Debt-Backed AI Financing Has Surged

Rising debt levels and capital expenditure projections highlight risks for infrastructure-heavy AI expansion.

Updated on Sept. 25, 2026 in Artificial Intelligence

Bold flat-color editorial illustration of a heavy industrial cooling pipe, representing the structural financing risks within AI data center infrastructure. — Rising debt levels linked to AI infrastructure investments have reached 60% of industry funding, raising concerns about the long-term viability of current capital expenditure projects. AI Illustration. Upload story photo >

Live Poll

Do you believe the current pace of investment in artificial intelligence is creating a financial bubble?

Venture capitalist Paul Kedrosky has likened the current surge in debt-backed AI financing to the 2008 financial crisis, noting that such financing now accounts for more than 60% of total industry funding. This increase coincides with massive capital expenditure projects, including Amazon's projected $220 billion outlay for 2026.

Why it matters

Investors are expressing uncertainty regarding the long-term ability of AI data centers to generate sufficient revenue to cover construction costs. This disconnect between massive infrastructure spending and actual output growth threatens the stability of existing five-year financing structures.

Required returns for some data-center projects have climbed to 10% to 12% due to higher Treasury yields and credit spreads. While autonomous coding agents have increased raw coding activity by 240% in a study of 500,000 GitHub developers, productivity gains for actual software releases are limited to 30%.

The players

Paul Kedrosky

A venture capitalist who tracks financial risks associated with the rapid scale-up of AI infrastructure.

Amazon

A global cloud provider and e-commerce giant currently directing massive capital expenditures toward AI and data-center capacity.

NBER

The National Bureau of Economic Research, a research organization that conducts studies on economic activity and technical productivity.

The details

Companies are leveraging AI to accelerate software production, investment decks, and pitch transactions, often funding the necessary cloud capacity through heavy capital expenditures. However, an NBER study indicates a significant gap between activity and results: while individual coding tasks rose 240%, project-level productivity reached only 80% and realized release output remained at 30%. This suggests that current AI infrastructure investments may be over-extending relative to the functional output realized by developers.

In 2025, debt-backed AI financing comprised 15% to 20% of the market total.
Amazon projects capital expenditures of $220 billion for the 2026 calendar year.
Prediction markets assign a 10% probability to an AI industry downturn by December 31, 2026.
Refinancing difficulties are expected for five-year debt structures in 2029.
Additional pressure on five-year financing structures is anticipated in 2030.

The Tech Race

This growth in debt-backed financing follows a pattern established by the 2008 financial crisis, where rapid, leverage-heavy investment preceded market corrections. Investors are currently weighing this debt accumulation against the long-term utility of the massive data-center capacity being built by companies like Amazon.

Users should watch for shifts in corporate capital expenditure reports and credit spreads that may impact the sustainability of current AI service pricing. Industry-wide changes to software development costs will likely depend on whether the 30% release productivity gain can be meaningfully improved in future software versions.

The takeaway

The gap between raw AI coding activity and actual software releases suggests that current infrastructure build-outs may face a significant reckoning. Investors and industry participants should monitor the 2029-2030 refinancing cycle for signs of systemic stress in AI infrastructure debt.