Global technology companies have issued debt at a record pace in 2025. Data reveals that an intensifying race to build artificial intelligence (AI) capacity is forcing even cash-rich firms to borrow heavily for infrastructure such as data centres and semiconductor procurement.
According to Dealogic data, global tech companies issued $428.3 billion of bonds through the first week of December. U.S. firms accounted for the vast majority, issuing $341.8 billion.
European and Asian tech companies followed, issuing $49.1 billion and $33 billion, respectively. This marks a significant shift for an industry traditionally reliant on internal cash flows.
Michelle Connell, president at Portia Capital Management, commented on the trend. She stated debt-funded AI capital expenditure reflects a structural shift. Rapid technological obsolescence and short chip lifespans now force continuous reinvestment.
Heavy borrowing has begun to affect corporate financial health. A Reuters analysis of over 1,000 major tech firms shows their median debt-to-EBITDA ratio rose to 0.4 in September. This is nearly double the level seen during the 2020 debt surge.
Big Tech’s dominance of AI infrastructure is shrinking as a host of new players invest in the data center gold rush https://t.co/R9jmso3s51
— Bloomberg Green (@climate) December 22, 2025While leverage remains moderate, the increase signals that debt is rising faster than earnings. This poses a risk if future AI cash flows disappoint.
Credit markets are showing signs of caution. Five-year credit default swap (CDS) spreads for major firms like Oracle and Microsoft have widened notably in recent months. This indicates rising investor-perceived risk.
Read: Forbes 2025 AI Boom Billionaires: Meet the Tech Titans Behind Artificial Intelligence Wealth
Scott Bickley, an advisory fellow at Info-Tech Research Group, expressed concern. He views the trend as the result of an “overheated marketplace.” It creates a “go big or go home” narrative focused on the stock price.
“This is neither sustainable nor repeatable as a permanent shift in operating modes for the hyperscalers,” Bickley stated.
