InfraComputeThe capital behind compute.
Menu

Tools  /  Term sheets  /  GPU delayed draw term loan

Annotated term sheet · 22 lines · Q3 2026 edition · data to 30 September 2026 · refreshed with the quarterly index

GPU delayed draw term loan

A model term sheet for a senior secured delayed draw facility to a GPU-owning project company, with each line annotated with what the filed facilities show and where lenders and borrowers push.

For: Lenders, GPU clouds and their counsel, sizing a GPU loan against a customer contract.

Randeep Bubbra | Lawyer | Making the next generation of infrastructure financeable · Firm profile ·

How to read it

Each line has model wording, what the filed deals show and where lenders and borrowers push. The first 3 lines are free here; subscribers see every line.

  1. Facility and size
  2. Margin
  3. Final maturity
  4. Advance rate
  5. Sizing DSCR
  6. What goes into the DSCR
  7. Minimum DSCR covenant
  8. Cash trap
  9. Cash trap escalation
  10. Contract value ratio
  11. Equity cure
  12. Customer contract
  13. Customer direct agreement
  14. Customer step-in
  15. Assignment and change of control
  16. Data centre access
  17. Security
  18. Serial numbers
  19. Spares
  20. Useful life
  21. Remarketing
  22. Casualty

Free PDF

Get the annotated term sheet

All 22 lines with the lender and borrower positions, as a PDF for your deal file.

Free. Work email only. The download starts here once the form is accepted. How we handle your details.

01

Facility and size

Senior secured delayed draw term loan of up to US$[•] to [Project SPV], drawn as GPUs are delivered and accepted under the Customer Contract.

What the filings showUS$34m to US$10bn across 30 checked US dollar facilities (median US$1.3bn).

LenderCommitments sized to the contracted fleet only, with any uplift for later phases subject to fresh credit approval.

BorrowerAn accordion for later phases on the same terms, so a new customer contract does not need a new facility.

DigitalOcean: GPU/CPU equipment finance facilityZankore: Senior term loan for GPU infrastructureNscale: North Carolina GPU DDTLNscale: Ward County GPU DDTLIREN: Mackenzie GPU financing (reported)Lambda: Investment-grade rated TLB

02

Margin

Term SOFR (zero floor) plus [•]% per annum, with commitment and upfront fees as agreed.

What the filings show2.25% to 7.5% over SOFR across 11 checked facilities in the register (median 4.25%). Fees, floors and OID are not included.

LenderA margin that reflects the customer's credit and the GPU residual risk, with a step-up if the facility is not refinanced by an agreed date.

BorrowerPricing off the customer's credit, with a step-down once the fleet is accepted and cash flow is proven.

Lambda: Investment-grade rated TLBCoreWeave: DDTL 5.5Nscale: Macquarie Iceland GPU facilityNebius: First secured GPU facilityIREN: Microsoft DDTL floating trancheCoreWeave: DDTL 5.0

03

Final maturity

The earlier of [•] years from the first drawing and [the end of the Customer Contract term].

What the filings show3.8 to 6.5 years from signing to stated maturity across 10 checked facilities (median 5.1 years).

LenderMaturity inside the contracted term, so the loan amortises from contracted cash flow and not from remarketing.

BorrowerMaturity matching the contract, with full amortisation over the term and no balloon.

DigitalOcean: GPU/CPU equipment finance facilityNscale: North Carolina GPU DDTLNscale: Ward County GPU DDTLLambda: Investment-grade rated TLBCoreWeave: DDTL 5.5Nscale: Macquarie Iceland GPU facility

Method

Ranges come from the checked records in the InfraCompute register and the filed documents behind the what's market cards, to 30 September 2026. The wording is a model for discussion, not a form for any deal, and the edition changes only with the quarterly index. General information drawn from public filings. It is not legal advice on any transaction, and using this page does not create a lawyer-client relationship. Views are personal.

Randeep Bubbra

InfraCompute Weekly

The week in compute finance, read from the filings.

Deals, terms and the clauses behind them, for the people financing data centres, GPU clouds and the power behind them. By Randeep Bubbra.

Want to run your own deal through these tools? Book 20 minutes with Randeep.

Fridays, one email. Work email only. Unsubscribe any time. Privacy.