Behind the $8.2 trillion AI feast, the banking industry is quietly entering the Asian GPU financing arena.

Banks are starting to get involved in GPU financing in Asia, an area previously dominated by riskier private lending funds, thus significantly expanding the funding pool for the next phase of the AI race.

In recent months, major international banks have played a key role in providing approximately $3.8 billion in GPU loans to three AI infrastructure providers: GMI Cloud, Zankore, and PaleBlueDot AI. According to sources familiar with the matter, Citigroup, JPMorgan Chase, Barclays, Deutsche Bank, Santander, and Sumitomo Mitsui Banking Corporation of Japan are currently evaluating GPU-linked loans.

This funding is crucial. PwC estimates that spending on data centers in Asia could reach $8.2 trillion by 2050, with the vast majority going towards hardware such as GPUs and servers.

International banks take the lead, and Asian banks follow suit.

Hundreds of data centers are being built across Asia, and developers are simultaneously raising funds to purchase chips. While GPU financing in the US has grown rapidly, the few such loans previously available in Asia were mostly secured through private credit.

Citigroup is acting as the sole debt advisor for Zankore's $3.1 billion loan in Indonesia, while JPMorgan Chase is acting as the placement agent for PaleBlueDot AI's $255 million credit facility. Approximately six bankers and financial advisors in Asia have indicated they are in talks with or aware of further GPU-related financing. Additionally, sources familiar with the matter revealed last month that GMI Cloud is in talks with banks and private lenders for a new $300 million loan to purchase chips for its data centers in Thailand.

According to sources familiar with the matter, among traditional lending institutions already involved in this area, global investment banks are currently the main players due to their experience in handling complex structures. One banker stated that some US banks have brought in headquarters experts to assess the chip's value. Citigroup, Barclays, and Deutsche Bank declined to comment, while JPMorgan Chase, Santander, and Sumitomo Mitsui did not immediately respond to requests for comment.

Asian banks are also becoming more active. According to Zankore Chairman Vikram Sinha, Singapore's United Overseas Bank (UOB), which co-underwrote Zankore's $3.1 billion loan with four other banks, is leading negotiations for a new round of financing. Sinha stated at a meeting on September 22 that Zankore needs continued financing to expand its AI data center capacity tenfold to 1 gigawatt, and partnering with banks is "the right way." "We know exactly the scale we need. We want to take the difficult path, working with banks and syndicates."

Valuing chips remains a challenge.

Eric Tan, a partner in banking and finance at Hogan Lovells Cadwalader, said, "As transaction volumes increase and borrowers demand more competitive pricing, the role of banks will become increasingly important." However, the rapid pace of technological iteration also exposes lenders to risks such as "rapid depreciation, technological obsolescence, and rental volatility" in GPU financing.

With banks becoming a more important source of funding, borrowers need to provide more substantial proof that project revenue is sufficient to repay the loan, and traditional lending institutions may require more conservative underwriting standards and higher debt service reserves.

Private lending has also not lowered its standards. Mike Arougheti, head of Ares Management, one of Asia's largest private lending firms, said that despite GPU financing being the area with the largest funding gap in the AI boom, Ares maintains high standards. "You have to prioritize risk appetite, not willingness to invest," he said at the Barclays Global Financial Services Conference in September. "At least nobody can clearly explain to me what the depreciation curve of this technology looks like." He also noted that these loans offer limited returns, typically only about 100 to 200 basis points higher than other AI infrastructure loans.

Repayment depends on the customer

Asian GPU loans largely follow the structure pioneered by CoreWeave, one of the earliest and largest users of this type of financing. In many transactions, loans are repaid with revenue from the sale of data center computing power, with customer contracts and the chips themselves typically serving as collateral. For lenders, the most critical issues are therefore the reliability of the customer and the length of the contract.

Deals backed by Nvidia are the easiest to get approved. In the GMI Cloud Taiwan project loan (provided by 14 banks) and the Zankore loan completed last month, Nvidia agreed to purchase all unsold computing power and cover customer contracts in case they fall through. In exchange, these two Nvidia Cloud Partner Program members will charge buyers higher rates than Nvidia promised and share revenue with Nvidia.

The debate surrounding the potential dangers of AI and whether safeguards should be erected for its development will continue to influence risk assessments. Eric Tan said, "It may take some time for the market to find equilibrium. Deal structures may tighten, sovereign backing may intervene, and deal patterns will evolve accordingly."

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.