What do you get when you combine a 2008 housing bubble, a 2000 dot-com bubble and a 1998 currency bubble?
This is the part of market history I keep coming back to.
What do you get when you combine the ingredients of three very different crises?
A 2008 style housing and credit bubble.
A 2000 style technology and valuation bubble.
And a 1998 style currency and leverage problem.
The scary part is that these three episodes broke in different places.
In 2000, the problem was concentrated in stocks and especially technology.
The Nasdaq eventually fell about 78% from its March 2000 peak to its October 2002 low.
The important part wasn’t simply that technology stocks were expensive.
A huge amount of capital had been committed to a story about future growth.
When expectations broke, the investment cycle broke with them.
In 2008, the housing bubble moved through mortgages, banks and credit markets before becoming a much larger economic crisis.
U.S. home prices eventually fell more than 20% nationally from their peak.
Housing construction collapsed.
Mortgage losses spread through the financial system.
Banks pulled back lending.
The recession then fed back into housing and credit.
Then there was 1998.
Thailand’s currency crisis had already spread through Asia.
Russia devalued the ruble and defaulted on its debt.
Investors ran toward safer assets.
Then Long Term Capital Management started coming apart.
LTCM had roughly $30 of debt for every $1 of capital at the end of 1997.
It lost 44% of its value in August 1998 alone.
Fourteen banks and brokerage firms eventually put together a $3.6 billion rescue to prevent a disorderly collapse and fire sale.
That is the part people forget about 1998.
The currency problem didn’t stay a currency problem.
It became a leverage problem.
The leverage problem became a liquidity problem.
And suddenly something that looked contained was threatening the wider financial system.
Now look at the ingredients people are worried about today.
Technology has become an enormous concentration of market value.
AI spending has become a gigantic capital expenditure cycle.
Housing remains extremely expensive relative to incomes.
Government debt is enormous.
Long term borrowing costs are elevated.
Private credit has grown into a huge source of financing.
And currencies can still become the pressure point that exposes leverage somewhere else.
The numbers around AI are already enormous.
The five largest U.S. technology companies spent $380 billion on capital expenditure in 2025.
They are expected to spend roughly twice that amount in 2026.
That means roughly $760 billion from just five companies.
And one recent estimate puts the broader AI infrastructure buildout at around $745 billion this year.
Another estimate says AI infrastructure could require $1.5 trillion of spending in 2026.
David Cahn, the Sequoia partner who has been tracking the economics of the buildout, estimates that the industry may ultimately need roughly $3 trillion in revenue to justify the infrastructure spending already being committed.
Those aren’t normal technology numbers.
This is becoming one of the largest investment cycles in the economy.
And that is exactly why Torsten Sløk, Apollo’s chief economist, made such an interesting comparison.
Sløk says AI infrastructure investment is growing at nearly twice the pace of the peak housing boom in terms of its annual contribution to GDP growth.
That matters.
Because housing wasn’t just another industry in 2008.
When housing investment reversed, the damage spread through construction, employment, household wealth, mortgages, banks and eventually the entire economy.
Now imagine a similar reversal in a much more complicated network.
Data centers.
Semiconductors.
Power generation.
Electricity transmission.
Construction.
Equipment.
Real estate.
Private credit.
Public equities.
If AI spending keeps accelerating, all of those businesses benefit from the same giant capital cycle.
If it suddenly stops, they can all get hit by the same reversal.
And the scale means the damage wouldn’t have to start with a bank.
It could start with a few hyperscalers deciding the return on another $100 billion of infrastructure isn’t good enough.
Cut the spending.
Then suppliers lose orders.
Construction projects get delayed.
Equipment orders disappear.
Power projects get cancelled.
Revenue forecasts get cut.
Stocks fall.
Then the financing problem begins.
This is where Jeff Schmid, president of the Kansas City Federal Reserve, becomes important.
Schmid has warned that the financial structures forming around the AI buildout deserve close attention.
He has raised the possibility that AI could become “too big to fail” and specifically discussed the danger of interconnected financing and cascading failures.
That is a very different warning from simply saying AI stocks might be overpriced.
He’s talking about connections.
And there are already signs of stress on the credit side.
Fitch says the U.S. private-credit default rate has reached a record 6%.
There were 99 default events in the 12 months through April, with 81 first-time defaults.
More than half involved interest deferrals or PIK structures.
PIK usage has risen to about 10% from 6% in early 2022.
In plain English, some borrowers aren’t generating enough cash to pay their interest.
So they’re adding the interest to the debt.
The private-credit market has grown beyond $1 trillion.
That doesn’t mean a 2008 style collapse is already happening.
It means the system is entering the next downturn with a credit market that is considerably larger and already showing stress.
Now connect the pieces.
AI investment keeps climbing.
The five largest tech companies potentially go from $380 billion of capex to roughly $760 billion in one year.
AI infrastructure approaches 3% of U.S. GDP in coming years under some estimates.
Sløk says its growth rate is already running at nearly twice the pace of peak housing investment.
Then demand disappoints.
Maybe AI revenue grows, but not fast enough to justify the infrastructure spending.
Maybe the economics of running enormous models deteriorate.
Maybe companies realize they have overbuilt.
Maybe investors simply decide that the returns won’t justify another trillion dollars of spending.
The first thing that happens is obvious.
Capex gets cut.
Then the multiplier runs backward.
Chip orders fall.
Data center construction slows.
Equipment companies lose contracts.
Power projects get postponed.
Industrial suppliers lose revenue.
Private-credit borrowers tied to those industries start missing targets.
The default rate moves above 6%.
PIK increases again.
Lenders become less willing to refinance.
Then you get the 2008 ingredient.
Housing is already carrying high prices and high financing costs.
If long-term rates stay elevated while employment weakens, transactions can freeze.
Builders cut starts.
Commercial real estate struggles to refinance.
Household wealth falls.
Banks tighten lending.
Now the AI slowdown has become a credit slowdown.
Then bring in 1998.
A currency shock hits.
The yen moves violently.
Carry trades unwind.
Leveraged investors have to sell.
Treasury yields rise.
Suddenly the normal escape route disappears.
Because what happens if stocks are falling while bonds are falling too?
The Fed can cut short-term rates.
But that doesn’t automatically force long-term Treasury yields down.
If investors are demanding higher yields because of inflation, fiscal risk or Treasury supply, the government can find itself trying to rescue the economy while its own financing costs are rising.
That is the ugly combination.
Technology valuation collapse.
Credit deterioration.
Housing weakness.
Currency stress.
A Treasury market that refuses to behave like the safe haven everyone expects.
Now look at what happened historically.
The Nasdaq fell about 78% after the 2000 peak.
U.S. home prices fell more than 20% nationally during the housing bust.
LTCM lost 44% in one month and had to be rescued by 14 financial institutions.
Those weren’t isolated losses.
Each crisis spread into something larger.
That is the important lesson.
If the next AI downturn happens by itself, it could simply be another major technology correction.
But if AI spending falls at the same time private-credit defaults are climbing, housing is weakening and a currency shock is forcing leveraged investors to liquidate, the consequences become much larger.
You could get trillions of dollars of equity losses.
Hundreds of billions of dollars of cancelled infrastructure spending.
A wave of corporate refinancing failures.
Falling construction.
Rising unemployment.
Lower consumer spending.
More bank losses.
More government borrowing.
And then another rescue.
But the rescue could create its own problem.
More Treasury issuance.
More fiscal spending.
More liquidity.
More pressure on the dollar.
More inflation risk.
And that changes what shines.
The winners wouldn’t necessarily be the companies with the fastest growth.
They would be the companies that can survive without cheap money.
Energy producers.
Gold.
Copper.
Power generation.
Electricity transmission.
Critical infrastructure.
Cash-rich companies.
Companies with little debt.
Businesses selling things people still need when credit disappears.
The biggest losers would be on the other side.
Companies that need constant refinancing.
Companies whose customers depend on cheap credit.
Companies valued on profits that are supposed to arrive five or ten years from now.
Companies whose entire business model depends on AI capex continuing to accelerate.
That is why I think combining 2000, 2008 and 1998 is more useful than simply asking whether today’s market “looks like” one of them.
The bubbles were different.
The transmission mechanisms were different.
But they can connect.
AI creates the investment boom.
A reversal destroys the spending behind it.
Private credit spreads the losses through leveraged companies.
Housing turns financial stress into an economic downturn.
A currency or Treasury shock makes the rescue harder.
And if all of that happens close enough together, the final break may come from somewhere nobody was watching.
A private-credit fund.
A highly leveraged company.
A data-center financing vehicle.
A major currency trade.
A Treasury auction.
The thing that starts the crisis may be AI.
But the thing that finishes it could be credit, housing or the Treasury market.
That is how history rhymes without repeating.
Not financial advice. This is market commentary and historical analysis, not a recommendation to buy or sell any security.