ByteDance is pursuing a record offshore loan of $20 billion to finance its AI development.
ByteDance is currently in initial discussions with banks regarding an offshore loan of approximately $20 billion, which would mark the largest amount the owner of TikTok has ever raised, according to a report by Bloomberg on Wednesday, citing sources knowledgeable about the talks.
The company is reportedly looking for a facility with a three-year term, with the possibility of extending it to five years. ByteDance has not yet provided a response to a request for comments. The mentioned figure stands out and is considerable.
A loan of $20 billion would almost double the roughly $9.5 billion financing that ByteDance secured in September 2024, which at that time was the largest dollar-denominated corporate loan in Asia outside of Japan. That previous arrangement, managed by Citigroup, Goldman Sachs, and JPMorgan, also featured a three-year term with a five-year extension option, with part of the proceeds being used to refinance an existing facility rather than for new expenditures.
Underlying this significant figure is a common factor affecting many substantial financings in the sector at present: the acquisition of chips and the infrastructure to house them. ByteDance has laid out initial plans for capital expenditures totaling around 160 billion yuan, or about $22.7 billion, in 2026, primarily focused on AI infrastructure.
A loan of this magnitude would meaningfully contribute to that effort and do so in dollars, which is important for a company sourcing computing and components that are largely priced outside of China. This procurement issue has become particularly challenging for ByteDance, partially due to geographical factors. US export controls have restricted access to Nvidia’s most advanced accelerators, prompting the company to seek out several alternative solutions simultaneously.
ByteDance is developing its own data-center CPUs through parallel Arm and RISC-V pathways, partly due to rising server-processor prices from Intel and AMD. The company has established a collaboration with Qualcomm for application-specific inference chips and is considering orders from several smaller Chinese suppliers, including Biren, MetaX, and Iluvatar CoreX. These options are costly, and prices show little sign of decreasing.
Additionally, the origin of the models themselves adds complexity to the situation, an issue the loan does not directly address. ByteDance is Microsoft’s largest AI customer, projected to spend over $1 billion annually purchasing OpenAI models via Azure in a market where OpenAI does not sell directly.
Thus, the company appears to be acquiring Western frontier models while simultaneously building a domestic hardware base, and a $20 billion loan would serve to finance both aspects of that strategy. The terms currently reported are minimal, which is typical at this phase. Initial discussions do not bind ByteDance to any commitments, and the total amount may fluctuate, with no confirmed lending group, pricing, or closing timeline yet established.
According to Bloomberg’s sources, ByteDance is exploring lending options rather than having finalized agreements. For a private company of ByteDance’s size, lacking a public listing to raise equity, a syndicated bank loan is among the few avenues available to quickly amass tens of billions of dollars, and historically, banks have been keen to lend against one of the world’s most lucrative internet businesses.
If the deal reaches $20 billion, it would set a new standard for offshore corporate borrowing in the region, marking the second time in under two years that ByteDance has reset such a record. The key question remains whether the amount will hold through negotiations. However, the direction of the intended spending is clearly defined.
Other articles
ByteDance is pursuing a record offshore loan of $20 billion to finance its AI development.
According to Bloomberg, ByteDance is in preliminary discussions for an approximately $20 billion offshore loan, nearly twice its record for 2024, due to increased spending on AI infrastructure and chips.
