ByteDance Closes $29.6 Billion Loan, Asia's Second-Largest Dollar Borrowing of 2026, to Fund AI Infrastructure Across Southeast Asia

ByteDance signed a $29.6 billion three-year unsecured syndicated loan in early September after orders exceeded $30 billion, allowing the TikTok parent to upsize from an initial $20 billion target, according to people familiar with the matter reported by Bloomberg, Reuters, and multiple outlets. The facility, coordinated by Citigroup and JPMorgan with participation from nearly 30 banks including ICBC, Bank of China, HSBC, and China Construction Bank, ranks as Asia's second-largest dollar-denominated loan of 2026 behind only SoftBank's $40 billion raise in March. Chinese banks provided roughly $18.9 billion or 64 percent of the total at 68 basis points over SOFR, tighter than the roughly 85 basis points ByteDance paid on its 2024 offshore borrowing. Proceeds are officially earmarked for general corporate purposes, though sources familiar with the matter said the loan primarily supports overseas AI infrastructure expansion in Southeast Asia, where ByteDance is building data center capacity and securing compute resources outside China. The tenor is three years with an option to extend to five. The deal reflects lender confidence in ByteDance's creditworthiness and the appeal of AI-related financing even as US export controls on advanced semiconductors push the company toward domestic Chinese chip suppliers for training infrastructure. ByteDance has been shifting semiconductor orders to suppliers like Iluvatar CoreX for inference work and has mapped out plans for up to $70 billion in AI capital expenditure for 2026, with the potential to reach $100 billion in 2027, according to separate Bloomberg reporting. For the global AI infrastructure market, the ByteDance loan underscores the capital requirements and geographic diversification strategies driving the sector, as both Western and Chinese developers race to secure compute capacity and data center footprint ahead of regulatory and supply-chain constraints.