Is an AI Bubble About to Burst? Fed Chair Warns of Risks: Could We See a Repeat of the 2000 Dot-Com Crash? Will Bitcoin Be Dragged Down Too?

Zusammenfassung:​In the second half of this year, an increasing number of prominent figures have been sounding the same alarm: the AI bubble could be on the verge of bursting.

In the second half of this year, an increasing number of prominent figures have been sounding the same alarm: the AI bubble could be on the verge of bursting.

On Monday this week, Fed Chair Kevin Warsh even publicly stated that policymakers need to prepare for a potential AI bubble collapse.

Ray Dalio, founder of Bridgewater Associates, has gone further, describing AI as a classic bubble fueled by massive debt financing and approaching a critical breaking point. He has compared current market indicators with those seen before the 1929 crash and the 2000 dot-com bust.

Ken Griffin, founder of Citadel, has also warned that the AI boom has almost certainly developed into a capital bubble, drawing parallels with the dot-com era.

Michael Burry, the investor portrayed as the inspiration for The Big Short, has taken an even more aggressive position. He has described the AI bubble as enormous and ultimately doomed to burst, and has reportedly positioned himself with bearish options linked to AI-related stocks.

One of the most dangerous aspects of today's AI bubble lies in guarantee agreements.

On the surface, AI companies are conducting legitimate business: selling chips, providing cloud services, and building data centers. But beneath the surface, an increasing amount of capital is no longer being supported by companies' own operating cash flows. Instead, it relies on borrowing, long-term contracts, guarantees, and circular financing arrangements in which companies buy from one another.

A stock market decline can be absorbed gradually. But when the underlying credit chain breaks, the damage can spread much faster.

The real question is not whether the dot-com bubble of 2000 will repeat itself exactly. It is whether today's AI bubble combines the market concentration of the 2000 tech bubble with the financial engineering and credit instruments that amplified the 2008 financial crisis.

Consider this analogy:

In 2000, companies raised equity capital and used it to build fiber-optic networks that nobody needed. When those investments failed, shareholders took the losses.

Today, the situation is different. Companies are using long-term contracts and guarantees to borrow money and build data centers. Before construction is even complete, stacks of guarantee agreements may already have been signed.

The former was primarily an equity bubble. The latter could evolve into a credit bubble.

So, has the AI bubble really reached a fragile breaking point? What would happen if it burst? How would Bitcoin be affected? And, most importantly, what should investors do?

AI Bubble vs. Dot-Com Bubble: A More Complex Bubble with Potentially Greater Destructive Power

Whenever we face a major issue, we should first ask whether it actually exists before trying to understand why.

So, is the AI bubble real?

The answer is yes. But that does not mean AI itself is a bubble.

AI technology is real, and so is the demand for it. NVIDIA continues to report strong growth, with quarterly data center revenue exceeding $107 billion. Microsoft, Google, Amazon, and Meta are still investing aggressively in data centers. AI is already being deployed in programming, customer service, healthcare, finance, manufacturing, and other industries.

The problem is that market valuations have moved far too quickly.

Goldman Sachs has highlighted some striking figures: AI accounts for only 21% of US GDP growth, yet contributes 62% of S&P 500 earnings growth and, at the market's peak, 92% of the increase in US stock market capitalization.

Put simply, AI is still relatively small in the real economy but has become a giant in the stock market.

Investors have already priced into today's share prices profits that may not materialize for many years.

But inflated valuations are not the most dangerous part of this story.

The real question is: Where is all the money coming from?

The 2000 dot-com bubble was primarily driven by equity financing. Startups burned through venture capital and IPO proceeds to build fiber-optic networks that nobody used. When those investments failed, shareholders absorbed the losses. Banks were not as deeply entangled in the process.

Today's situation is fundamentally different.

AI investment has become so enormous that even the cash flows of major technology companies are struggling to keep up. As a result, companies are increasingly relying on:

Corporate bond issuance

Bank loans

Long-term leasing arrangements

Private credit

Special Purpose Vehicles (SPVs)

Several figures in Federal Reserve reports deserve particular attention:

Approximately $680 billion in direct AI-related debt accumulated during the first ten months of 2026.

Off-balance-sheet leases, guarantees, and contractual commitments total approximately $3 trillion.

Debt can account for 85%–90% of SPV asset values, meaning that a decline of just over 10% in asset prices could wipe out the equity cushion.

The IMF has also warned that circular financing in the AI sector has exceeded $1 trillion, with approximately two-thirds of that amount financed through debt.

So, the AI bubble is real. But it is not simply another valuation bubble like the one in 2000.

It is a hybrid bubble combining genuine technology, real earnings, excessive capital expenditure, and hidden credit leverage.

What appears to be a stock market problem on the surface may actually be a much deeper credit problem underneath.

This is why the potential damage from an AI bubble collapse could exceed that of the dot-com crash.

Today's US stock market is more concentrated. The five largest AI-related stocks now account for approximately 28% of the S&P 500, exceeding the concentration levels seen during the dot-com bubble.

More importantly, the nature of this concentration has changed.

In 2000, market concentration centered on companies such as Cisco, Microsoft, and Intel — the companies building the infrastructure for the internet.

Today, concentration is centered on NVIDIA, Microsoft, Alphabet, Amazon, and Meta — companies that are simultaneously building the infrastructure, collecting tolls, and operating the marketplaces.

If the AI narrative starts to unravel, the impact will not be confined to one segment of the technology supply chain. It could undermine the valuation foundations of the entire stock market, triggering a global financial shock.

There is another critical difference: credit contraction can be much faster and more destructive than valuation contraction.

A valuation decline is like slowly squeezing air out of a bubble.

A credit contraction is like suddenly shutting off the oxygen supply.

If falling AI stock prices trigger credit events at the SPV and project-financing levels, the damage will go far beyond a normal valuation correction.

It could trigger a vicious cycle:

Collateral values fall → lenders demand additional collateral → borrowers are forced to sell assets → asset prices fall further → collateral values deteriorate even more.

When SPV debt accounts for 85%–90% of asset values, a decline of just over 10% could wipe out the equity cushion.

The higher the leverage, the smaller the margin for error and the faster a collapse can unfold.

The five largest AI-related stocks accounting for 28% of the S&P 500 also means that passive index funds, pension funds, and retirement accounts are heavily exposed to the same group of companies.

Under a baseline scenario in which the S&P 500 falls 25%, the value of US household equity holdings through direct investments, mutual funds, and retirement accounts could shrink by approximately $19 trillion.

Many investors believe Microsoft, Google, Amazon, and Meta can withstand a crisis because they have real earnings and substantial cash flows.

That assumption needs to be treated with caution.

Over the 12 months through March 2026, the four major US cloud providers collectively spent approximately $464 billion on capital expenditure, while their free cash flow continued to shrink.

S&P Global Ratings expects six major hyperscalers could experience negative free operating cash flow in 2026 and 2027, with cash flow potentially not recovering until 2029.

The problem is not that these giants are unprofitable. It is that their earnings are no longer growing fast enough to cover their spending.

They are increasingly borrowing money to finance their investments.

When even the largest technology companies must rely on debt and off-balance-sheet commitments to sustain capital expenditure, it suggests that the industry's ability to generate enough cash internally is not as strong as it appears.

The Bank for International Settlements (BIS) has characterized the current AI investment boom through a combination of high capital expenditure, debt financing, SPVs, circular financing, and concentrated credit exposure.

It has drawn comparisons with historical investment cycles involving canals, railways, and the internet.

This is not simply a repeat of 2000.

It is the market concentration of 2000 combined with the financing techniques of 2008.

If the AI Bubble Bursts: Five Dominoes Could Fall in Succession

Unlike the dot-com crash, which primarily involved a stock market and telecommunications industry collapse, an AI bubble bursting could trigger a chain reaction across five interconnected areas:

Credit → Wealth → The Real Economy → Global Trade → Cryptocurrency.

The transmission mechanism is much more extensive today.

Domino No. 1: The US Stock Market Shifts from Valuation Compression to Credit Contraction

Jason Furman, a former White House economic adviser, has outlined a baseline scenario in which the S&P 500 falls 25%, a more pessimistic scenario involving a 35% decline, and an extreme scenario involving a 50% decline.

Under the baseline assumption of a 25% drop, US household equity wealth held through direct investments, mutual funds, and retirement accounts could shrink by approximately $19 trillion.

A 25% decline alone would not necessarily constitute a systemic financial crisis.

Between 2000 and 2002, the Nasdaq fell by a cumulative 78%, yet the US economy only entered a mild recession in 2001.

What really determines the scale of the damage is the transmission mechanism.

If the decline begins with AI-related stocks and subsequently triggers credit events at the SPV and project-financing levels, the impact will no longer be limited to a normal valuation correction.

Instead, it could set off a vicious cycle:

Falling collateral values → Margin calls → Forced selling → Further declines in collateral values.

Goldman Sachs has already pointed out that the wave of AI-related bond issuance is putting upward pressure on credit spreads. It also expects earnings growth to be cut in half, from 36% to 11%.

This is precisely why financial markets may be more fragile than the broader economy.

AI's direct contribution to GDP remains relatively limited, but its integration into stock market valuations and credit networks is far greater than its actual economic weight.

Domino No. 2: Global Trade Faces a Dangerous Single Point of Dependence

AI capital expenditure directly contributed approximately 0.3–0.4 percentage points to US GDP growth in 2025. That figure could rise to 0.4–0.6 percentage points in 2026.

These numbers may not appear particularly significant at first glance.

However, AI investment has an enormous marginal influence on global trade. As much as 80% of recent growth in global exports has been concentrated in AI-related goods.

This means that if AI capital expenditure peaks and begins to contract, the damage will extend far beyond US technology stocks.

Economies heavily dependent on exporting AI hardware to the United States, particularly South Korea and Taiwan, could also face substantial pressure.

Goldman Sachs expects the year-over-year change in AI investment as a share of GDP to peak in the first half of 2027.

Even if the absolute amount of AI investment continues to increase, a peak in its growth rate represents an important turning point in the investment cycle.

That second-order inflection point alone could be enough to trigger a meaningful market reaction.

The IMF's worst-case scenario assumes that if AI fails to meet market expectations, US stocks could plunge 20% in 2026, while global output could be 1.1% below the baseline forecast by 2028.

Domino No. 3: Bitcoin Is Not a Safe Haven. It Is the High-Beta End of the Same Leverage Trade

The narrative that “capital will rotate into crypto when the AI bubble bursts” has already been disproven by November 2026.

Bitcoin fell 16% that month, moving in tandem with NVIDIA and AI-related ETFs, which had retreated approximately 30% from their highs. US spot Bitcoin ETFs recorded net outflows of $2.35 billion over just two weeks.

The transmission mechanism is clear and unforgiving.

AI stocks and Bitcoin are both long-duration, liquidity-sensitive risk assets. For years, they have been held within many of the same leveraged and momentum-driven investment portfolios.

When volatility rises on the AI side and margin calls begin, institutional investors typically respond by reducing overall exposure and raising cash. They do not simply rotate from one high-beta asset into another.

The 30-day rolling correlation between BTC and the Nasdaq 100 has climbed to 0.68. During periods of market stress, Bitcoin has increasingly behaved like a shadow version of the Nasdaq.

Arthur Hayes has proposed a longer-term framework for understanding this relationship:

During the initial stages of a crisis, Bitcoin will fall alongside other risk assets as investors deleverage. However, if a credit crisis eventually forces the Federal Reserve to inject massive amounts of liquidity into the financial system, the resulting monetary expansion could ultimately benefit Bitcoin.

The logic behind this framework makes sense, but it relies on one critical assumption: there must first be a sufficiently severe deleveraging event, followed by a timely and aggressive policy response.

Domino No. 4: The Real Economy and the Wealth Effect

A decline in stock market wealth can affect consumer spending through the wealth effect, while a reduction in corporate capital expenditure can weigh on investment and employment.

Once AI capital expenditure growth peaks, the entire supply chain — including data center construction, power equipment, servers, and networking hardware — could begin to cool.

The economy will not necessarily grind to a halt overnight, but one of its important marginal growth drivers will weaken significantly.

Domino No. 5: The Credit Markets

The areas that deserve the closest attention are SPVs, private credit, banks, and bond markets.

If off-balance-sheet commitments begin creating financial pressure, credit spreads could widen and financing costs could rise. This would further constrain AI investment, creating a negative feedback loop.

An AI bubble collapse may therefore not look like one clean, sudden valuation crash.

Instead, it could unfold gradually through several stages:

Earnings expectations deteriorate → Off-balance-sheet commitments come under pressure → Credit spreads widen → Financing becomes more expensive → AI investment contracts further.

This gradual transmission from equity valuations into the credit system could ultimately prove more dangerous than a conventional stock market correction.

What Should Ordinary Investors Do? Shift from Betting on Direction to Managing Vulnerability

In the current environment, the greatest danger is not misunderstanding the long-term value of AI.

It is taking on short-term vulnerabilities that your portfolio cannot withstand, particularly when you are operating on the wrong investment time horizon.

First, distinguish between companies that use AI to save money and companies that borrow money to build AI infrastructure.

Large non-technology companies using AI to improve efficiency and reduce operating costs have AI exposure on the cost side of their businesses. They may be more resilient when credit conditions tighten.

By contrast, companies aggressively borrowing to build data centers, particularly those with fragile cash flows, could become the first dominoes to fall.

1. The HALO strategy does not mean abandoning technology.

HALO (Heavy Assets, Low Obsolescence) focuses on tangible, capital-intensive assets with relatively low technological obsolescence.

Examples include electricity grids, energy infrastructure, transportation networks, and selected commodities.

These assets may offer portfolio value under both scenarios: an impaired AI investment narrative and continued AI adoption.

After all, AI development itself depends on electricity and physical infrastructure.

This is not simply about seeking safety. It is about owning assets that the AI industry cannot operate without.

2. Within AI, avoid buying everything. Focus on quality and execution.

Goldman's trading desk has explicitly stated that this is no longer a market in which investors can simply “buy everything AI-related.”

Valuations of hyperscale cloud providers have fallen to approximately 19.7 times earnings, below those of traditional defensive sectors such as consumer staples.

If the bubble bursts, the first assets to be sold may not necessarily be the most important AI companies.

They could instead be the peripheral players with the highest leverage, weakest cash flows, and business models least supported by verifiable earnings.

3. For crypto investors, stop treating Bitcoin as a diversification tool. Manage positions according to actual correlations.

Until the leverage embedded in the AI financing chain has been substantially unwound or absorbed, the idea that Bitcoin is a safe haven independent of technology stock risk is not supported by market evidence.

If you still believe in Bitcoin's long-term investment thesis, keep your position at a level that allows you to withstand a severe deleveraging event in which Bitcoin falls alongside AI stocks, without being forced to sell.

4. Most importantly, preserve cash and short-duration assets. This is not about missing out. It is about retaining your options.

Bank of America has recommended reducing risk exposure, shifting toward defensive sectors, duration assets, and the US dollar.

When markets are approaching a potential bubble collapse, the most important thing is not to predict the exact moment it will burst.

It is to ensure that when the collapse actually happens, your portfolio still contains assets that you are not forced to liquidate, along with enough flexibility to take advantage of new opportunities.

Conclusion: Remember These Three Things

First: The cracks in the AI bubble are not primarily in P/E ratios. They are in the guarantee agreements.

Valuation bubbles can sometimes be absorbed through future earnings growth. A broken credit chain, however, can transmit financial stress much faster.

Second: This is not a replay of 2000. It is the market concentration of the dot-com bubble combined with the financing techniques of the 2008 financial crisis.

Stock market risks are increasingly spreading into the credit markets.

Third: Manage vulnerability before betting on direction.

The technological revolution is real, but market pricing can be dangerously detached from reality.

Before AI proves that it can fundamentally transform the economy, the market may first undergo a brutal repricing that wipes out the leverage and financial commitments built on the assumption that “this time is different.”

Haftungsausschluss

Die Ansichten in diesem Artikel stellen nur die persönlichen Ansichten des Autors dar und stellen keine Anlageberatung der Plattform dar. Diese Plattform übernimmt keine Garantie für die Richtigkeit, Vollständigkeit und Aktualität der Artikelinformationen und haftet auch nicht für Verluste, die durch die Nutzung oder das Vertrauen der Artikelinformationen verursacht werden.
Letzter Artikel

US-Richter spricht Brüder in Ethereum-MEV-Fall über 25 Millionen US-Dollar nach Fehlprozess frei