Free report
What aircraft and infrastructure finance teach GPU lenders
GPU lending has rebuilt in three years what aircraft finance, project finance and securitisation took decades to learn. Ten lessons from the older markets, and how far the 2026 deals have already taken them.
An old playbook for a new asset.
A GPU cluster is as mobile as an aircraft, financed like a power plant and housed in a building that is itself being securitised. Each older market has already solved part of the problem GPU lenders face now.
Key findings
What the deals show.
Drawn from credit agreements, rating releases and company filings. The full report sets out the sources for every figure.
- 01
GPU debt runs for most of the hardware's life
Lenders' models assume six years; 2026 GPU loans run up to about six. Aircraft, data centre and renewables debt covers a fraction of the asset's life.
- 02
There is no Cape Town for compute
GPUs sit in third-party buildings with no asset registry, so enforcement depends on direct agreements and landlord waivers.
- 03
Valuation is still a formula
GPU loans test cost less straight-line depreciation, not independent appraisals as in aircraft finance.
- 04
ABS warns earlier
Data centre ABS trap cash at 1.30x to 1.35x. GPU project loans trap at 1.10x and default at 1.05x.
- 05
The second customer is coming
Remarketing rights and re-leasing tests have arrived in the 2026 deals, borrowing from aircraft finance.
GPU debt runs for most of the hardware's life
Show the data table
| Asset | Useful life (years) | Typical debt tenor (years) | Note |
|---|---|---|---|
| GPU servers | 4 to 6 | 2.5 to 6 | Lenders' models assume six years straight-line (IREN, CoreWeave DDTL 2.0); 2026 loans run from 30 months to about six years |
| Data centre buildings | 20 to 40 | 5 to 23 | ABS anticipated repayment dates of about five years to amortising lease-backed bonds to 2049 |
| Commercial aircraft | 25 to 30 | 7 to 12 | Typical secured loan, EETC and ABS amortisation profiles |
| Wind and solar | 25 to 35 | 15 to 20 | Typical contracted project finance tenors |
| Telecom towers | 30 to 40 | 5 to 10 | Typical tower ABS anticipated repayment dates |
Inside the report
Read the terms.
Then see the structure.
12-page PDF, written for lenders, investors, sponsors and their advisers.
- 01
Ten lessons
Valuation, repossession, maintenance, remarketing, servicers, construction, liquidity, diversification, refinancing and third-party support.
- 02
Two structure diagrams
A typical aircraft ABS and a contracted infrastructure project finance, set against the GPU structures.
- 03
The tenor problem
Useful life against debt tenor across GPUs, aircraft, data centres, renewables and towers.
- 04
Triggers compared
Cash trap and amortisation levels in data centre ABS against GPU project loans, from the filed documents.
- 05
Protections side by side
How aircraft, project finance, data centre ABS and GPU lenders protect themselves, in one table.
- 06
Questions for the market
Five open questions on registries, appraisals, servicers, lessors and infrastructure investors.
Questions
The short answers.
How long do GPUs last compared with the loans against them?
Lenders' models in the public GPU loans assume a six-year straight-line life, and 2026 GPU loans run from 30 months to about six years. In aircraft, data centre and renewables finance, debt covers a much smaller share of the asset's life.
Is there a Cape Town Convention for GPUs?
No. Aircraft lenders rely on an international registry and fixed repossession timelines. GPU lenders rely on contracts instead, such as direct agreements with colocation providers and landlord waivers.
Why do data centre ABS trap cash earlier than GPU loans?
Data centre ABS trap cash at 1.30x to 1.35x DSCR because the leases may be renewed. GPU project loans are sized to repay in full within the contract, so they trap at 1.10x and default at 1.05x.
What is in the full report?
Ten lessons from aircraft finance, project finance and securitisation, two structure diagrams, two charts, a side-by-side table of lender protections and questions for the market.