The AI infrastructure arms race has officially breached corporate balance sheets. Tech conglomerates are pivoting from self-funded balance-sheet cash reserves to colossal corporate debt syndicates. Just months after locking down $10.8 billion from 20 lenders in September 2024, ByteDance expanded its credit line dramatically. The TikTok owner closed a staggering $29.6 billion, three-year loan facility from nearly 30 international banks to underwrite overseas data centers and AI capacity expansion.
Coordinated by Citigroup and JPMorgan, the syndicate drew heavy interest across Chinese, US, European, and Singaporean institutions. ByteDance originally floated a $20 billion target, but lender appetite pushed the final unsecured package to nearly $30 billion. Notably, Chinese lenders supplied over 60% of the capital, and ByteDance pledged zero collateral—neither hard assets nor corporate equity.
Compute Scale and Infrastructure
The transaction represents the second-largest dollar-denominated facility in Asia this year, trailing only SoftBank's $40 billion raise in March. It includes an option to extend the three-year maturity by an additional two years, giving ByteDance essential runway as global data center construction drags on.
This capital is earmarked directly for compute survival: accelerating inference and training throughput for the Doubao foundation model family while navigating tighter Western export curbs. Hardware constraints remain the operational bottleneck. In response, ByteDance has diversified beyond Western supply chains, negotiating silicon procurement with Shanghai-based Iluvatar CoreX and sourcing alternative compute from Baidu.
"ByteDance is competing with local hyperscalers on AI data centers and with global hyperscalers on multimodal AI models,"
noted Lian Jye Su, chief analyst at Omdia, highlighting how ByteDance must fund double CapEx pressure across both local cloud infrastructure and frontline foundation models.
Financing Reality Check
A three-year unsecured loan creates an aggressive depreciation clock. Data centers require multi-year buildouts, while GPU clusters face rapid hardware obsolescence and fierce unit-cost compression against OpenAI, Google, and Meta. Without direct cash monetization of enterprise compute and Doubao API margins, servicing a $30 billion debt stack will severely test ByteDance's operational leverage before the first extension option even kicks in.