In 2008, eager homebuyers and developers found themselves lured by adjustable-rate mortgages and flooded into the market, creating the housing craze of the mid-2000s. Then came a credit crunch as lenders lost confidence in borrowers’ ability to make good. In 2008, the destruction that started with overstretched homeowners and developers walloped the widespread economy.

Now there’s a similar danger brewing. But the catalyst isn’t reckless consumers. Now it’s the U.S. Treasury choosing daredevil financing. And because of the particulars of how the government has structured its debt, this plan could backfire down the road—and lead to a new financial crisis.

The parallels between the two borrowing booms should concern all Americans because in both cases, they left the economy highly vulnerable to unforeseen shocks. The problem isn’t just the gigantic scale of the new debt we’re shouldering to blunt the lockdown. If we were relying on long-term Treasuries that locked in low, fixed interest costs for decades, the threat wouldn’t be nearly so dire.

But in fact, the U.S. has doubled down by raising $3.6 trillion, equivalent to over one-fifth of all federal pre-pandemic debt, using ultra-short-term loans at never-before-seen, unsustainably low rates that are bound to jump. If they jump far enough, the U.S. would endure gigantic budget shortfalls that would panic our creditors and cripple the private sector with soaring interest costs. Only giant tax increases would halt the damage.

To weigh what the U.S. risks by wagering on short-term borrowing, it’s instructive to examine what happened in the financial crisis. In 2008 and 2009, the upheaval arose from a real estate bubble that shattered. But the type of mortgages held by U.S. homeowners undermined their ability to weather the hurricane. Between 2003 and 2006, around 20 million households gorged on $4 trillion in adjustable-rate mortgages, or ARMs, priced at low fixed rates for, say, two years. After that initial period, the ARMs reset at the much-higher market levels for floating-rate debt.

From 2001 to 2006, ARMs as a share of all home-loan originations soared more than threefold, to over 50%. But in late 2009, between $50 billion and $100 billion a month in ARMs were resetting at rates as high as 9%, far above the initial “teasers” that first attracted the homeowners. Millions of families couldn’t afford the new payments, and the credit crunch caused by the mortgage meltdown made it impossible for most of them to roll the ARMs into safe, 30-year home loans. Put simply, the ARM holders took a flier, and it flopped in an onslaught of foreclosures and bankruptcies, amid a deep recession that sank national income 4.3% from late 2007 to mid-2009.

The proliferation of ARMs helped cause the financial crisis by inflating housing prices. But it also made the outcome far worse by heightening the risk faced by America’s households. Families who could have muddled through the downturn with a 30-year mortgage at 5% lost their homes because they had wagered on an ARM that reset at rates almost twice that high, nearly doubling their monthly payments. We don’t know what will cause the next upheaval—widespread trade wars that bring on a new era of protectionism, a deep recession coupled with a collapse in stock prices, or another black swan like COVID-19. But whatever the shock, today’s explosion in super-short-term borrowing could turn what would have been a passing tremor into a devastating earthquake.

John Cochrane, an economist at Stanford’s Hoover Institution, warns that the U.S. is ignoring the lessons from the financial crisis at great peril. “America is using the equivalent of ARMs to finance the relief programs to fight the coronavirus,” warns Cochrane. “Rates on the short-term federal debt are super-low right now, but if rates go up, the U.S. is—what’s the polite word for ‘screwed’? What if the Treasury suddenly can’t find the new people to pay off the old borrowers, except at much higher rates, like the ones that sank the ARM borrowers?” That’s the formula for a fiscal catastrophe, Cochrane adds. He notes that in his home state of California, earthquakes are unlikely, but do happen, and unpredictable financial earthquakes happen too. He adds that it was precisely the inability to refinance short-term debt that caused the Greek debt crisis a decade ago.

The risks of short-term debt

To fund the four relief and stimulus packages designed to ease damage from the lockdown, the Treasury has embarked on by far the biggest one-year borrowing program in U.S. history. Keep in mind that the U.S. is adding that COVID-19 debt on top of the borrowings needed to fund the $1 trillion deficit for fiscal 2020 (ending September 30) projected before the outbreak. On May 4, the Treasury announced that it will borrow $2.999 trillion in the three months from April to June, versus the previous plan to retire $56 billion.

The structure of that new borrowing is a sharp departure from the program in recent years. Since 2012, the U.S. has been capitalizing on historically low longer-term interest rates to extend the average maturities of its fast-growing debt. The reason is obvious: Shifting the balance toward 10-year notes and 30-year bonds is a lot safer than relying on “T-bills,” securities that come due in a time period from a couple of weeks to a year. From 2012 to 2019, that swing toward conservatism enabled the U.S. to extend the average maturity from 60 months, which has been the norm for several decades, to around 70 months, meaning that 17% versus 20% of all U.S. debt comes due each year.

In the first quarter from October to December, the Treasury kept that strategy in place. Of the $330 billion in new debt, 42% came in maturities of seven years or more. The “go-short” reversal went from pronounced in the January to March quarter (Q2), to overwhelming in Q3. In Q2, the U.S. borrowed a total of $477 billion. Of that number, 26% came in maturities of seven years or more, and just over half came due in a year less.

In Q3, borrowing exploded, and so did the shift to selling bonds carrying the cheapest rates, meaning the ones with the shortest maturities. Of the staggering, nearly $3 trillion jump, Treasury forecasts that it will fund $2.750 trillion, or 92%, in bills running a year or less, and another $92 billion at terms of three years or shorter, for a total of $2.840 trillion, 95% of the total. The Treasury has also disclosed that it intends to sell another $677 billion in securities in the final quarter ending Sept. 30. We don’t yet know the breakdown of maturities, but it’s reasonable to predict the Treasury will follow the crisis-driven pattern of Q3, so that around 90% of the offerings will have terms of a year or less, and almost all of them will carry durations of no more than three years.

If Q4 resembles Q3, as expected, the U.S. debt picture will undergo a historic transformation in just 12 months, both in the well-publicized surge in total debt, but also in the mostly overlooked move to shorter maturities. The binge on bills, as opposed to going long, makes the outlook far more dangerous. By Fortune’s estimates, by the close of September, the U.S. will have added $4.15 trillion in new debt, of which $3.6 trillion, or 86%, will be due in a year or less—generally a few weeks or months. And $3.78 trillion, or 91%, will need to be refinanced by mid-2023.

Why aren’t we going long?

The reason for the gigantic shift both to the size and share of Treasury bills is clear: In this time of exploding deficits, Treasury sees its role as minimizing the extra costs of borrowing.

It’s easy to see Treasury’s rationale: Going short has never been cheaper. Since the outbreak signaled a sharp economic slowdown, rates on short-term debt have collapsed. In late February, the six-month Treasury was paying 1.52%; by May 22, it was yielding 0.14%, and 13-week bills now trade at 0.12%. On a current run rate, a good guess at the extra interest the U.S. will pay this year on that entire $4-trillion-plus in new debt might be a tiny 0.5%, or $20 billion, an addition of just 5% to last year’s interest bill.

That break won’t last. The average market rate on six-year Treasuries was 5% from 1960 to early 2007, and it reached 2.6% in December 2018. So Treasury is taking a big risk, since anything like a return to historic averages would push interest expense to ruinous levels. On the other hand, long-term debt is also extremely inexpensive right now. The 10-year note is yielding just 0.64%, and the 30-year is a screaming deal at 1.27%; it stood at 4% as recently as 2013.

So why isn’t the U.S. picking long-term protection at bargain rates? “The U.S. should be going with long-term financing,” says Cochrane. “That way, if there is any problem and rates go up, the U.S. isn’t stuck with massive interest costs and deficits. Borrowing long-term insulates the economy from a financial crisis.” In fact, Treasury Secretary Steven Mnuchin has shown a strong interest in choosing a safer path. In early March, he told the House Ways and Means Committee that he had explored interest in new, 50-year Treasury bonds and got a less-than-enthusiastic reception. “We went out to a large group of investors and solicited feedback, and I was somewhat surprised that there’s some interest but not enough that it would make sense to issue those bonds,” Mnuchin told the legislators.

It’s also likely that Treasury isn’t pushing hard on 30-year bonds because investors balk at buying tons of those too. In early March, the demand for new offerings of 30-year Treasuries collapsed, driving rates skyward. Enter the Fed, which calmed the waters by purchasing $17 billion in long-dated Treasuries. “Treasury was having a hard time selling 30-year bonds in March,” says Cochrane. “It was a warning sign that the Fed had to step in to buy lots of the longer-term bonds Treasury wasn’t able to sell in the private markets.”

For Cochrane, the blowup may explain why Mnuchin isn’t championing long-term debt that would lock in rates at historically low levels for decades to come. In the absence of strong demand for 10- or 30-year borrowings, Treasury is pursuing a strategy of funding where debt is cheapest. That means replacing maturing Treasury bills with the new bills of the same term, simply exchanging three- or six-month Treasuries that come due with new three- or six-month Treasuries. Clearly, Treasury isn’t worrying about defending the U.S. against rising rates.

Danger ahead

Though slender rates are restraining interest costs right now, the scale of the extra debt puts the U.S. in danger. And the ARM-style borrowing magnifies the menace. By the end of the September fiscal year, the U.S. will hold approximately $21.7 trillion in debt, a 29% increase of almost $4.9 trillion on last year’s $16.8 trillion. In a single year, the national burden will jump from 79% to over 100% of GDP, and that’s using last year’s national income figure, not this year’s stricken number, as the benchmark. “When you have debt that huge, it doesn’t take much of a rise in rates to bury you,” warns Brian Riedl, an economic policy expert at the conservative Manhattan Institute.

Riedl points out that a 2% increase in average rates on that $21-trillion-plus load, leaving rates still well below their historic average, would add $420 billion interest expense, $45 billion more than America’s total interest expense in 2019, and almost 10% of all outlays.

What scenario could trigger a fiscal crisis? Cochrane isn’t worried about a rise in rates caused by stronger economic growth. “In that case, rates are rising because there’s more demand for capital, and companies are getting more productive,” he explains. “Tax receipts go up, and the need for spending goes down. If the rate increase is driven by a strong economy, we can afford it.”

The real danger, he says, comes from a confluence of a bad economy and a crisis of confidence, with one force-feeding the other. In a recession, tax receipts would dwindle, causing deficits to explode, and investors would fret that the U.S. can’t make good on its bonds without stoking inflation that would crush their value. Result: a flight from Treasuries that would hike rates and throw the U.S. economy into a tailspin. Once again, cautions Cochrane, the chances that endgame will arrive are far greater because we’re choosing the fix of short-term financing.

The danger of ARM-style financing isn’t the only lesson from the financial crisis. In effect, the U.S. is paying for fixed and growing long-term obligations, everything from Social Security to Medicare to the military, with more and more short-term funding. In the real estate frenzy in the early years of the century, Lehman Brothers spent billions of dollars on apartments and office buildings, often carrying long-term mortgages, using short-dated securities such as commercial paper. General Electric followed the same playbook, issuing $150 billion in commercial paper to back a gigantic real estate portfolio.

Lehman and GE were doing just what the U.S. is doing now, using securities due in weeks or months to cover fixed payments stretching years into the future. When it was clear that Lehman couldn’t cover those payments on assets collapsing in value, it failed to raise the funds to survive. On Sept. 15, 2008, Lehman folded in the biggest bankruptcy in U.S. history, exposing the folly of that mismatching funding and obligations.

That’s a warning America should take to heart while there’s still time.

More must-read finance coverage from Fortune: