How ByteDance Is Funding Its AI Infrastructure Push With a $29.6 Billion Loan

ByteDance is seeking financial flexibility for a far larger AI infrastructure push through a reported $29.6 billion offshore syndicated loan. The facility is notable not just for its size, but for its pricing: an opening margin of 68 basis points over the Secured Overnight Financing Rate (SOFR), a level reported to be below the roughly 85-basis-point margin on ByteDance’s prior offshore borrowing. 45

The key distinction is important: the loan’s stated use is mainly general corporate purposes. It can support AI-related investment, but it does not publicly earmark a fixed amount for a particular data center, chip order, or location. 1

The financing: a larger-than-planned offshore loan

ByteDance originally sought a $20 billion facility, then increased it to $29.6 billion after receiving more than $30 billion in lender orders, according to reports citing people familiar with the matter. That would make it Asia’s second-largest dollar-denominated borrowing of 2026. 12

Citigroup and JPMorgan coordinated the reported financing. The facility has a three-year tenor with options to extend it to five years. 34

At SOFR plus 68 basis points, the initial margin is unusually tight for a borrowing of this scale by a Chinese technology company. Combined with the order book exceeding the original target, that pricing is a market signal: participating banks appear willing to compete for exposure to ByteDance. 45

Still, the reporting requires a timing caveat. As of the September 3 reports, the transaction had not yet been signed, with banks still confirming allocations. 45

What the money can—and cannot—tell us about AI spending

The loan gives ByteDance a large pool of liquidity rather than a transparent AI-infrastructure budget. General-corporate-purpose financing can support equipment purchases, data-center construction, working capital, cloud services, and other company needs; it does not establish that the entire $29.6 billion will flow directly into AI infrastructure. 1

Reporting has said ByteDance is weighing capital spending of up to $70 billion for AI infrastructure in 2026. That figure illustrates the potential scale of its ambitions, but it is not the same as a disclosed capex budget, a signed loan-use schedule, or audited company guidance. 1

The practical effect of the facility is therefore financial headroom. ByteDance can fund a mixture of owned infrastructure and external services while retaining flexibility as hardware prices, supply conditions, and policy constraints change.

A hybrid compute strategy

ByteDance’s reported AI strategy spans proprietary infrastructure, specialized chips, and cloud-model access.

The company is developing custom CPUs for its servers and data centers amid chip-price increases and supply constraints, Reuters reported in May. It has also reportedly reached an agreement to procure millions of Qualcomm application-specific integrated circuits for AI data centers, indicating a strategy that does not depend on a single chip supplier.

At the same time, ByteDance has remained a major buyer of external AI and cloud services. Bloomberg reported that it has generally been Microsoft’s largest AI customer in recent years and is on track to spend more than $1 billion annually on Microsoft AI and cloud services, largely using OpenAI models.

Taken together, those reports suggest a hybrid model: build more internal capacity and chip capability while continuing to use outside cloud infrastructure and models where they remain strategically useful.

The Ulanqab claim needs careful handling

A report said ByteDance was in preliminary discussions to add between five and six gigawatts of computing capacity in Ulanqab, Inner Mongolia, over the next two years. But the available evidence does not establish finalized contracts, construction milestones, power approvals, or a firm early-2028 delivery date for that capacity.

There is also room for confusion because separate reporting described DeepSeek—not ByteDance—as pursuing roughly one gigawatt of computing capacity in Ulanqab.

For now, the most defensible interpretation is that ByteDance is exploring a potentially substantial expansion in the region, not that a 5–6 GW buildout has been independently confirmed as fully financed and scheduled.

What strong bank demand says about transparency

The lender response is a meaningful private-credit endorsement, but it is not public disclosure.

A syndicate can review financial information and other materials needed for internal credit approval. Reporting on ByteDance’s loan process said the company did not share audited financial statements with potential lenders, instead providing the financials required for credit approvals and requiring their deletion afterward.

That arrangement helps explain how banks can assess a borrower that does not publicly disclose the level of detail expected of a listed company. It may give lenders confidence in cash flow, repayment capacity, and deal protections, but outsiders do not receive the same information.

As a result, the loan’s size, low margin, and demand are best understood as a private-market credit signal. They support the view that lenders see ByteDance as an attractive borrower. They do not independently verify every reported AI-spending figure, data-center target, chip-sourcing plan, or long-term capacity claim.

Bottom line

ByteDance is using a huge, competitively priced offshore loan to expand its room to invest in AI. The facility’s increase from $20 billion to $29.6 billion and its 68-basis-point SOFR margin show unusually strong bank appetite. 14

But the financing is broad corporate capital, not a public blueprint for exactly how ByteDance will deploy up to $70 billion in capex. Its AI buildout may combine internal data centers, custom CPUs, Qualcomm chips, and continued Azure-based model access, while the most ambitious infrastructure targets remain reports rather than publicly documented commitments. 1