Gross federal debt crossed $40 trillion on August 19, months ahead of the forecast path, in part because the revenue from the invalidated tariffs never arrived. The administration’s answer is growth. “The way you take care of debt is with growth,” Trump said. Bessent, on CNBC: there is nothing magic about the number, and the economy can grow out of it. Kent Smetters, who runs the Penn Wharton Budget Model, called it a fantastic story that is pretty clearly not feasible, and argued the causality runs the other way — you fix the debt in order to get the growth, not the reverse.
Smetters is right, but the interesting part is not the arithmetic he cites. It is that the specific growth engine being invoked in 2026 is the same engine currently bidding against the Treasury for capital.
The condition has two terms, and the story only addresses one
A debt ratio stabilises when nominal GDP growth exceeds the effective interest rate on the outstanding stock, with the primary balance at or near zero. Interest expense over the trailing twelve months has reached roughly $1.4 trillion against $40 trillion of gross debt, an average cost of about 3.5%. That number is a lagging artefact of a decade of sub-2% coupons still sitting in the book. It is not the marginal cost. The marginal cost is what the Treasury pays today, and the thirty-year has been printing above 5.3% — high enough that Bessent deployed over $4 billion in unscheduled buybacks to steady it.
So the effective rate is rising toward the marginal rate mechanically, on a schedule, with no policy decision required. Interest expense is projected to reach $1.7 trillion by late 2028. Net interest in 2026 already exceeds every mandatory program except Social Security and Medicare.
The second term is the one the story ignores entirely. The United States does not run a primary balance of zero. It has not run one in a generation. Growth above the effective rate would stabilise the ratio only if the government stopped adding new principal, and it is adding new principal at a pace that took the debt through $40 trillion ahead of schedule. Growth has to beat the interest rate and then beat it again by the width of the primary deficit. That is not a faster-growth problem. It is a different problem wearing the same word.
The growth being cited is partly the spending, not the return on it
The strongest version of the administration’s case is not the one it makes out loud. It is that AI represents a genuine productivity step-change, and that a step-change in productivity is the one thing that has historically moved fiscal arithmetic without legislation. That argument deserves to be taken seriously on its merits rather than dismissed with the ratio.
It still has a sequencing problem. Data centre construction, server procurement, power interconnection and the memory and networking content inside the racks are all counted as investment in the current GDP print. That is what “tremendous growth” is partly measuring right now: the outlay. The tax base does not widen from the outlay. It widens when the deployed capital raises output per worker across the rest of the economy, which shows up with a lag measured in years, and which the productivity statistics have not yet resolved. Between those two moments the capex flatters the growth number being used to argue that no fiscal adjustment is necessary.
The CBO checkpoint is fiscal 2028, when it expects debt of $43.3 trillion, and debt held by the public at 120% of GDP by 2036. Even a real productivity surge starting today does not reach the tax receipts inside that window.
The buildout is bidding against the Treasury for the same dollar
This is where the argument inverts on itself. The AI infrastructure cycle has moved decisively from equity funding to debt funding — corporate issuance, private credit, GPU-collateralised lending, SPV structures around individual sites. The calendar is running toward $1.5 trillion. That paper competes for the same fixed-income allocation the Treasury needs for the deficit and the rollover.
The long end has been repricing accordingly. What moved the thirty-year to its highest level since 2007 was not the policy path — the curve is steep, not flat, which is a term premium signal rather than a front-end one. It is supply. Deficit issuance and AI corporate issuance arriving into the same book at the same time, with Warsh signalling he does not regard hikes as his preferred instrument against a price level that keeps consumer expectations above 4%.
Which means the growth engine the administration is pointing at as the solution is, right now, a direct contributor to the numerator of the problem. Every basis point the AI issuance calendar adds to the term premium raises the rate at which $40 trillion refinances. The buildout has to deliver enough productivity to outrun the borrowing cost it is itself creating.
The inflation escape hatch is shorter than it looks
There is a fourth path nobody says out loud, which is to inflate the real value of the stock down. It worked after 1945 because the debt was long, the coupons were fixed, and the holders were captive.
None of those conditions holds. Weighted average maturity on marketable debt sits near 70 months. Bills alone are roughly $6.8 trillion of about $28 trillion marketable, repricing inside a year. TIPS index directly. Floating rate notes reset quarterly. An inflation surprise buys perhaps two years of real erosion before it is fully absorbed into the coupon and starts compounding the interest line instead. The Treasury’s own maturity profile has removed the option.
What actually settles this
Smetters made a point about credibility that is easy to skip past: markets can tolerate a bad fiscal trajectory, but they price a gap between what is promised and what arrives. The promise on the table is that growth closes the gap without legislation. The test is not GDP.
The test is the term premium — the spread between the thirty-year and the ten-year. If the growth story is being believed, that spread compresses as the buildout matures and issuance normalises. If it keeps widening while nominal GDP prints strong, the market has decided that stronger growth in this configuration means more supply rather than less, and the story has been priced as a story. Watch the spread, not the level. The level tells you about policy. The spread tells you what the bond market thinks of the plan.
