AI Crash Risk: How Much Wealth Could a Bubble Burst Wipe Out?
Quick Summary
From $20 to $40 trillion in potential losses — analysts are sizing up AI crash risk. Here's what the data says and why diversification may not protect you.
In This Article
The Two Stocks Moving the Entire Global Market
In a single month, two memory chip companies — Micron and SK Hynix — generated 17% of the entire global stock market's return. Not 17% of the chip sector. Not 17% of US equities. Seventeen percent of every listed company, in every country, added together.
Those two stocks represent roughly 1% of the MSCI All Country World Index, which tracks thousands of companies across dozens of markets. For every dollar the global market rose that month, 17 cents came from just those two names. That single data point, sourced from Aadian Asset Management, crystallises the concentration risk at the heart of the AI trade — and it raises a question every serious investor should be stress-testing right now: if two chip stocks can drag the entire planet's equity markets upward, what happens to your portfolio on the way back down?
This is not a prediction that the AI bubble will burst. Plenty of credible, well-informed investors believe AI is a genuine technological revolution that will more than justify the capital being poured into it. The optimists are the majority, and their conviction is the reason this boom exists at all. But sizing up downside risk is not pessimism — it's discipline. And the downside risk here, by almost every credible estimate, is measured in the tens of trillions of dollars.
Why Diversification Is Failing AI Crash Risk Management
Diversification is the closest thing finance has to a free lunch. Spread your money across uncorrelated assets, and when one blows up, the others cushion the fall. The problem is that the AI trade has quietly colonised almost every corner of the market — including the corners where cautious investors thought they were hiding.
Consider the Russell 2000, the small-cap index that most investors treat as exposure to ordinary American businesses. In a recent strong first half, 16 of its 50 best performers were semiconductor and chip equipment firms. Companies like MaxLinear and Aerojet Rocketdyne equivalents in the testing and cabling space surged 250–380%. You didn't dodge the AI boom by buying small caps. You bought the companies selling it cables and testing gear.
The Russell 1000 Value Index tells an even more instructive story. Value investing appeared to stage a comeback, with the index outperforming growth by a meaningful margin in a recent period. The reason, however, had nothing to do with stock-picking skill. The annual index rebalance moved high-flying semiconductor names — Micron, AMD, Western Digital — out of value and into growth at almost exactly their peak. Then it absorbed Amazon, Apple, and Microsoft at near their lows. A calendar reminder outperformed active management. And the result? A "sensible value portfolio" now stuffed with three of the biggest tech companies on Earth — effectively the AI trade wearing a false moustache.
The conclusion is uncomfortable: for most investors, getting out of the AI trade is harder than it looks. The sector doesn't stay in its lane. It has spread into utilities (data centres need enormous amounts of power), real estate (server warehouses and land), construction (a data centre building boom is underway across the US), and even consumer spending, as AI-related IPO wealth disperses into the broader economy through home purchases, luxury goods, and private jet charters.
What the Numbers Actually Say: Sizing the Potential AI Crash
So how much money is actually at stake? Three credible estimates, three different methodologies — and they all land in the same ballpark.
Economist Dean Baker runs what he calls the AI Bubble Monitor. His calculation: the total US stock market sits at roughly $80 trillion. If price-to-earnings ratios simply drifted back to their long-run historical average — not a crash, just mean reversion — approximately $40 trillion of stock market wealth would be erased. That averages out to nearly $300,000 per US household, though that figure is heavily distorted by concentration: the wealthiest 10% of Americans own roughly 90% of shares, so the median household would feel far less.
Gita Gopinath, former chief economist at the IMF, estimates a dot-com-style correction today would destroy roughly $20 trillion of American wealth, plus an additional $15 trillion held by foreign investors. That $20 trillion figure represents approximately 70% of US GDP.
Oliver Wyman consultants ran their own model and landed near $33 trillion in value destruction — more than the entire US economy produces in a year.
For perspective: the actual dot-com bust destroyed around $6 trillion in equity value. The mainstream estimates for an AI crash run five to six times larger. The order of magnitude is consistent across all three approaches, even if the precise figures differ.
The Wealth Effect: Why a Stock Crash Hits Main Street Now
Here's the structural shift that makes this moment different from 2000. According to Goldman Sachs and Federal Reserve data, stocks overtook real estate as the single biggest component of American household wealth — for the first time since the Second World War.
For most of modern history, the average family's net worth lived in their house. A stock market crash was, relatively speaking, a rich person's problem. That is no longer true. The stock market is now where the median household's wealth actually resides, which means a crash today has a broader reach into middle-income America than any previous tech correction.
The mechanism is the wealth effect. Research suggests that for every $100 of paper stock wealth, people spend approximately $3 in the real economy. That sounds modest until you multiply it across tens of trillions in notional losses. The arithmetic runs cleanly in reverse: if $30 trillion of household wealth evaporates, consumer spending contracts sharply. The new kitchen gets cancelled. The contractor loses revenue. The car dealership has a bad quarter and lays someone off. The damage cascades through the real economy without a single bank needing to fail — it just requires people to feel meaningfully poorer.
Harvard economist Jason Furman calculated that AI-related infrastructure spending accounted for something like 90% of US economic growth in one recent six-month period. More conservative estimates put the figure around 25%. Either way, a significant share of recent economic growth is, in effect, data centre construction. The electricians, concrete crews, transformer manufacturers, cable installers, and truck drivers who depend on that spending don't own Nvidia shares. They don't need to. If AI capital expenditure slows, their income slows with it.
Nobel laureate Joseph Stiglitz has noted the additional complication: a market correction and AI-driven labour displacement could hit simultaneously. Households facing falling savings at the same moment their jobs become less secure is not a theoretical risk — it's a plausible scenario that policymakers are not visibly prepared for.
The $3 Trillion Iceberg: What Big Tech's Balance Sheets Don't Show
The banking system is in substantially better shape than it was in 2008. Capital ratios are higher, liquidity buffers are real, and annual stress tests have genuine teeth. A Lehman-style cascade through the financial system is not the most likely vector for pain here.
But when regulators make one part of the financial system safer, risk doesn't disappear — it migrates. A Wall Street Journal analysis found that the major tech companies' reported capital expenditure on AI infrastructure — data centres, chips, computing capacity — totalled roughly $600 billion over a recent 12-month period. Reading the footnotes of their SEC filings tells a different story. The Journal identified approximately $3 trillion of additional AI commitments that don't appear on the balance sheet: long-term data centre leases, locked-in chip purchase agreements, energy contracts. Alphabet alone carries over $800 billion of this type of off-balance-sheet obligation.
Analysts have labelled this the AI spending iceberg. None of it is hidden — it's all disclosed in the filings, for anyone who reads footnotes. But it means the true scale of what these companies are committed to spending is roughly five times larger than the headline capex figures suggest. And every dollar of it rests on the assumption that AI revenues will materialise to cover it.
The credit markets have begun to take notice. The cost of insuring big tech debt against default has recently hit record highs. Meanwhile, private credit — the lightly regulated, $2–3 trillion market of funds that lend directly to companies outside the traditional banking system — has seen troubled loans at the 20 largest listed funds climb to their highest level since 2017. Fitch reported private credit defaults hit a record in a recent month. One major fund disclosed that 7% of its entire loan book was in distress. These are not systemic warnings yet, but they are early signals worth monitoring.
What Investors Should Actually Be Thinking About
None of this is an argument for selling everything and sitting in cash. The bull case for AI is serious, well-funded, and backed by genuine technological progress. The companies investing at scale believe the payoff will be enormous — and they may well be right.
But a few practical considerations are worth internalising:
- Concentration is higher than most portfolios reveal. If you hold a broad index fund, a value fund, and a small-cap fund, you may have more AI exposure than you think — often overlapping in the same names.
- The wealth effect is now a mainstream risk, not a tail risk. With stocks surpassing real estate as the dominant household wealth vehicle, a market correction has real-economy consequences that didn't apply in earlier tech cycles.
- Off-balance-sheet commitments matter. The $3 trillion in undisclosed tech spending obligations is a legitimate analytical data point. Investors and analysts who only read headline capex figures are working with incomplete information.
- Private credit stress is worth watching. It's not 2008, but risk doesn't vanish — it relocates. Rising default rates in private credit funds that have lent heavily to AI and software companies is a canary worth monitoring.
- Diversification requires active scrutiny. Mechanical index rebalancing can inadvertently concentrate exposure. Understanding what's actually inside your funds — not just the label on the tin — is more important now than it was five years ago.
The scale of potential losses, if the bear case proves correct, runs between $20 trillion and $40 trillion by credible estimates. That is not a number to be dismissed as abstract. It is, depending on whose model you use, somewhere between the entire annual output of the US economy and twice that figure — in paper wealth that could, under adverse conditions, simply stop existing.
Free Weekly Newsletter
Enjoying this guide?
Get the best articles like this one delivered to your inbox every week. No spam.
The party may well continue for years. But knowing the size of the tab is basic financial hygiene.
Frequently Asked Questions
How much wealth could an AI market crash realistically destroy?
Estimates from credible sources vary but cluster in the tens of trillions. Economist Dean Baker suggests $40 trillion if US price-to-earnings ratios revert to historical norms. Former IMF chief economist Gita Gopinath estimates roughly $20 trillion in US wealth destruction, plus $15 trillion held by foreign investors. Oliver Wyman consultants modelled approximately $33 trillion in losses. For context, the dot-com crash destroyed around $6 trillion — making these estimates five to six times larger in scale.
Is diversification enough to protect against an AI bubble crash?
Standard diversification is less protective than many investors assume in this cycle. The AI trade has spread into small-cap stocks (through chip equipment and testing companies), value indices (via semiconductor stocks added during rebalances), utilities, real estate, and construction. Investors who believe they've avoided the AI trade may find significant indirect exposure when they look inside their funds. Scrutinising actual holdings — not just fund labels — is advisable.
Why would an AI crash affect people who don't own tech stocks?
Two main channels. First, the wealth effect: research suggests every $100 of stock market losses reduces real consumer spending by around $3. Multiply that by tens of trillions and the consumer economy contracts meaningfully. Second, AI infrastructure spending (data centre construction, chip manufacturing, power infrastructure) now accounts for a significant share of US economic growth — by some estimates, a quarter or more. Electricians, contractors, manufacturers, and logistics workers who depend on that spending are exposed without holding a single share.
What is the AI spending iceberg and why does it matter?
Wall Street Journal analysis found that major tech companies' publicly reported AI capital expenditure — approximately $600 billion over a recent 12-month period — understates their true financial commitments by a factor of roughly five. An additional $3 trillion in obligations (long-term data centre leases, chip purchase agreements, energy contracts) sits in the footnotes of SEC filings rather than on balance sheets. Alphabet alone has over $800 billion in such commitments. These obligations are legally binding and rest on the assumption that AI revenues will eventually materialise to cover them — making them a meaningful risk factor if AI monetisation disappoints.
How does an AI crash differ from the 2008 financial crisis?
The major banks are significantly better capitalised than in 2008, with stronger liquidity buffers and mandatory stress testing. A Lehman-style banking cascade is not the primary risk scenario analysts identify. The more likely transmission mechanism is the wealth effect — falling asset prices reducing consumer spending and real economic activity — combined with a slowdown in AI infrastructure spending that ripples through construction, manufacturing, and utilities. The risk also lives partly in private credit markets, which have grown to $2–3 trillion and have seen rising default rates, though these remain well below systemic crisis levels.
This article is for informational purposes only and does not constitute financial advice. Always consult a qualified financial professional before making investment decisions.
Frequently Asked Questions
The Two Stocks Moving the Entire Global Market
In a single month, two memory chip companies — Micron and SK Hynix — generated 17% of the entire global stock market's return. Not 17% of the chip sector. Not 17% of US equities. Seventeen percent of every listed company, in every country, added together.
Those two stocks represent roughly 1% of the MSCI All Country World Index, which tracks thousands of companies across dozens of markets. For every dollar the global market rose that month, 17 cents came from just those two names. That single data point, sourced from Aadian Asset Management, crystallises the concentration risk at the heart of the AI trade — and it raises a question every serious investor should be stress-testing right now: if two chip stocks can drag the entire planet's equity markets upward, what happens to your portfolio on the way back down?
This is not a prediction that the AI bubble will burst. Plenty of credible, well-informed investors believe AI is a genuine technological revolution that will more than justify the capital being poured into it. The optimists are the majority, and their conviction is the reason this boom exists at all. But sizing up downside risk is not pessimism — it's discipline. And the downside risk here, by almost every credible estimate, is measured in the tens of trillions of dollars.
Why Diversification Is Failing AI Crash Risk Management
Diversification is the closest thing finance has to a free lunch. Spread your money across uncorrelated assets, and when one blows up, the others cushion the fall. The problem is that the AI trade has quietly colonised almost every corner of the market — including the corners where cautious investors thought they were hiding.
Consider the Russell 2000, the small-cap index that most investors treat as exposure to ordinary American businesses. In a recent strong first half, 16 of its 50 best performers were semiconductor and chip equipment firms. Companies like MaxLinear and Aerojet Rocketdyne equivalents in the testing and cabling space surged 250–380%. You didn't dodge the AI boom by buying small caps. You bought the companies selling it cables and testing gear.
The Russell 1000 Value Index tells an even more instructive story. Value investing appeared to stage a comeback, with the index outperforming growth by a meaningful margin in a recent period. The reason, however, had nothing to do with stock-picking skill. The annual index rebalance moved high-flying semiconductor names — Micron, AMD, Western Digital — out of value and into growth at almost exactly their peak. Then it absorbed Amazon, Apple, and Microsoft at near their lows. A calendar reminder outperformed active management. And the result? A "sensible value portfolio" now stuffed with three of the biggest tech companies on Earth — effectively the AI trade wearing a false moustache.
The conclusion is uncomfortable: for most investors, getting out of the AI trade is harder than it looks. The sector doesn't stay in its lane. It has spread into utilities (data centres need enormous amounts of power), real estate (server warehouses and land), construction (a data centre building boom is underway across the US), and even consumer spending, as AI-related IPO wealth disperses into the broader economy through home purchases, luxury goods, and private jet charters.
What the Numbers Actually Say: Sizing the Potential AI Crash
So how much money is actually at stake? Three credible estimates, three different methodologies — and they all land in the same ballpark.
Economist Dean Baker runs what he calls the AI Bubble Monitor. His calculation: the total US stock market sits at roughly $80 trillion. If price-to-earnings ratios simply drifted back to their long-run historical average — not a crash, just mean reversion — approximately $40 trillion of stock market wealth would be erased. That averages out to nearly $300,000 per US household, though that figure is heavily distorted by concentration: the wealthiest 10% of Americans own roughly 90% of shares, so the median household would feel far less.
Gita Gopinath, former chief economist at the IMF, estimates a dot-com-style correction today would destroy roughly $20 trillion of American wealth, plus an additional $15 trillion held by foreign investors. That $20 trillion figure represents approximately 70% of US GDP.
Oliver Wyman consultants ran their own model and landed near $33 trillion in value destruction — more than the entire US economy produces in a year.
For perspective: the actual dot-com bust destroyed around $6 trillion in equity value. The mainstream estimates for an AI crash run five to six times larger. The order of magnitude is consistent across all three approaches, even if the precise figures differ.
The Wealth Effect: Why a Stock Crash Hits Main Street Now
Here's the structural shift that makes this moment different from 2000. According to Goldman Sachs and Federal Reserve data, stocks overtook real estate as the single biggest component of American household wealth — for the first time since the Second World War.
For most of modern history, the average family's net worth lived in their house. A stock market crash was, relatively speaking, a rich person's problem. That is no longer true. The stock market is now where the median household's wealth actually resides, which means a crash today has a broader reach into middle-income America than any previous tech correction.
The mechanism is the wealth effect. Research suggests that for every $100 of paper stock wealth, people spend approximately $3 in the real economy. That sounds modest until you multiply it across tens of trillions in notional losses. The arithmetic runs cleanly in reverse: if $30 trillion of household wealth evaporates, consumer spending contracts sharply. The new kitchen gets cancelled. The contractor loses revenue. The car dealership has a bad quarter and lays someone off. The damage cascades through the real economy without a single bank needing to fail — it just requires people to feel meaningfully poorer.
Harvard economist Jason Furman calculated that AI-related infrastructure spending accounted for something like 90% of US economic growth in one recent six-month period. More conservative estimates put the figure around 25%. Either way, a significant share of recent economic growth is, in effect, data centre construction. The electricians, concrete crews, transformer manufacturers, cable installers, and truck drivers who depend on that spending don't own Nvidia shares. They don't need to. If AI capital expenditure slows, their income slows with it.
Nobel laureate Joseph Stiglitz has noted the additional complication: a market correction and AI-driven labour displacement could hit simultaneously. Households facing falling savings at the same moment their jobs become less secure is not a theoretical risk — it's a plausible scenario that policymakers are not visibly prepared for.
The $3 Trillion Iceberg: What Big Tech's Balance Sheets Don't Show
The banking system is in substantially better shape than it was in 2008. Capital ratios are higher, liquidity buffers are real, and annual stress tests have genuine teeth. A Lehman-style cascade through the financial system is not the most likely vector for pain here.
But when regulators make one part of the financial system safer, risk doesn't disappear — it migrates. A Wall Street Journal analysis found that the major tech companies' reported capital expenditure on AI infrastructure — data centres, chips, computing capacity — totalled roughly $600 billion over a recent 12-month period. Reading the footnotes of their SEC filings tells a different story. The Journal identified approximately $3 trillion of additional AI commitments that don't appear on the balance sheet: long-term data centre leases, locked-in chip purchase agreements, energy contracts. Alphabet alone carries over $800 billion of this type of off-balance-sheet obligation.
Analysts have labelled this the AI spending iceberg. None of it is hidden — it's all disclosed in the filings, for anyone who reads footnotes. But it means the true scale of what these companies are committed to spending is roughly five times larger than the headline capex figures suggest. And every dollar of it rests on the assumption that AI revenues will materialise to cover it.
The credit markets have begun to take notice. The cost of insuring big tech debt against default has recently hit record highs. Meanwhile, private credit — the lightly regulated, $2–3 trillion market of funds that lend directly to companies outside the traditional banking system — has seen troubled loans at the 20 largest listed funds climb to their highest level since 2017. Fitch reported private credit defaults hit a record in a recent month. One major fund disclosed that 7% of its entire loan book was in distress. These are not systemic warnings yet, but they are early signals worth monitoring.
What Investors Should Actually Be Thinking About
None of this is an argument for selling everything and sitting in cash. The bull case for AI is serious, well-funded, and backed by genuine technological progress. The companies investing at scale believe the payoff will be enormous — and they may well be right.
But a few practical considerations are worth internalising:
- Concentration is higher than most portfolios reveal. If you hold a broad index fund, a value fund, and a small-cap fund, you may have more AI exposure than you think — often overlapping in the same names.
- The wealth effect is now a mainstream risk, not a tail risk. With stocks surpassing real estate as the dominant household wealth vehicle, a market correction has real-economy consequences that didn't apply in earlier tech cycles.
- Off-balance-sheet commitments matter. The $3 trillion in undisclosed tech spending obligations is a legitimate analytical data point. Investors and analysts who only read headline capex figures are working with incomplete information.
- Private credit stress is worth watching. It's not 2008, but risk doesn't vanish — it relocates. Rising default rates in private credit funds that have lent heavily to AI and software companies is a canary worth monitoring.
- Diversification requires active scrutiny. Mechanical index rebalancing can inadvertently concentrate exposure. Understanding what's actually inside your funds — not just the label on the tin — is more important now than it was five years ago.
The scale of potential losses, if the bear case proves correct, runs between $20 trillion and $40 trillion by credible estimates. That is not a number to be dismissed as abstract. It is, depending on whose model you use, somewhere between the entire annual output of the US economy and twice that figure — in paper wealth that could, under adverse conditions, simply stop existing.
The party may well continue for years. But knowing the size of the tab is basic financial hygiene.
Frequently Asked Questions
How much wealth could an AI market crash realistically destroy?
Estimates from credible sources vary but cluster in the tens of trillions. Economist Dean Baker suggests $40 trillion if US price-to-earnings ratios revert to historical norms. Former IMF chief economist Gita Gopinath estimates roughly $20 trillion in US wealth destruction, plus $15 trillion held by foreign investors. Oliver Wyman consultants modelled approximately $33 trillion in losses. For context, the dot-com crash destroyed around $6 trillion — making these estimates five to six times larger in scale.
Is diversification enough to protect against an AI bubble crash?
Standard diversification is less protective than many investors assume in this cycle. The AI trade has spread into small-cap stocks (through chip equipment and testing companies), value indices (via semiconductor stocks added during rebalances), utilities, real estate, and construction. Investors who believe they've avoided the AI trade may find significant indirect exposure when they look inside their funds. Scrutinising actual holdings — not just fund labels — is advisable.
Why would an AI crash affect people who don't own tech stocks?
Two main channels. First, the wealth effect: research suggests every $100 of stock market losses reduces real consumer spending by around $3. Multiply that by tens of trillions and the consumer economy contracts meaningfully. Second, AI infrastructure spending (data centre construction, chip manufacturing, power infrastructure) now accounts for a significant share of US economic growth — by some estimates, a quarter or more. Electricians, contractors, manufacturers, and logistics workers who depend on that spending are exposed without holding a single share.
What is the AI spending iceberg and why does it matter?
Wall Street Journal analysis found that major tech companies' publicly reported AI capital expenditure — approximately $600 billion over a recent 12-month period — understates their true financial commitments by a factor of roughly five. An additional $3 trillion in obligations (long-term data centre leases, chip purchase agreements, energy contracts) sits in the footnotes of SEC filings rather than on balance sheets. Alphabet alone has over $800 billion in such commitments. These obligations are legally binding and rest on the assumption that AI revenues will eventually materialise to cover them — making them a meaningful risk factor if AI monetisation disappoints.
How does an AI crash differ from the 2008 financial crisis?
The major banks are significantly better capitalised than in 2008, with stronger liquidity buffers and mandatory stress testing. A Lehman-style banking cascade is not the primary risk scenario analysts identify. The more likely transmission mechanism is the wealth effect — falling asset prices reducing consumer spending and real economic activity — combined with a slowdown in AI infrastructure spending that ripples through construction, manufacturing, and utilities. The risk also lives partly in private credit markets, which have grown to $2–3 trillion and have seen rising default rates, though these remain well below systemic crisis levels.
This article is for informational purposes only and does not constitute financial advice. Always consult a qualified financial professional before making investment decisions.
About Zeebrain Editorial
Zeebrain publishes independent analysis of markets, investing, personal finance, and business. We disclose affiliate relationships, never accept payment for coverage, and fact-check all claims against primary sources. Read our editorial policy →
Disclaimer: Content on Zeebrain is for informational and educational purposes only and does not constitute financial advice or a recommendation to buy or sell any security. Always conduct your own research and consult a qualified financial adviser before making investment decisions. Past performance is not indicative of future results.
More from Business & Money
Related Guides
Keep exploring this topic
AI Bubble or Bull Run? What Investors Need to Know
Business & Money · AI bubble · S&P 500
How AI Debt Hidden in Insurance Could Trigger the Next Crisis
Business & Money · AI bubble · life insurance
Citi & Goldman Issue Stock Market Warnings: What It Means
Business & Money · stock market · Goldman Sachs
Why the U.S. Is Buying Japanese Yen — and What It Means
Business & Money
Explore More Categories
Keep browsing by topic and build depth around the subjects you care about most.