ByteDance raises $29.6bn, Asia’s second-biggest loan this year, and pays less for it

ByteDance has landed a $29.6bn loan, the second-largest in Asia this year and roughly three times what the TikTok owner borrowed offshore in 2024. Banks put in more than $30bn of orders for a facility that started life at $20bn.

The terms were first reported by Bloomberg. Citigroup and JPMorgan are coordinating the deal, which runs for three years with an option to extend to five, and it has not yet been signed while banks confirm allocations.

The pricing is the part lenders will be arguing about. ByteDance is paying an opening margin of 68 basis points over SOFR, down from the 85 basis points it paid on its last offshore facility, which means it borrowed half again as much money at a materially lower price.

That happened in a year when Asia’s loan market has otherwise been slow. Demand for this particular borrower has moved in the opposite direction to the market around it.

The trajectory is steep. ByteDance raised $10.8bn from around 20 lenders in 2024, and was in preliminary talks over a record $20bn facility as recently as June, a target the final deal has now overshot by nearly half.

That 2024 facility was, at the time, the largest dollar corporate loan in Asia outside Japan. Two years on, the same borrower has taken almost three times as much and still finished second in the region, which says as much about the year as it does about ByteDance.

The stated use is general corporate purposes, which in ByteDance’s case currently means data centres. The company is weighing capital spending of up to $70bn a year on AI infrastructure, a figure that would put it in the same bracket as the American hyperscalers.

It is also spending under constraint. US export controls limit its access to Nvidia’s advanced chips, so the build-out runs through custom silicon on Arm and RISC-V, Qualcomm inference parts, and domestic Chinese suppliers, which makes the same amount of compute more expensive to assemble.

Some of that budget goes to a competitor. ByteDance has been paying more than $1bn a year to use OpenAI’s models through Microsoft Azure while financing the domestic hardware base intended to end that dependence.

Only one Asian borrower has raised more this year. SoftBank signed a $40bn bridge facility in March to fund its OpenAI position, and the two deals now bookend the region’s AI borrowing.

The structures are not comparable, though. SoftBank’s facility is a bridge against an equity stake, repaid when longer-term financing arrives, while ByteDance is a cash-generating business borrowing against its own operations.

That distinction is what the 68 basis points is pricing. Lenders are treating TikTok and Douyin revenue as the collateral, rather than a mark on a private company that could move.

The wider pattern is now well established, with Big Tech’s AI-related debt passing $350bn as companies fund data centres with borrowed money rather than cash flow. ByteDance joining at this size moves a large share of that borrowing outside the US.

What the money buys is a run at the frontier. ByteDance is training a ten-trillion-parameter model and expanding its data centre cluster in Inner Mongolia, neither of which is affordable on operating cash flow alone.

There is an unresolved figure in the reporting. The $70bn capex number sits awkwardly against the 160 billion yuan, around $22.7bn, that ByteDance was reported to be planning for 2026 earlier in the year, and the sources differ on whether $70bn is a decision or a scenario under consideration.

The company has not commented publicly on either number. It remains privately held, publishes no financial statements, and discloses capital spending only through the people who arrange its financing.

Which is the quiet significance of a syndicated loan this size. Twenty-odd banks now have a view of ByteDance’s books that its users, its advertisers and most of its regulators do not.