Tech companies are no longer funding their investments in artificial intelligence (AI) primarily through equity or internal cash flows, but are increasingly turning to debt, prompting fixed income managers to reassess how much credit risk is building beneath the sector’s rapid expansion.
For much of the past decade, large technology companies sat comfortably in investment-grade portfolios. Strong balance sheets, conservative funding models and large net cash positions meant exposure to the sector was widely seen as low risk. That assumption is now being tested as capital expenditure accelerates and financing choices change.
As Al Cattermole, senior fixed income portfolio manager at Mirabaud Asset Management, put it: “Debt-funded capex introduces greater execution risk, raises leverage and increases the likelihood of rating pressure over time, particularly in a sector where the eventual winners and losers of the AI investment cycle are still far from clear.”
The turning point came late last year, when borrowing began to rise sharply to support data centre construction and related infrastructure. Brent Finck, senior research analyst for global investment grade at Neuberger, said the scale of issuance marked a clear break with previous funding patterns. “The fourth quarter of 2025 marked an inflexion point as companies turned to increased debt issuance to fund capex,” he said.
Around $93bn of AI-related debt was issued in 2025, accounting for more than 5% of total investment-grade supply for the year and almost three times the sector’s average annual issuance over the previous decade, according to Bank of America.
Borrowers, including Meta, Alphabet and Oracle, all raised capital to expand data centre capacity and secure long-term energy supply.
From a credit perspective, this raises two related concerns. The first is balance sheet pressure. Although earnings are still growing, free cash flow is increasingly being absorbed by capital spending, dividends and share buybacks, leaving less headroom than before.
Finck said that many issuers can still fund investment internally in aggregate, but “the cushion is narrowing and, for some issuers, capex will exceed earnings in 2026”.
The second concern is uncertainty over returns. Unlike previous technology investment cycles, the infrastructure being built for generative AI has no track record, making it difficult to judge how durable revenues will be over time.
Cattermole warned that this introduces a speculative edge to lending decisions, particularly given the pace at which the technology itself is evolving.
“Many of the data centres being built could quickly be rendered obsolete by technical improvements that make chips more efficient and reduce the requirement for so much capacity,” he said. As a result, there is a growing risk that today’s assets fail to generate the cash flows investors expect over their full economic life.
These risks are magnified in parts of the high-yield market, where transparency is more limited and contractual structures are harder to analyse.
Cattermole also highlighted the problems of circular arrangements (already known in the equity market), in which major players buy and sell computing power from one another, making it difficult for bondholders to assess true underlying demand.
Delays to construction or changes in contract terms could materially alter credit outcomes, yet are often hard to model in advance.
Concerns extend beyond public markets. Anton Dombrovskiy, fixed income portfolio specialist at T. Rowe Price, said the growing role of private credit in financing AI investment added another layer of complexity, particularly given the lack of disclosure relative to public bonds.
While he does not see a systemic threat at this stage, he warned that rapid growth in less regulated markets has historically been a source of hidden fragilities.
For now, managers stress that the situation is manageable rather than alarming. Data-centre lending still represents a relatively small portion of major credit indices and the largest hyperscalers retain significant balance-sheet capacity.
However, most expect issuance to rise further over the coming years, increasing dispersion between stronger and weaker borrowers.
That is already shaping portfolio positioning. At Neuberger, Finck said the firm remains constructive on the sector over the long term but is more cautious in the near term as supply builds and free cash flow comes under pressure.
“We are tactically cautious on the sector heading into 2026,” he said, citing elevated spending levels and the likelihood of high net new issuance.