From $10.8 Billion to $29.6 Billion in Two Years: ByteDance’s Record Loan Is the AI Debt Era in One Number

When ByteDance last tapped the global loan market in 2024, it walked away with $10.8 billion from about 20 lenders — a record for the Chinese internet sector at the time. Two years later, the TikTok and Doubao parent has nearly tripled that figure in a single transaction. According to Bloomberg News and Reuters, ByteDance has secured a $29.6 billion syndicated loan from nearly 30 banks, making it Asia’s second-largest US dollar-denominated borrowing of 2026. The Straits Times reports the total facility size at $37.6 billion including refinancing components.

The purpose is unambiguous: funding an artificial intelligence infrastructure buildout that could reach $70 billion this year — data centers, computing systems, and the training runs behind models like Doubao and Seedance. In an industry where OpenAI, Anthropic, and the hyperscalers dominate headlines, the quiet fact is that the world’s most aggressive AI spender outside the United States is now also one of its biggest debt issuers.

The Anatomy of a Mega-Loan

The numbers themselves tell the story of a company that banks cannot afford to say no to.

The size. ByteDance originally sized the deal at $20 billion. When orders topped $30 billion — roughly a 1.5-times oversubscription — the company upsized to $29.6 billion in new money. Reuters reports nearly 30 banks joined the syndicate, spanning international and Chinese lenders. Citigroup and JPMorgan coordinated the deal.

The price. The loan carries an opening margin of just 68 basis points over the Secured Overnight Financing Rate (SOFR) — among the lowest margins for a Asian private-sector borrower in recent memory, and notably tighter than the roughly 85bp ByteDance paid on its 2024 facility. When lenders accept a lower spread on a bigger check, they are making a statement about the borrower’s perceived credit quality — and about how badly they want the relationship.

The structure. The tenor is three years, with extension options stretching to as long as five. That is a deliberately short-dated structure for capital that will fund assets — GPUs and data centers — that have multi-year depreciation curves. It is a bet that ByteDance’s operating cash flow, reportedly amplified by roughly $50 billion in 2025 profit, will comfortably service the debt long before maturity.

The trajectory. As TechTimes notes, ByteDance’s borrowing history has climbed from $1.335 billion in 2019 to $29.6 billion in 2026 — a 22-fold expansion in seven years, all while remaining privately held.

Why ByteDance Is Borrowing Instead of Just Spending

ByteDance does not need rescuing. The company is one of the most profitable enterprises on the internet, with Doubao alone serving roughly 300 million monthly users and its advertising machine generating cash at a scale few companies on earth can match. So why borrow $29.6 billion?

Three reasons stand out.

First, the capex curve has gone vertical. In May, Bloomberg reported ByteDance was weighing AI capital expenditures of as much as $70 billion for 2026 — more than double its 2025 spending, and dramatically above the roughly $23 billion plan floated earlier in the year. Reports suggest the budget allocates on the order of 230 billion RMB to GPUs alone, plus 120 billion RMB for CPU and other equipment. Even a company generating $50 billion in annual profit cannot fund that from cash flow without strain.

Second, debt preserves optionality. ByteDance remains private, and an IPO — long rumored, never confirmed — would be complicated by geopolitical scrutiny in Washington. Borrowing in the syndicated loan market lets the company raise tens of billions of dollars without disclosure obligations, equity dilution, or a public listing. As one analysis put it, this is the AI race “funded by debt, not disclosure.”

Third, the competitive clock is ticking. Alibaba has pledged hundreds of billions of RMB to AI over three years. Tencent spent roughly $12 billion and Alibaba $19 billion last year — figures ByteDance is now spending into the stratosphere. In chatbots (Doubao) and video generation (Seedance 2.5, which can produce 30-second 4K clips in a single take), ByteDance holds an early lead in key segments, and management has clearly decided that lead is worth defending at almost any cost.

The Bigger Picture: AI’s Debt-Financed Era

The most important context for this deal is that ByteDance is not an outlier — it is a data point in a rapidly forming pattern. Goldman Sachs noted in August that hyperscale technology companies issued $108 billion of debt globally in 2025, and the pace has accelerated since. CoreWeave, Oracle, Meta, and Microsoft have all turned to credit markets this year to fund data center construction, and PwC now sizes the total data centre capital mobilization through 2035 at $31.6 trillion.

For two decades, the defining feature of big tech was net cash — balance sheets so rich they could buy back stock by the hundred-billion. The defining feature of the AI era is the reverse: even the richest companies in the world are borrowing to keep up. Compute has become the scarce strategic resource, and no player can convert profit into GPUs fast enough organically.

ByteDance’s loan is also a signal about where lenders think the AI value chain is durable. Banks extended nearly $30 billion at 68 basis points over SOFR — pricing usually reserved for blue-chip corporates — to a privately held Chinese company operating under persistent US regulatory pressure around TikTok. Lenders are effectively underwriting the durability of ByteDance’s advertising cash flows and its AI infrastructure bet simultaneously. That is a remarkable vote of confidence, and it would not have been available at any price two years ago.

There are risks, of course. A three-year tenor on multi-year infrastructure means refinancing exposure if credit markets tighten. Chip export controls could strand parts of the buildout — ByteDance has already navigated the odd arrangement of receiving Nvidia H200s that Beijing confines to Hong Kong. And $70 billion of annual capex assumes the returns on AI inference and advertising keep compounding, a bet whose payoff timeline remains genuinely uncertain.

But the direction is unmistakable. The AI arms race has entered its leverage phase. ByteDance just wrote the largest single line item so far.