ByteDance seeks $20B offshore loan to fund $70-100B annual AI capex race

<cite index="62-1,62-2">ByteDance, the developer of TikTok, is in preliminary talks with banks for a borrowing of about $20 billion, what would be the firm's largest offshore loan yet at a time when it's boosting investments in artificial intelligence</cite>. <cite index="62-2">The Beijing-based social media conglomerate has approached banks for the new-money loan, which could carry a three-year tenor, with the option to extend to as long as five years</cite>. <cite index="63-3">ByteDance has joined an intensifying global push into AI, with the Chinese tech firm weighing plans to boost capital spending to as much as $70 billion this year to expand its data centers and other AI infrastructure, with the figure potentially climbing to $100 billion next year if economic and business conditions remain favorable</cite>.

<cite index="63-3">Four US hyperscalers — Amazon, Alphabet, Microsoft and Meta — are planning as much as $725 billion in capital spending this year, largely focused on AI data center equipment; meanwhile SoftBank recently secured a $40 billion bridge loan to fund its investment in OpenAI</cite>. <cite index="66-2,66-3">Qualcomm is reportedly in negotiations to provide custom chip design services to ByteDance, potentially targeting mass production of video processing units by year-end; meanwhile, the chips discussed will be partly based on high-speed connectivity technology from Alphawave Semi (acquired by Qualcomm), involving video processing unit design, with a goal of starting mass production by the end of the year</cite>.

<cite index="63-4">ByteDance has long been regarded as a prime IPO candidate, though it's given few signs it's in a hurry to do so</cite>. For infrastructure architects: ByteDance spending $70–100B annually on AI capex (while pursuing custom silicon design) is not an isolated bet on one model or region—it's structural positioning to own the entire inference stack for Chinese commerce and export markets. The $20B loan signals serious downstream infrastructure capital needs and a push toward in-house silicon. This matters because it changes the global HBM and GPU utilization math for 2027–2028.