Industry reaction to the USD.AI and BSQ Capital joint venture is focused on one thing: the collateral. The hardware itself. Not cashflows, not operator credit ratings — the actual Nvidia H100s and their successors sitting in racks.
“That is the key,” said one Singapore-based infrastructure financier who declined to be named. “Traditional lenders look at a GPU cluster and see depreciation risk.
These guys look at it and see an asset they can seize and redeploy. It changes the underwriting math completely.” The joint venture targets $300 million in initial GPU-backed financings for operators across Asia-Pacific. There is a path to $1 billion if performance milestones are hit.
BSQ Capital, headquartered in Singapore, brings regional knowledge of what it calls “next-generation” digital infrastructure — AI data centres, interconnection facilities, the power and cooling systems that make them work. USD.AI brings the platform designed specifically to underwrite and manage compute assets.
The Asia-Pacific region is a pressure cooker for AI compute. Demand for training and inference hardware is surging from Tokyo to Jakarta. But the financing side has been a bottleneck.
Data centre lenders, accustomed to long-lived real estate assets, have been uneasy with GPU hardware that can lose value fast as new chips hit the market. “The financing gap in Asia-Pacific is real,” the financier said. “Operators have the demand.
They have the customer contracts. But they cannot get the bank debt to buy the hardware.
This structure is a direct response to that.” The structure is senior secured. That means the GPU hardware itself is the collateral — not the operator’s balance sheet, not the project cashflows, not a corporate guarantee. For lenders, it is a tangible asset class with a secondary market.
For operators, it is a way to fund enormous capital outlays without tying up equity. One regional AI startup founder, speaking on condition of anonymity, said the structure matters because it aligns incentives.
“If I miss a payment, they take the GPUs. That is brutal but it is also clean. There is no negotiation about restructuring cashflows.
The asset is the asset.” The joint venture is not the first attempt at GPU-backed financing in Asia, but it is the largest announced so far. The $300 million initial deployment is the headline number. But market participants are watching the $1 billion upsizing target more closely.
That would represent a significant scaling of the asset class in the region. BSQ Capital’s focus on “next-generation” digital infrastructure has increasingly meant AI.
The firm has been building a portfolio of assets that support AI workloads — not just data centres but the power and cooling systems that make them viable. The joint venture extends that thesis into direct hardware financing. USD.AI, based in New York, has built its platform specifically for this kind of underwriting.
The company does not disclose its underwriting criteria publicly, but the structure suggests a focus on hardware valuation, resale markets, and the operational viability of the clusters being financed. One banker who works on data centre financings in Southeast Asia said the joint venture fills a gap that traditional lenders have been unwilling to address.
“Banks here will finance the building. They will finance the power connection. But the GPUs?
That is a bridge too far for most credit committees. This is a different risk appetite.” The joint venture was announced June 30. No timeline for the first deployment was provided.
But the demand side is clear. Asia-Pacific is a battleground for AI compute capacity, and the financing mechanisms are finally catching up.
