoft, Amazon, Alphabet, and Meta—firms with long operating records and robust balance sheets. These companies are not speculating with borrowed funds; they are reinvesting substantial free cash flow into data centers, semiconductors, and model development.

Michael Browne from Franklin Templeton makes this contrast explicit: unlike the dot-com era or earlier industrial manias, the AI boom is “notably funded by internal cash flow, not debt.” This reduces fragility and provides a buffer if expectations moderate. The underlying businesses powering this wave are real, diversified, and profitable.

2. A Market That Emits Capital, Not One That Absorbs It

Another important difference is how public markets are being treated by the companies at the center of the boom. Owen Lamont of Acadian Asset Management emphasizes that U.S. firms today are returning capital to shareholders through dividends and buybacks. In other words, the market is emitting money, not drawing it in to finance speculative expansion—very different from the pre-2000 period, when firms aggressively issued equity to take advantage of inflated prices.

Lamont contrasts this directly with the Enron and dot-com era: those firms required continuous inflows to sustain their valuations. Today, four of the five largest U.S. companies are net repurchasers. In Lamont’s view, this is not what a Ponzi-like bubble looks like.

To be sure, buybacks are not a perfect sign of health. Charlie Munger’s famous warning applies: some companies repurchase shares partly to keep prices elevated (as he put it: “it’s crazy and it’s immoral. But other than that, it’s completely fine.”). Still, the aggregate signal remains meaningfully different from the 1990s.

3. Less Visible Leverage, Fewer IPOs, and Fewer Classic Bubble Markers

One of the most striking divergences from 1999 is the absence of an IPO wave. Lamont points out that every major historic equity bubble—from the South Sea Bubble to the bicycle mania to the dot-com era—featured aggressive issuance. Firms exploit euphoric markets by selling stock. Today, issuance is negative, with net repurchases running near $1 trillion.

Similarly, the AI boom has so far avoided the explosive, margin-fueled speculation that characterized the late 1990s tech run-up.

4. A More Mature Monetization Landscape—At Least for Some Participants

While many AI companies still lack proven paths to profitability, the ecosystem around them is more commercially mature than during the internet’s early years. J.P. Morgan highlights AI’s visible value in productivity, cybersecurity, customer service, and software development.

It is critical not to overstate this point. Monetization remains ambiguous for several leading-edge firms. But compared with 1999—when business models were aspirational—the installed base of paying enterprise customers is materially larger today.

5. A Better Starting Point—But Not a Safe One

Taken together, these structural improvements seem to give today’s AI cycle a stronger foundation than the conditions that preceded the dot-com crash. The companies are larger, the financing is more disciplined, and the underlying technology is generating real economic value. These factors explain why many sophisticated institutions, from J.P. Morgan to Franklin Templeton, argue that the AI boom is not a bubble.

But some stronger foundations do not immunize the market from excess. They merely set the stage for why this bubble will look different than the previous ones.

II. Why Some Things Are Worse Than Before the Dot-Com Bubble

After considering the strengths that make this cycle better than the dot-com boom-bust, we will now examine significant elements that are considerably worse. These weaknesses are just as real as the strengths—and arguably more consequential. Many of the underlying risks resemble prior bubbles, but several are new in scale, complexity, or opacity. These vulnerabilities offset the advantages described earlier and, taken together, make the bubble scenario the most convincing.

The factors that are distinctly worse today are, among others: hidden leverage, circular financing, private-market exuberance, ideological narratives, market-wide overvaluation, macroeconomic dependence on AI spending, and the erosion of regulatory guardrails.

1. Reinforcing Loops Among Hyperscalers

Investors like Mark Tilbury have commented on the AI financing diagram that Bloomberg published a few weeks ago, which raised alarms about the soundness of the cashflows of large AI companies. The chart shows how companies such as Microsoft, Nvidia, OpenAI, and Oracle form a tightly intertwined money-sharing ecosystem. For instance:

• Big tech firms invest billions into AI model developers.
• Those developers purchase vast amounts of compute from the same big tech firms.
• Semiconductor providers sell hardware to hyperscalers, who also fund the AI startups that then buy more hardware.

Tilbury argues that these dynamics can inflate revenue and valuation signals because they create economic activity that remains largely inside the same closed loop.

This is not fraud. It is simply a structure that makes the AI economy appear as if more independent capital is entering the system than is truly the case. This counteracts the alleged cash flow strength that is pointed by J.P. Morgan or Lamont and which is considered one of the greatest strengths of the current AI market.

1. The Debt That Sits in the Shadows

One of the most important structural differences between today and 1999 is the type—and visibility—of the debt supporting the technology buildout. AI infrastructure requires enormous capital: data centers, power generation, energy grids, long-term chip procurement agreements, and specialized networking.

Many of these commitments do not appear in traditional balance-sheet debt figures, but they are being performed by highly leveraged firms, most of the time operating outside the loop of the big companies that are pushing the S&P 500 up. Contrary to the dot-com bubble, it is not the actual AI companies that people invest in on the market the leverage ones, but key providers which no one pays attention to. In any case, the magnitude and extension of the leverage is as bad or even worse taken globally, than internet boom companies.

Andrew Ross Sorkin, famous author of “Too Big to Fail” and “1929: The Inside Story of the Greatest Crash in Wall Street History” has warned of another endogenous factor not directly linked to AI but that can worsen the market conditions: the broader financial system is absorbing illiquid, opaque assets at a scale not seen in prior cycles, particularly as private equity and venture capital find their way into retirement accounts. This creates a slower-moving, harder-to-measure form of leverage—one that can amplify fragility without raising immediate red flags.

2. A Heavier Sovereign Backdrop

At the macro level, unlike the late 1990s, the United States also enters the AI boom with substantially higher federal debt. Servicing costs are elevated, fiscal flexibility is constrained, and the Treasury’s ability to intervene aggressively in a downturn is reduced.

Meanwhile, as Forbes analyst Kieran Meadows highlights, margin debt as a percentage of GDP is at its highest level since 1995, suggesting that retail and institutional investors alike are taking on more leverage than at any point since the dot-com era. When both private and public balance sheets are stretched, shocks propagate more quickly.

3. Private Markets Overvaluation

On another topic, Amazon’s CEO, Jeff Bezos, noted in a public appearance that firms with minimal staff and no product are receiving multibillion-dollar valuations. Contrarily to 1999, many of the most extreme valuations are not occurring on public exchanges; they appear in private markets that lack transparency, liquidity, and price discipline.

The Fobes piece adds that the PitchBook Unicorn 30 Index—tracking private giants such as OpenAI and SpaceX—has returned to all-time highs, showing that the supposed “plateau” in private AI valuations companies signaled in Lamont’s piece is not entirely correct.

2. Retail Investors Entering Markets With Higher Leverage and En Masse

Directly related to the previous theme is a major evolution since the dot-com era: the expansion of retail access to private and pre-IPO investments. Evergreen funds, asset tokenization, and other similar structures are permitting retail investors to enter private markets in a way that is historically novel. AI private companies are one of the big beneficiaries of this new development. Historically, private markets soften before public ones, but the never-before-seen retail access to these markets might mean that a bubble may detonate, probably for the first time in history, here.

On the other hand, the ease of leverage on micro investment digital platforms like Robinhood, Accorns, or eToro has facilitated retail leverage. Not only do these investors have access to aggressive leverage tools (options, CFDs, margin accounts), but these platforms regularly provide buttons that make it extremely easy to leverage 2x, 5x, and even getting to the absurd numbers of 30x for certain asset classes. These options are provided in the same style as “one click buying”, and are ludicrously easy for retail investors. Due to this, the barriers to leveraged speculation have fallen dramatically.

These platforms have also made retail investment accessible to ever smaller accounts. In the past, certain minimums were needed to access a broker, but now literally anyone can invest, with the added detail that some platforms also offer access in across the globe. Retail investments are often called “dumb money”, and dumb money is now dominating the market in unprecedented ways. That frictionless environment accelerates boom-and-bust cycles.

3. Social Media as an Engine of Speculation

Something that is not regularly considered when studying the state of the AI investment theme is social media. In the 90s, social media did not exist, and neither did investment influencers who promote all types of investment bets and whose reach is now the entire population of Earth. Stories of overnight wealth spread rapidly, creating the same self-reinforcing psychology that defined Tulip Mania, the South Sea Bubble, and the late 1990s—only now with more reach, and amplified by algorithms designed to maximize engagement and suppress moderated views.

4. Existential Narratives Inviting Exponential Expectations

While the internet narrative was strong, the AI narrative today is more extreme than any past narrative surrounding technology. Many public figures claim that AI will replace millions of jobs, usher in the fourth industrial revolution, or render prior technological leaps insignificant.

Palantir’s Alex Karp, for example, argues that institutions that fail to adapt to AI will be “permanently left behind.” Bezos emphasizes AI’s role as a “horizontal enabling layer” touching every industry. These visions may be directionally correct—but markets tend to overprice truths, and drive prices up before fundamentals catch-up.

5. AI as a Civilizational Imperative

As if that narrative wasn’t enough, some public figures (like the same Alex Karp) frame AI adoption as essential for maintaining Western geopolitical leadership. This elevated tone—“invest in AI or lose the future”—encourages capital allocation that is fueled by deeper-seated fears and motivations (unlinked to financial reality). During the dot-com era, the narrative was optimistic and commercial; today it is also ideological, implying moral or strategic failure if one abstains. This might very well be one of the factors fueling an unprecedented mania for AI.

6. The S&P 500 Has Never Been This Concentrated

It is well known that seven companies (Mag 7) now represent over one-third of the S&P 500. This concentration exceeds levels seen in 1999, when tech dominated but did not exert this degree of index-level influence. Today, the equity market is effectively a leveraged bet on a handful of firms whose valuations and capex plans drive the entire index.

If sentiment turns—for any reason—the feedback loop into the broader market will be immediate.

7. AI Spending Is Propping Up the Economy

A very concerning piece by Deutsche Bank concludes that, absent massive AI-related capex, the U.S. might already be in recession. This implies that the economy is becoming structurally dependent on the investment decisions of a few technology giants.

This is a critical difference from the dot-com era: then, tech exuberance inflated markets but did not serve as the macroeconomic engine keeping GDP growth afloat. This also makes the market sensitivity to any change in confidence or stability much greater, and its effects, exponential.

8. A Market Without New Equity Issuance

Owen Lamont observed that the U.S. stock market today shows negative net issuance—roughly -0.9%—meaning firms are repurchasing stock rather than issuing it. During the dot-com era, by contrast, issuance surged as companies capitalized on euphoric valuations.

Lamont sees the lack of issuance as proof we are not yet in a classic bubble. But it also raises a more uncomfortable point:

If the AI boom is not being financed by new public equity, where is the risk capital coming from?

Often, from the very companies whose valuations depend on AI’s success. This is clearly different from the dot-com era, but equally momentous. That circularity, mentioned in a point above, can magnify fragility if expectations shift.

9. System-Wide Overvaluation

The AI boom is unfolding amid a market environment that is broadly overvalued, not just in technology. As the Forbes piece observes:

• The Shiller CAPE ratio is near 40, approaching dot-com extremes.
• Gold, bitcoin, coffee, and junk bonds trade near record highs.
• Housing affordability is severely stretched, with prices at 5× median income.
• High-yield spreads show investors demanding unusually low compensation for high credit risk (junk bonds).

In 1999, excess was concentrated. Today, it is pervasive—and therefore more dangerous. Even if AI valuations were somewhat rational, the broader backdrop would still make downside risk soar.

10. AI Spending Is Orders of Magnitude Larger Than Internet Investment in 1999

Tilbury estimates that the “Magnificent Seven” spent $330 billion on AI in 2024 alone. During the dot-com era, investment in the entire telecommunications industry, including investments made for cellular phones and others, only amounted to 81 billion. The percentage of that spending that went directly to the internet was only a fraction of that figure.

The scale matters: the bigger the buildout, the more painful an unwind becomes if monetization disappoints.

11. Deregulation and the Weakening of Financial Guardrails

Finally, Sorkin warns that some of the regulatory protections introduced after prior crises are eroding. Disclosure standards have loosened, risky private investments are increasingly accessible to non-professional investors, and leverage is easier to obtain.

This erosion is occurring at the very moment speculative behavior is rising—making the possibility of a multi-causal crisis more prevalent.

III. Why Some Things Are the Same

Even with all the structural differences between the AI boom and the dot-com bubble, many underlying behaviors and dynamics remain strikingly similar. These repeating patterns matter because financial history shows that bubbles do not depend on technology. They depend on human psychology, incentives, and the misperception of risk.

There are recurring themes linking today’s AI enthusiasm with the market cycles of the past—particularly the late 1990s. Despite the improved balance sheets and the genuine utility of the technology, the behavioral drivers of bubbles have not changed.

1. Investing Without Clear Monetization—A Familiar Red Flag

One of the most defining features of the late 1990s was the willingness of investors to fund companies whose business models were aspirational rather than proven. That pattern is visible today, especially among leading-edge AI firms whose profitability remains uncertain.

Mark Tilbury highlights that many AI companies—OpenAI being the most frequently cited example—operate without a clearly established path to durable profits. Revenues may be rising, but they rely heavily on unusually favorable funding structures and customer concentration. Tilbury warns that soaring valuations paired with weak or untested monetization are classic ingredients of market exuberance.

This echoes the dot-com dynamic: transformative technology is not a guarantee of near-term or even medium-term returns. The story can be true while the valuation is wrong.

2. “Keeping Up” Spending and an Investing Bubble

A major psychological similarity between 1999 and today is the pressure on companies and investors to participate simply because others are doing so.

Corporate leaders today often justify enormous AI budgets not on quantified return expectations, but on the strategic necessity of keeping up. In a honest interview with Access, Mark Zuckerberg mentioned that he rather misspend a couple hundred billion dollars than loose the AI race, that an AI bubble “is quite possible”, because “most other major infrastructure build-ups in history”, including the dot-com bubble, “include a phase were the infrastructure gets build out, people take on too much debt, a lot of the companies end of going out of business, and then the assets get distressed.”

He also said that “there is definitely a possibility… based on past large infrastructure build-outs, and how they led to bubbles, that something like that would happen here”. This needs no comments.

Jeff Bezos, similarly, commented on AI, saying that it might not be a financial bubble, but rather a, industrial one. In his view, AI today shows signs of an industrial bubble: legitimate innovation paired with indiscriminate investment. Still, Bezos clearly states that industrial bubbles, while positive for society as a whole, burst as any other bubble, making investors the big losers.

3. Retail FOMO Is Back

Digital trading platforms—with easy leverage, TikTok-style financial commentary, and frictionless order execution—have made speculative behavior even more accessible.

The presence of aggressive leverage tools (options, CFDs, margin accounts) on retail platforms amplifies this effect. This differs from the dot-com era only in accessibility; the underlying behavior is the same.

4. Overconfidence and Narrative-Driven Exuberance

Jeff Bezos notes that in periods of technological enthusiasm, “virtually every startup and experiment gets funded.” He argues that AI is experiencing exactly this dynamic: both promising and weak ideas are attracting capital simply because they are labeled “AI.”

This indiscriminate funding mirrors the dot-com era, where capital chased “internet” in the same way it now chases “AI.” Investors struggle to separate the transformative from the trivial in the early stages of a new technological development and the market rewards optimism.

Two psychological pillars of all bubbles—overconfidence and narrative dominance—are clearly present:

• Overconfidence that AI will rewrite economic laws.
• Overconfidence that adoption curves will be straight lines.
• Overconfidence that early market leaders will inevitably become long-term monopolies.

Analysts used the same arguments to justify extreme valuations for Cisco, AOL, and Qualcomm in 1999.

5. The Eternal Promise That “This Time Is Different”

Forbes’ video emphasizes a recurring myth of financial history noticed by economist John Kenneth Galbraith: the belief that the present innovation is so profound that old valuation principles no longer apply. In his words, every generation falls for the idea that past bubbles are irrelevant because new technology changes everything.

In the late 1990s, investors believed that the internet would abolish the business cycle.
Today, some claim AI represents a secular mega-trend that renders cyclical thinking obsolete.

Franklin Templeton’s Michael Browne adds nuance—he believes AI could become essential infrastructure akin to electricity. Yet even he acknowledges this does not immunize markets from mispricing. Technologies can be transformative while their investment phases are speculative.

The mindset that “AI is too big to fail” is nearly identical to the dot-com belief that the internet made recessions and valuation resets unlikely.

6. The Market Being Propped Up by a Single Theme

Another parallel is the degree to which a single narrative disproportionately supports the broader equity market. In 1999, it was the internet; today, it is AI.

Sorkin has pointed to this similarity explicitly, warning that the market’s dependence on a narrow theme increases vulnerability if confidence erodes. As during the dot-com era, the capex cycle of a handful of companies mantains the performance of entire economy.

7. Overvaluation Relative to Fundamentals

Finally, Sorkin has also warned that the pattern of sharp price appreciation unmoored from fundamentals resembles all financial crisis periods. While technology might be even more transformative today than it was 25 years ago, the relationship between valuations and underlying earnings remains stretched.

This mirrors 1999, when investors correctly identified many companies as transformative, but incorrectly priced as if transformation equated to immediate, massive cash flow.

AI companies today face the same challenge: even if the technology becomes indispensable, valuations may already be assuming outcomes that are years—if not decades—away.

Conclusion: A Better Technology Does Not Guarantee a Better Market Outcome

The rise of artificial intelligence marks one of the most significant technological breakthroughs of our time. On this point, nearly every serious analyst agrees.

But markets are not priced on technological truth alone. They are priced on expectations—how fast the truth will materialize, how widely, and at what margin. And this is where today’s environment becomes precarious. The strengths of the AI economy are offset by risks that are deeper, more opaque, or more systemic than those seen before the dot-com crash.

On the “bright” side, we do not see the same explosion of equity issuance that defined past bubbles. Firms are net repurchasers, not capital absorbers. Balance sheets are stronger. Profitability is higher. The largest AI investors generate enormous free cash flow and can fund their ambitions internally.

Yet on the “worse” side, several dynamics raise questions that did not exist in 1999:

• hidden leverage embedded in private markets and multi-year infrastructure commitments;
• circular financing among hyperscalers and model developers that can inflate revenue signals.
• unprecedented retail participation in speculative assets;
• ideological narratives that encourage investors to see AI exposure not as a decision but as a civic duty;
• hyperconcentration of the equity market in seven firms that now make up 33% of the S&P 500;
• valuations stretched across all asset classes, not just technology;
• and an economy increasingly reliant on AI capex to avoid recession.

And then there are the features that are simply the same: unclear monetization pathways, investors chasing momentum because prices are rising, firms spending “to keep up,” and the recurring belief that a new technology invalidates historical valuation frameworks. These behavioral constants were present in 1999 and remain present today, even if the underlying technology has advanced dramatically.

What emerges from this landscape is a paradox:

Some of AI’s commercial foundations are stronger than the internet’s in 1999, but the financial and provider ecosystem surrounding it is more fragile.

This does not imply an imminent crash. Bubbles rarely end on schedule. Bubbles often leave behind extraordinary innovations—even when investors lose money. If AI follows that pattern, society may benefit enormously while AI heavily reliant portfolios experience severe downturns.

For long-term investors—particularly families and entrepreneurs focused on capital preservation—the implication is clear:

AI’s promise is real, but its current market pricing may be assuming too straight a line between innovation and profit. The safest path might be to get out before it is too late and take advantage on the long-term returns on AI to buy de dip.

The lesson from the dot-com era is not that transformative technologies should be avoided. It is that transformative technologies do not protect investors from overpaying, especially in the inicial stages.

AI may indeed reshape the economic landscape. But the market built around it still behaves like a market—with all the exuberance, cyclicality, and vulnerability that entails. As history shows, the challenge is not identifying the winning technology. It is navigating the cycle that surrounds it.

Periods of technological transformation often reward innovation while punishing indiscriminate exposure. Tiempo Capital helps high-net-worth families and individuals evaluate concentrated themes like AI within a broader investment management and family office framework—balancing opportunity, valuation risk, liquidity, and long-term objectives. Learn more about our approach to portfolio construction and contact us to discuss how your exposure to AI fits within your overall strategy.

This material is for informational purposes only and does not constitute financial, legal, tax, or investment advice. All opinions, analyses, or strategies discussed are general in nature and may not be appropriate for all individuals or situations. Readers are encouraged to consult their own advisors regarding their specific circumstances. Investments involve risk, including the potential loss of principal, and past performance is not indicative of future results.