ByteDance Secures $29.6 Billion Loan from 28 Banks
According to Bloomberg, ByteDance, the parent company of TikTok, has signed a loan agreement totaling $29.6 billion with a syndicate of 28 banks to support its AI business expansion. The loan was recently finalized.
The initial planned size of the loan was $20 billion, but it was expanded to $29.6 billion due to strong demand from banks; the loan term is 3 years, extendable to 5 years, with an interest rate of SOFR (Secured Overnight Financing Rate) plus 68 basis points. The financing is for "general corporate purposes," covering AI-related investments.
Chinese banks dominate the syndicate, with 15 Chinese banks collectively lending about $18.9 billion, accounting for approximately 64% of the total loan amount. Industrial and Commercial Bank of China is the largest single lender with $3 billion, followed by Bank of China with $2.5 billion and China Construction Bank with $1.5 billion. Among foreign banks, HSBC is the largest foreign lender with $1.5 billion.
This scale makes it the second-largest dollar financing in Asia this year, second only to the $40 billion bridge loan obtained by SoftBank Group in March; however, the financing costs of the two loans differ significantly—SoftBank's bridge loan had an initial interest rate of about 250 basis points above the benchmark rate, nearly four times that of ByteDance's loan, reflecting different assessments of the creditworthiness of the two companies by banks.
ByteDance has previously disclosed that its capital expenditure plan for 2026 may increase to about $70 billion, primarily directed towards AI infrastructure such as data centers. In March, it was reported that ByteDance plans to build an AI data center in Malaysia, purchasing about 500 NVIDIA Blackwell systems and approximately 36,000 B200 chips, with hardware investments expected to exceed $2.5 billion. This move is interpreted as a common path for Chinese tech companies to circumvent U.S. export restrictions on advanced chips by operating through third countries.
For the participating Chinese banks, this is an opportunity to compete for quality loan assets from leading tech companies amid weak domestic credit demand, with state-owned banks like ICBC and Bank of China further solidifying their financing relationships with ByteDance. For ByteDance, securing nearly $30 billion in funding at a much lower cost than SoftBank means it can more comfortably advance its overseas AI infrastructure construction without relying on equity financing or high-interest bridge loans like its American counterparts such as OpenAI and Anthropic.
ABAB AI Insight
This financing by ByteDance is not its first large-scale debt issuance—previously in 2024, the company completed $10.8 billion in financing through about 20 banks, and this $29.6 billion loan is nearly three times larger, reflecting its rapidly expanding funding needs as its AI strategy progresses. Meanwhile, ByteDance has been reported multiple times over the past year to be establishing AI data centers in Malaysia and other Southeast Asian countries, procuring high-end NVIDIA chips to circumvent U.S. export controls. This "overseas computing power construction" approach is consistent with the strategies of other Chinese tech giants like Alibaba and Tencent in response to chip bans in recent years.
The flow of this $29.6 billion funding clearly points towards AI infrastructure—ByteDance has disclosed that its capital expenditure could reach $70 billion by 2026, primarily for data center construction, with the Malaysian project alone requiring over $2.5 billion in hardware procurement. Notably, the loan was obtained at a premium of 68 basis points, significantly below the market average, indicating banks' confidence in ByteDance's cash flow and asset quality. The core motivation for mobilizing funds is to lock in overseas computing power resources at relatively low-cost debt leverage before the global AI computing power competition intensifies.
Compared to the high-interest bridge loans that SoftBank took on to increase its stake in OpenAI (with rates around 250 basis points above the benchmark rate, and later seeking to increase it to a maximum of $40 billion), ByteDance's financing costs are significantly lower. This difference is akin to the divide in traditional industries where "cash flow stable giants borrow at investment-grade rates, while leveraged investors reliant on single asset valuations can only accept high-interest terms." Currently, global AI infrastructure financing is in a phase of "record debt scale, but sharply diverging bargaining power among different companies," with ByteDance benefiting from the stable cash flow of its cash cow businesses like TikTok, placing it in a more advantageous position in this differentiation.
This essentially reflects a capital concentration effect—when AI infrastructure investment scales have ballooned to hundreds of billions of dollars, only companies with stable cash flows and healthy balance sheets can obtain large-scale credit from the banking system at low costs, while more companies reliant on equity pledges or asset mortgages are forced to accept higher financing costs. Mechanically, this differentiation in financing costs will further amplify the speed advantage of leading companies in expanding computing power, making the AI infrastructure race gradually evolve into a competition of financing capabilities rather than purely a competition of technology or products.
ABAB News · Cognitive Law
- Companies with stable cash flows borrow the cheapest; those without cash flows can only mortgage the future.
- In the end, the computing power competition is about whose loan interest rate is lower.
- Banks don't care about the stories you tell; they only look at your cash flow from the last quarter.