• AI overvaluation is now the biggest concern among credit investors, surpassing traditional risks like inflation or geopolitics.
  • Investment-grade respondents citing AI as their top risk surged to 23% in early February 2026, up from 9% in December.
  • Investors forecast hyperscaler issuance to reach $285 billion in 2026, amid record fears of excessive AI capex.

For the first time, an AI bubble has become the primary worry for credit investors, according to a Bank of America (BAC) Global Fund Manager Survey from early February 2026. The poll of 190 managers overseeing $512 billion revealed that a record near-third of participants view corporate AI investments as "overinvesting" since data collection began in 2005. Only 20% now favor more capital expenditure, down from 34% previously, signaling a sharp shift in sentiment.

Efforts to gauge market stability have hit a snag as corporate bond spreads for U.S. hyperscalers reached three-year highs, with private credit strains becoming evident. For instance, Blue Owl (OWL) recently suspended retail redemptions, according to people familiar with the matter. Delinquencies remain low, but lower-rated software loans—comprising 26% of commitments—pose significant risks if capital injections falter. Without sustained funding, some firms could face restructuring pressures.

"What institutional investors are really focused on is regulatory stability and sustainable growth," said one anonymous fund manager, echoing concerns from the survey. Attempts to reach hyperscaler representatives for comment were unsuccessful, but sources indicate they signal no slowdown in spending despite warnings. The AI-adjacent debt totals approximately $1.2 trillion, dwarfing bank commercial and industrial commitments of $450 billion, which account for 13% of large bank totals—up from 9% in 2015.

Bank exposure averages 9% of Tier 1 capital outstanding but 25% committed, with data center lending at $14.9 billion through the third quarter of 2025. The Chicago Fed has warned of tail risks in AI lending, including indirect non-bank financial institution exposure, adding to the unease. In a slight tone shift, some analysts note that fund inflows may offset potential bond weakness, keeping markets on a "tightrope," as BofA's Michael Hartnett described it.

Historically, AI capex enthusiasm peaked post-2024, shifting the risk narrative away from inflation and geopolitics. This echoes dot-com era overinvestment, though current delinquencies are low, at less than 50 basis points for software loans. Looking ahead, short-term risks include potential credit events from hyperscaler borrowing if returns disappoint, with BofA downgrading semiconductors to underweight due to stretched valuations. Long-term, tail risks involve borrower defaults drawing on lines beyond the 9% Tier 1 capital threshold, which could slow AI buildouts and dip private credit returns to 5.4% in 2026.

Correction: An earlier version misstated the survey date; it was conducted from February 6-12, 2026.