As AI infrastructure spending climbs to unprecedented levels, market attention has turned from how much the technology giants are investing to how they are paying for it. Bond issuance, private credit deals, long-term leases, and several trillion dollars in off-balance-sheet commitments have raised a familiar question: is this another leveraged capex boom headed for a debt reckoning, like the railway, real estate, and telecom cycles before it?

The aggregate data, however, point to a more nuanced picture. Relative to the size of the economy, US nonfinancial corporations are carrying an unusually low net interest burden, while the federal government's interest burden has risen sharply. The scale of the AI commitments also overstates the leverage involved, since roughly 85% are undelivered purchase and lease contracts rather than concealed debt. The key question is therefore not the headline size of AI financing, but whether these commitments are creating a material increase in corporate debt-service costs and whether the returns generated by the resulting infrastructure are sufficient to cover its financing costs.

I. The Same Rate Cycle, Two Very Different Balance Sheets

Corporate and government net interest payments moved together for most of the postwar period, rising during the inflationary 1970s and peaking during the high-rate years of the late 1980s and early 1990s, when corporate net interest payments approached 3% of GDP.

However, the two have diverged sharply since the pandemic. Government net interest has surged to near-record highs, while corporate net interest has fallen below 1%, levels last seen in the 1960s. The divergence reflects three channels:

Different monetary-policy transmission channels and debt maturity structures: During the pandemic, the Federal Reserve cut rates to near zero and used quantitative easing and corporate credit facilities to compress corporate bond yields and credit spreads. Companies responded by issuing long-term, fixed-rate debt at low interest rates, locking in their financing costs. By contrast, the federal government issued large amounts of short-term Treasury bills to finance fiscal expansion. After the Fed began raising rates in 2022, existing corporate debt costs adjusted gradually, while the government had to refinance short-term debt at increasingly higher rates. Higher rates therefore passed through much more quickly to government interest expenses than to corporate debt-servicing costs.
Higher corporate interest income offset more than half of interest expenses: Corporate cash balances were elevated because strong operating profits generated substantial internal cash flow, while the 2017 US tax reform reduced the tax disincentive to repatriate foreign earnings, prompting some companies to bring funds back to the United States. Companies held these funds in cash, bank deposits, money market funds, short-term Treasury securities, and other financial assets. As rates rose after 2022, these assets generated more interest income, offsetting interest payments on existing debt and lowering corporate net interest payments. In 2024, corporate interest income totaled approximately $471.9 billion, offsetting nearly 60% of the $790.9 billion in corporate interest expenses.
Tax cuts benefited corporations while expanding the government deficit: In addition to the TCJA’s 21% corporate tax rate and accelerated depreciation, the One Big Beautiful Bill Act (OBBBA), enacted in 2025, restored and made permanent several business tax benefits, including 100% bonus depreciation, immediate expensing of domestic R&D, and changes to interest-deduction rules. Alongside pandemic relief and fiscal support, these measures strengthened corporate cash flow and debt-servicing capacity, reducing reliance on new financing. At the same time, tax cuts and fiscal support widened the federal deficit, increased public debt, and pushed government interest expenses higher once rates rose.

The corporate sector therefore entered this investment cycle with an unusually low debt-service burden and an unusually large interest-earning cash position. The fixed-rate advantage fades as maturities roll over, and AI capacity is being financed at current rates rather than 2020 rates. Even so, the buildout is starting from the strongest corporate debt position in decades.

II. The Economics Supporting the AI Buildout

Whether technology giants can sustain their AI-related borrowing ultimately depends on whether those investments generate enough cash flow to service the debt. For AI, that requires two conditions: demand that keeps growing into the capacity being built, and returns on that capacity that stay above the cost of the capital funding it. On current evidence, both hold.

First, end-market demand is compounding. Revenue at the leading frontier labs is growing faster than at almost any company in history, and the growth now extends well beyond consumer subscriptions into coding agents and enterprise deployment. Falling token prices have not capped this growth, as each round of price cuts on model access has produced a multiple-fold jump in token consumption, which is the Jevons Paradox at work as cheaper compute expands total demand. This surge in compute demand is also visible in hardware markets. Shortages of the newest GPUs are forcing operators to extend leases on older accelerators and redeploy them from frontier training to inference. Rising GPU rental prices across several chip generations provide direct evidence that demand for compute exceeds available supply.

Second, AI investments are earning more than they cost to finance. Large cloud providers generate returns on invested capital (ROIC) of around 20%, well above their weighted average cost of capital (WACC) of roughly 11–13%, even after recent increases in borrowing costs. This gap means that new AI infrastructure, including debt-funded capacity, can still create value. Estimates suggest that hyperscaler AI data centers can pay back their investment in three to four years, compared with hardware depreciation periods of about six years. As long as ROIC remains above WACC, additional borrowing should remain manageable.

Read more: $3.5 Trillion in Shadow Financing: The Hidden Financial Architecture Behind the AI Boom

MM Research Insights

The divergence in net interest burdens puts the AI debt debate into perspective. Fears of a corporate debt crisis overstate the significance of headline commitment figures. With nonfinancial corporate net interest near historic lows relative to GDP, strong operating cash flows, and returns on invested capital well above the cost of capital, the corporate sector is funding this buildout from a position of unusual financial strength. A lower corporate interest burden also provides a buffer against potential spillovers from stress in the Treasury market into corporate credit markets. As long as AI demand keeps compounding, AI-related corporate debt is not a systemic concern.

The more critical monitoring point is the surge in government net interest payments. Infrastructure cycles have historically turned when financing conditions tightened. A record federal interest burden and persistent deficits could keep long-term yields elevated, raising financing costs across the economy and compressing the spread between investment returns and the cost of capital. At the same time, if AI investment translates into sustained productivity and corporate earnings growth, stronger economic growth could eventually improve public-debt dynamics through higher tax revenues and a wider growth–interest-rate differential (g − r).

In our view, the primary financial risk to the AI cycle is therefore not corporate overextension, but whether persistently high government borrowing and interest costs eventually raise the economy-wide cost of capital enough to erode the returns supporting the buildout.

Spotlight

More Spotlights...