Three hundred million dollars buys a lot of GPUs. That is the bet behind a new joint venture between USD.AI and BSQ Capital Partners, announced June 30, targeting senior secured financing for AI compute operators across the Asia-Pacific region. The initial deployment is $300 million.
The stated ambition is to scale that to $1 billion, contingent on hitting performance milestones. The numbers tell a specific story about risk.
This is not unsecured lending. The financing is “senior secured and GPU-backed.” The hardware itself — Nvidia H100s and their successors — serves as collateral. For lenders, that turns an abstract bet on AI cashflows into a bet on a tangible asset with a resale market.
For operators, it means they can fund capital-intensive GPU clusters without tying up their own balance sheet equity. The structure is designed to bridge a gap in the market: traditional data centre lenders are often uncomfortable with how fast GPU hardware depreciates.
This joint venture is built to tolerate that depreciation cycle, because the collateral is the hardware itself. BSQ Capital Partners brings the regional presence. The firm is headquartered in Singapore and specializes in what it calls “next-generation” digital infrastructure — a category that increasingly means AI data centres, interconnection facilities, and the power and cooling systems that make them viable.
USD.AI brings the financing architecture, having built a platform specifically to underwrite and manage compute assets. The partnership is a division of labor: one side knows the region and the asset class; the other knows how to structure the deal. The Asia-Pacific region is the battleground.
Demand for training and inference hardware is surging from Japan to Southeast Asia. The report states that financing mechanisms to support that buildout have lagged behind.
That is the gap this venture is designed to fill. The $300 million is the first deployment, not the ceiling. The path to $1 billion depends on performance milestones — meaning the venture itself must prove the model works before it scales.
Consider what $300 million buys in GPU hardware. A single Nvidia H100 can cost $30,000 or more on the open market.
That is roughly 10,000 H100s at initial deployment, enough to power a significant training cluster. At the $1 billion target, the number rises to over 30,000 GPUs. Those are not small numbers.
They represent serious compute capacity for the region. The joint venture is designed for operators who need to stand up clusters quickly. The senior secured structure means lenders have first claim on the hardware if an operator defaults.
That is not a hypothetical concern. GPU hardware depreciates fast.
New chips arrive every 18 to 24 months. The resale value of an H100 drops the day a Blackwell or a Rubin or whatever comes next hits the market. The venture is betting that the financing structure can absorb that depreciation risk better than traditional lenders can.
BSQ Capital Partners focuses on “next-generation” digital infrastructure. That is a broad category, but in practice it means assets tied to AI data centres — the facilities themselves, the interconnection, the power and cooling.
The joint venture adds GPU hardware to that portfolio. It is a logical extension. AI data centres need GPUs.
The financing for those GPUs has been a bottleneck. This venture is designed to unblock it. The announcement is short on names and long on structure.
No specific operators are named. No specific projects are listed.
The joint venture is a framework, not a deal book. The $300 million is a target for initial deployment, not a committed fund. The path to $1 billion depends on hitting performance milestones that are not publicly defined.
That is typical for this kind of announcement. The details will come when the first deals close.
For now, the numbers are the story. $300 million. $1 billion. Senior secured. GPU-backed.
Asia-Pacific. That is the equation the joint venture is trying to solve.
Sources
- USD.AI and BSQ Launch $300M APAC JV | USD.AI
- USD.AI, BSQ Capital launch $300m AI infra financing venture in APAC
