USD.AI Closes $128.9M Asset-Backed Deal to Deploy 32 NVIDIA GB200 Systems in British Columbia
The transaction — USD.AI's largest loan originated to date — signals that asset-backed lending is becoming a serious financing mechanism for high-density AI infrastructure.
On September 23, 2026, USD.AI announced a $128.9 million asset-backed financing facility — the largest single loan originated through its platform to date. The capital is earmarked for one purpose: deploying 32 NVIDIA GB200 NVL72 systems in British Columbia, Canada.
The deal is not a venture round, not a corporate bond, not a hyperscaler's internal budget line. It is a secured loan against hardware — and that distinction matters more than the headline number.
What the Deal Actually Is
Asset-backed financing treats physical infrastructure — in this case, NVIDIA GB200 NVL72 rack-scale systems — as collateral. The lender underwrites the residual value and revenue-generating capacity of the hardware itself, not just the creditworthiness of the borrower's balance sheet. For GPU clusters that command significant market value and generate measurable compute revenue, that structure is coherent.
USD.AI positioned itself as the arranger of this facility, meaning it structured and placed the debt rather than holding it entirely on its own books. The $128.9 million figure makes it the platform's record transaction — a reference point that tells you something about the scale at which this financing model is now operating.
Why GB200 NVL72 Systems, and Why British Columbia
The NVIDIA GB200 NVL72 is a rack-scale unit — each system integrates 72 Blackwell GPUs in a dense, liquid-cooled configuration designed for large-scale AI training and inference workloads. At 32 systems, this deployment represents substantial compute density in a single facility.
British Columbia is not a random choice. Canada's regulatory environment, relatively stable power costs, and access to hydroelectric generation have made it an attractive destination for data-center operators who need both grid reliability and a defensible cost structure. The province has drawn attention from infrastructure investors looking to build outside the most saturated US markets.
The geographic specificity here is worth holding onto: this is not a paper commitment or a letter of intent. The systems are being deployed — capital allocated, hardware moving, facility contracted — in a defined jurisdiction.
Asset-Backed Lending as AI Infrastructure Rails
The broader signal in this transaction is structural. For years, AI infrastructure investment flowed primarily through equity — venture rounds, hyperscaler capex, and sovereign wealth commitments. Debt financing for GPU clusters existed but remained niche, limited by lender unfamiliarity with hardware depreciation curves and compute market dynamics.
That is changing. A $128.9 million asset-backed facility for a discrete GPU deployment, arranged through a platform purpose-built for this asset class, suggests that underwriting standards for AI hardware are maturing. Lenders are getting comfortable with GB200-class systems as collateral — assessing utilization rates, secondary market liquidity, and contract structures well enough to extend nine-figure loans against them.
For founders and operators building AI compute businesses, this matters tactically. Asset-backed debt is cheaper than equity and non-dilutive. If the lending infrastructure continues to develop — more platforms, more lenders, tighter underwriting models — it could meaningfully change how mid-scale AI data-center operators capitalize their buildouts. The equity-heavy model that has dominated AI infrastructure finance is not the only path available anymore.
The Bigger Shift
A single $128.9 million transaction does not remake a market. But it is a useful marker. USD.AI's record deal on September 23, 2026, in British Columbia, against 32 NVIDIA GB200 NVL72 systems, represents something more durable than a one-off financing: it represents the normalization of asset-backed debt as a tool for scaling AI compute infrastructure. The builders who understand that shift — and move early to access it — will have a structural cost advantage over those still treating equity as the default. That is the real stake here.