It almost seems too good to be true.

Right now, the U.S. has a rare opportunity to borrow at such slender rates that our total interest expense will keep falling even as our total debt load rises rapidly. The U.S. gets to lavish money on worthwhile initiatives without risking a surge in interest expense that, at most times, would swamp the federal budget.

But there’s just one problem, one that Larry Summers singles out. President Clinton’s Treasury secretary is an outspoken champion of exploiting today’s super-low interest rates to expand federal borrowing for investment in infrastructure, green energy, and social programs. The size of the debt and deficits, he has argued, is much less important than the costs of carrying that debt.

But Summers stamped a key caveat on his blueprint. The strategy would only skirt big risks if the Treasury centered its strategy on borrowing at today’s long-term rates, now at their narrowest in history, locking in affordable payments far into the future. Summers is appalled that the U.S. government is going in precisely the opposite direction, pursuing the notoriously dangerous tack of essentially funding its bonds at longer maturities with overnight debt. “In effect, the government now has a floating rate, short-term liability outstanding rather than long-term, fixed interest rate liability,” Summers told Bloomberg Television’s Wall Street Week on Aug. 13. “At a moment of super uncertainty, at a moment when many people think rates are remarkably low, a decision to fund more short seems bizarre.”

Summers was specifically addressing the Federal Reserve’s “quantitative easing” (QE) campaign of deploying day-to-day borrowings as a vehicle for buying the vast bulk of the long-dated Treasuries now being issued. Those purchases are famously running at $80 billion a month. The U.S. is saving tens of billions a year because it’s sending the interest payments on those Treasuries right back to the Treasury in a closed loop. But the program could easily backfire. The U.S. is paying ultracheap overnight rates to keep the newly created money that it’s using to buy those Treasuries from leaking into new lending that would stoke inflation. Specifically, the central bank is luring financial institutions to park their “excess reserves” on its balance sheet. Right now, the Fed gets to hike the money supply and still hold off inflation by simply paying a minuscule 0.15% to sequester all those trillions.

Here’s a road map for the journey to peril, and it’s all too plausible. The economy continues to roar back from the lockdown, causing today’s strong inflationary pressures to persist or worsen. The Fed is forced to pay much higher rates on those $4.2 trillion in bank reserves to stanch the flood of money from flowing into car, home, and consumer loans, and driving prices higher. That extra expense would greatly increase the interest payments on the federal budget. Since the Fed would be buying fewer newly issued bonds, tapering would further upset the fragile balance by shrinking the effective “subsidy” the central bank is handing the Treasury. Until now, a remarkable blend of the lowest bond yields in history and the Fed’s trick of buying gobs of newly issued Treasuries and rebating the interest to the Treasury has kept that bill falling as our debt load explodes. The Fed’s walking a slippery tightrope to make that happen. A gust of inflation could send the extraordinary balancing act tumbling.

History’s biggest ARM

What Summers labels as the Fed’s “bizarre” path accounts for only part of what makes the high wire so treacherous. The Treasury is also funding a big portion of the huge stimulus outlays shouldered to counter the COVID crisis with Treasury bills due in four weeks to a year. All told, around half of U.S. federal debt that has expanded 30% since the close of 2019 to a towering $22.2 trillion, is now backed by the Fed’s overnight borrowing, bonds due in 12 months or less, or floating rate securities. As economist John Cochrane of the Hoover Institution puts it: “The U.S. is facing the same danger as Americans who bought homes during the 2006 housing bubble with adjustable-rate mortgages.”

Today, the world’s largest economy is harboring the equivalent of history’s biggest ARM. Put simply, the U.S. is courting major risks to hold today’s interest expense artificially low and make the fiscal picture look less precarious than it really is. Summers is calling on the Fed to “bring QE to an end,” in part because of the program’s mismatch of fast-rolling credit backing multiyear bonds. At the Fed’s meeting in late July, a majority of its Open Market Committee members advocated that tapering commence later this year. Economists at Charles Schwab predict that the Fed will start doing so in November, lowering the dollar amount of its purchases by $8.5 billion per month.

As we’ll see, the Fed can conceivably keep the U.S. on its current course by continuing its artful acrobatics. But that’s only if inflation is transitory. In July, the consumer price index waxed at 5.4%, matching the largest jump since August of 2008. The Fed is already predicting a 3% increase for all of 2021, well above its average target of 2%. Besides promoting full employment, the central bank’s top job is ensuring stable prices. If inflation is here to stay, fulfilling that mission could unleash a giant rate-reset on America’s ARM-style funding. That reset could take the annual interest tab for the biggest debt run-up in U.S. history from a seeming bargain to staggeringly expensive. The Larry Summers warning is right on.

The Treasury financed much of the COVID-driven spending boom on a short-term basis

The fresh debt the U.S. shouldered to battle the pandemic is making our fiscal future much riskier. Even though the Treasury is now shifting to longer-term borrowing, the extra trillions due in a year or two make the U.S. far more vulnerable to a sudden rise in rates.

At the end of December 2019, the federal government owed “debt held by the public” of $17.2 trillion. That category encompasses all Treasuries held by individuals, corporations, foreign governments, and the twelve Federal Reserve banks. Of that total, $2.4 trillion or 14% were in T-bills with maturities of four weeks to a year. The bulk at 58% or $9.3 trillion were T-notes starting with two-year terms extending to the benchmark 10-year at the long end. T-bonds at 20- and 30-year maturities accounted for $2.4 trillion, or 14% of all debt. Overall, more than 72% of U.S. borrowings sat in longer-dated T-notes and T-bonds.

That relatively prudent profile shifted radically when the U.S. undertook gigantic emergency spending and borrowing to combat the COVID-19 crisis. From the close of 2019 to August of last year, the U.S. Department of the Treasury went short in a big way. It issued an astounding $2.7 trillion in T-bills, securities rolling over, on average, every couple of months. The Treasury deployed T-bills for three out of every four dollars in new borrowing. In contrast, it sold just $1.4 trillion in longer dated T-notes and T-bonds—only half the proceeds from far riskier T-bills. (All figures for bonds issuance are expressed net of Treasuries that matured in the same periods.)

As of August 2020, the share of total borrowings in T-bills jumped by over 10 points to 23.4%. Add floating rate bonds, and the portion of less-than-one-year Treasuries and those indexed to short-term rates jumped from one-quarter to over one-third of all federal debt outstanding. The average maturity on federal bonds in the first three quarters of 2020 fell from 69 months to 62 months, one of the sharpest sudden drops in history.

Since then, the Treasury has reversed course. It’s been restoring the balance in its “outstandings” by originating far fewer T-bills that mature in less than a year, and gunning the volumes of new, longer-dated T-notes and T-bonds. The trends are detailed in a recent report from the Treasury’s Office of Debt Management updated through the federal government’s third quarter ended June 30. It includes projections for fresh debt issuance through the September close of the fiscal year. The Treasury forecasts that for the 12 months ended Sept. 30, it will retire $742 billion more in T-bills than it sells.

Of total new borrowing of almost $2 trillion for the 2021 fiscal year, $1.45 trillion will be funded by T-notes at maturities of five years or more, as well as T-bonds at 20 years and 30 years. The U.S. is returning to a much more conservative, traditional course by raising almost three-quarters of new borrowing at longer-dated maturities. In a jiffy, the average maturity of U.S. debt has jumped back to its pre-COVID level of 69 months.

It’s a good thing that the Treasury has restored the array of different maturities to their pre-crisis weights. The problem: Even if the proportions remain at their traditional levels, the U.S. now has much more short-term debt outstanding than at the close of 2019. The dollars matter more than the pieces of the pie. As of July, the U.S. owed $6.6 trillion in T-bills due in a year or less, plus floating rate debt. That’s $2 trillion more than it held across those categories in December of 2019. Now that the proportion of T-bills is back to normal, the Treasury began issuing them again in its fourth quarter.

From now on, the Treasury is likely to keep selling T-bills and floating rate bonds at a pace that maintains their current 30% share of total debt held by the public. The Treasury will need to borrow around $2 trillion annually in future years to fund our gigantic deficits. Hence, dollars owed in two-month to one-year and floating rate bonds will keep mounting rapidly—all from the much higher plateau established by COVID emergency funding. The need to constantly roll over or reset rates on that $6.6 trillion-and-growing nut makes the budget far more exposed to an inflation outbreak and rising rates versus just 18 months ago.

The Fed could face much higher costs on the bank reserves it must corral to forestall inflation

The new trillions in short-term borrowing plant fresh hazards in America’s fiscal path. But the Fed’s policy of paying whatever is necessary to keep the banks from pumping its newly minted money into loans that would overheat the economy could also cause a sharp rise in the nation’s interest burden. Today, the Fed is famously swelling the money supply at a rapid rate. Its challenge is blocking all those trillions from doing what they usually do in a super–easy-money era, fueling inflation.

The Treasury raises funding by selling T-notes and T-bonds to financial institutions. At the same time, the Fed is sprouting new money by crediting funds to its own account. It increases the money supply, and the availability of credit to families and businesses, by purchasing Treasuries and mortgage-backed securities from the banks with that new money. The lenders then have more liquidity to finance credit card balances and home loans. Their customers deposit the dollars in their checking accounts, and the banks lend part of those new deposits to seed more deposits that spawn more loans from themselves and rival banks in a cycle that expands consumer and business spending across the economy.

Traditionally, the Fed didn’t buy most of the newly issued Treasuries from the financial institutions that just bought them from the Treasury. It purchased just enough to provide additional credit when the economy’s clicking. When inflation was looming, it sold Treasuries to the banks at attractive rates to dial back the excess dollars chasing goods and services. The banks held a large portion of Treasuries on their balance sheets as reserves, and sold some of them to the Fed when demand for loans increased. But the QE program that started in October 2008 at the height of the Great Financial Crisis changed that paradigm. In part, the campaign is designed to lift home, stock, and other asset prices by holding longer-term rates extremely low, raising the wealth for families and corporations. Under QE, the Fed is devouring gigantic chunks of the Treasury’s even more gigantic issuance of T-notes (two to 10 years) and T-bonds (20 and 30 years).

In the 12 months to mid-August 2021, the Fed bought an epic $978 billion in T-notes and T-bonds. By my estimate, it purchased well over 80% of the total of around $1.2 trillion in those securities maturing in five years or more. All told, the Fed’s holdings of the two longer-dated classes has doubled since mid-2016 to $4.7 trillion.

That trove of T-notes and T-bonds sits on the asset side of the Fed’s famed balance sheet. Keep in mind that the Fed is buying those Treasuries from banks that just bought the same Treasuries from the U.S. Department of the Treasury. In effect, the banks are just flipping the bonds to the Fed. The financial institutions know that the central bank will hoover all the T-notes and T-bonds they just acquired, rewarding them with commissions for serving as middlemen. “That’s one reason rates are so low,” says William Luther, a professor of economics and monetary policy at Florida Atlantic University. “The banks know that the Fed will instantaneously buy all the Treasuries they want to sell at today’s high prices.”

The Fed is using trillions in newly generated money to buy Treasuries from the banks in record amounts. If the banks turned the Fed’s fount—almost $1 trillion in the past year—into fresh loans, prices would rocket. “The Fed needs a lever to ensure that huge increase in the money supply doesn’t translate into extra bank credit,” says Luther. “The Fed’s goal is to ring-fence that money so that it doesn’t flow into loans that cause inflation.” The Fed’s power tool: paying interest on deposits.

What the Fed pays to isolate those funds has varied greatly over the years. It’s called the interest on reserve balances or IORB rate. In the six years following the Great Financial Crisis, the number was a minuscule 0.07% to 0.15%. But as the economy improved in 2018, the rate jumped to 2% and hovered at 2.4% from January to August of 2019. As recently as February of last year, it stood at 1.58%. The slashing of the Federal funds rate to nearly zero last year drove the IORB back to around today’s level of 0.15%.

Why would the banks park trillions in reserves at such a paltry return? One reason, Luther points out, is that they don’t have a choice. The Basel regulations require that the banks maintain a far higher ratio of capital to risk-weighted assets, including their loans adjusted for the risk of default. A second motive: Though the IORB sounds extremely modest, the Fed typically arranges matters so that it’s just slightly higher than the Fed funds rate at which the banks lend to one another. In addition, the banks are taking none of the usual risks on the trillions collecting interest at the Fed. “The reserves the banks hold at the Fed are totally safe,” says Luther. “That money doesn’t risk defaults or require servicing. And it earns a competitive return, though that return is now tiny.”

The magic of the “consolidated balance sheet”

Today’s fiscal story is the saga of the Fed aiding the Treasury as never before. The Fed is creating trillions in new money and, under QE, using it to buy the record amounts of longer-term debt that the Treasury is selling to the banks. Today, the Fed owns $4.7 trillion in those T-notes and T-bonds. Instead of owing all the money to outsiders, the Treasury instead owes it to the Fed. Because both arms fall under the government’s consolidated balance sheet, the U.S. owes that money to itself. Hence, all the interest the Fed collects from the Treasury it sends right back to the Treasury. That arrangement turns what could be a huge interest bill into a virtual wash.

Here’s my estimate of the money flow from the Fed to the Treasury. In 2020, the Treasury paid around 2% to the Fed on that $4.7 trillion that the Fed owns in long-dated Treasuries, or $94 billion. Buying all those Treasuries does come at a price to the Fed, but for now, it’s a small one. The central bank can only safely generate the “free” money available to amass those bonds if it also pays interest to keep the money it hands the banks for the T-notes and T-bills from rushing into inflation-spurring credit. Right now, the Fed’s paying a tiny 0.15% on the $4.2 trillion in reserves. In effect, it’s backing all those Treasuries bought via new money by debt that cost just $6 billion a year (0.15% of $4.2 trillion in reserves).

Last year, the Fed sent the Treasury $88 billion in profit, consisting of the $94 billion in interest that it collected from the Treasury, minus the $6 billion in interest it paid on reserves. What a deal for the Treasury! Instead of paying $94 billion in interest to banks, hedge funds, and other outsiders, the Treasury shuttled that money to itself, courtesy of the Fed. The Fed helped the U.S. Department of the Treasury keep America’s interest expense $94 billion lower than it otherwise would have been. But in providing that epic lift to the Treasury, the Fed is bearing big risks by effectively financing long-term bonds with the overnight debt it’s paying on reserves. The downside of the consolidated balance sheet: The Fed is taking a flier that could easily blow up, and if so, it will pass the damage to the Treasury, and on to the federal budget.

The danger if inflation ignites

Indeed, what happens if short-term rates spike? The big threat is that inflation has already arrived, and it’s settling in. To keep prices in check, the Fed will need to raise interest on reserves. The central bank is stuck. Remember, it absolutely cannot afford to let those reserves ooze into new credit that would inflame inflationary pressures that are already starting to boil.

Say inflation keeps running at 5%, where it is now. The Fed might need to pay as much as the 2.4% charge that prevailed just over two years ago. If tapering begins as expected later this year, the Fed will be purchasing smaller and smaller amounts of newly issued bonds.

But that will trigger another big change that’s also bad for the Fed’s bottom line.

It will no longer be rebating the interest payments on those securities to the Treasury. Instead, they will be owned by outside investors who will collect the coupons that used to go back to the Treasury. So QE would contribute to higher carrying costs as well. But let’s just consider how a swing of almost $100 billion in yearly interest expense, caused by much higher rates on bank reserves at the Fed, would ripple through the federal budget.

In 2019, the U.S. paid $376 billion in interest on average debt of $17.5 trillion. Since then, rates across the economy have cratered. That made new borrowing a lot cheaper. But the big drop in interest paid on reserves also helped a lot. Amazingly, the Congressional Budget Office is predicting that although debt will grow by almost 40% from 2019’s level to $24.3 trillion in 2022, the U.S. will miraculously be paying far less in interest, just $304 billion. But imagine that interest on reserves rises by $100 billion. That’s a 33% increase in all interest costs. And it won’t just be interest on reserves. Higher inflation will also hugely raise the costs of the $6.6 trillion in T-bills due in less than a year and floating rate debt. Overnight, the nation’s budget projections would be thrown into turmoil. That’s the peril Summers was talking about. He’s right. We don’t know when the greatest of all ARMs will reset. But we can picture the wreckage when it does.