ByteDance has successfully secured a massive $29.6 billion syndicated loan, significantly expanding its initial target due to overwhelming demand from global banks.
Coordinated by major financial institutions including Citigroup and JPMorgan Chase, this borrowing marks Asia’s second-largest US dollar-denominated loan of the year.
ByteDance originally sought a $20 billion facility, but the scale was expanded after drawing over $30 billion in orders from participating lenders.
Despite borrowing nearly three times the amount of its previous $10.8 billion offshore facility in 2024, ByteDance secured a lower opening margin of 68 basis points over SOFR (down from 85 basis points previously). The facility features a three-year maturity with an option to extend to five years, designated for general corporate purposes and massive infrastructure scaling.
While officially designated for general corporate use, the historic influx of capital is heavily tied to ByteDance’s aggressive, multi-billion-dollar push into artificial intelligence.
The funds are expected to support massive data center expansions—including clusters in regions like Inner Mongolia—and secure the specialized hardware required to train advanced large language models.
ByteDance has stepped up its AI spending. Reuters reported in June that the tech firm was in talks with Shanghai-based Iluvatar CoreX to purchase AI chips for inference work and is also considering a similar deal with Baidu.
The new dollar-denominated facility is unsecured, meaning ByteDance is not pledging any asset or shares as collateral, reflecting how confidently lenders view the company’s creditworthiness.
“It is very rare to see such a mega loan unsecured,” said one of the sources. “The banks practically are counting purely on ByteDance’s name.”
Facing mounting pressure from both domestic and international competitors, ByteDance’s skyrocketing capital expenditure underscores how tech giants are increasingly relying on debt markets to fund the immense computing power needed for next-generation AI models
