AI's Debt-Fueled Buildout Echoes Telecom Bubble Risks, Says Growth Capital Founder
As major tech companies shift from equity to debt financing for AI infrastructure, a veteran venture capitalist and lender warns that the financing structure mirrors patterns that preceded the telecom sector's $1 trillion debt crisis.

The funding landscape for artificial intelligence development is undergoing a fundamental shift, moving away from venture equity and toward borrowed capital. For those who witnessed the telecom industry's collapse two decades ago, the parallels warrant close attention.

Stephen Hawking's "A Brief History of Time" contains a parable about infinite regress: a woman insists the world rests on a giant turtle, and when asked what supports that turtle, she answers "turtles all the way down." The AI sector's financial architecture increasingly resembles this problem. Valuations justify borrowing, borrowing creates capacity, and that capacity then validates the original valuations. Each layer depends on the one beneath it, with no solid foundation in sight.
Through the first three quarters of 2026, global venture funding totaled $679 billion, the highest recorded for that period, according to Crunchbase data. AI companies captured $102 billion in the third quarter alone, representing approximately 64% of worldwide venture investment. During that same quarter, 27 companies secured funding rounds of $1 billion or more, collectively accounting for roughly one-third of the total.
Equity financing, while risky, typically avoids fixed repayment schedules. When equity-backed companies underperform, investors absorb losses through valuation declines, and companies retain flexibility to pursue long-term objectives. Debt operates under different constraints.
Large-scale borrowing has now arrived. Meta issued $30 billion in bonds last October, one of the largest corporate debt offerings ever recorded, then returned to markets in April for an additional $25 billion, with lenders demanding higher compensation on the second transaction. Amazon, Alphabet and Oracle have each borrowed tens of billions. A joint-venture partner financed Meta's massive Louisiana data-center facility with a record $27 billion bond issuance rather than having Meta borrow directly. Morgan Stanley projects that major AI and cloud infrastructure companies will spend approximately $3 trillion on infrastructure through 2028, with only about half expected to come from internal cash reserves. The remainder must come from external sources, increasingly the bond market and private credit facilities.
The demand narrative
The most troubling echo concerns how demand is being projected. The current narrative asserts that data centers are essentially pre-leased before construction begins and that the primary risk is building insufficient capacity. This claim may prove accurate, but it circulates with the same universal acceptance that surrounded the telecom industry's famous forecast in 1999: that internet traffic was doubling every 100 days. Much of this narrative originates from the companies undertaking the construction themselves. Recent months have brought investor scrutiny to these claims.
Anthropic's recent public offering marks a turning point. Once companies of that scale begin reporting quarterly financial results, investors can test the demand narrative against audited figures rather than projections.
Some observers reasonably counter that the comparison breaks down because the largest borrowers may represent among the most profitable enterprises ever created, with cash flows far exceeding those of telecom carriers. Several AI companies are also reaching revenue scale faster than any previous businesses. The internet genuinely did transform everything, and the traffic growth was real. None of this, however, addresses the fundamental question a lender must pose: will cash arrive on the schedule debt repayment demands? Equity investors can generally tolerate delayed timelines; debt cannot. When borrowed funds build capacity faster than revenue materializes to service obligations, corrections typically occur through restructurings rather than gradual valuation adjustments.
Enduring lessons
This analysis does not predict that AI infrastructure will necessarily follow telecom's trajectory. The most important takeaway from the telecom collapse concerns capital structure—the factor determining which companies survived when demand eventually matched capacity, long after the underlying technology had proven its worth. The author has long maintained that growth-stage companies unnecessarily fear debt, a position that remains valid but conditional.
Debt functions effectively for borrowers whose obligations align with demonstrable revenue and for lenders who evaluate fundamentals rather than narratives. Everyone else faces harsh consequences. The borrowing supporting the AI expansion will ultimately be measured against this standard. Every assumption in the stack may hold true, but the lesson from the last crisis is clear: once investors begin questioning what the turtles stand on, they rarely cease asking until they reach something solid.
David Spreng is founder, co-CEO and co-CIO of Runway Growth Capital. He brings more than 30 years of experience as a Silicon Valley venture capitalist and growth debt lender, having been involved in forming and developing nearly 50 technology companies. He has appeared four times on the Forbes Midas List of top venture capitalists, co-founded Decathlon Capital Partners in 2010, served on the board of the National Venture Capital Association and has participated as a panelist at the World Economic Forum in Davos.


