GPU financing
Contract-backed financing
Debt sized on a signed customer contract for compute, not only on the hardware. It is the structure most GPU lenders start from, and the one that turns a customer agreement into capital for the next cluster.
What it is
A GPU operator signs a contract with a customer who commits to pay for compute for a fixed term. A lender then provides debt against that contract, often with the hardware as collateral as well. The loan is repaid out of the contract revenue, in instalments within the contract term.
The contract carries the loan. The hardware is the second line of defence. That is why the strength of the customer and the wording of the contract decide more than the GPU generation does.
What makes a contract financeable
Lenders read the customer contract before anything else. These are the terms they look for.
01
Customer credit quality
An investment grade or large, profitable customer weighs far more than a young company. Lenders run their own assessment of the customer.
02
Take-or-pay or reserved capacity
The customer pays whether it uses the capacity or not. Usage-based pricing with no minimum is weak security.
03
Term at least as long as the loan
Typically 2 to 5 years. The loan is repaid within the contract term, so a short contract means a short loan.
04
Limited termination rights
A contract the customer can end at will is hard to lend against. Termination with penalty or long notice is workable.
05
Fixed or floored price
A price per GPU-hour that is fixed, or has a floor, makes the revenue predictable. Pure market pricing does not.
06
Assignable to the lender
The contract, or the receivables under it, can be pledged or assigned to the lender as security.
How much the contract can carry
Indicative, from public sources. Every figure depends on the customer, the structure and the market, and no lender is bound by them.
- Starting point
- Contract revenue after operating costs, over the contract term
- Cover
- Lenders commonly want that cash flow to exceed interest and instalments by 1.3 to 1.5 times
- Hardware cap
- Even with a strong contract, the loan is usually capped at a share of the hardware's resale value, roughly 50 to 70 percent
- Equity
- The sponsor is expected to fund a meaningful share of the project, often 20 to 30 percent or more
- Concentration
- A single customer above about a quarter of revenue is questioned, and a single-customer deal is priced accordingly
From contract to drawdown
1
Contract and quote
Signed customer contract, hardware quote and colocation agreement. Missing one is normal at the inquiry stage.
2
Inquiry and first read
We read the material against what lenders ask for and say plainly what is financeable today.
3
Introduction
To lenders whose appetite fits the customer, the term and the size, and only after you have agreed.
4
Term sheet and due diligence
The lender assesses the customer, the contract and the owners. Often a month or more.
5
SPV, security and documents
Hardware and contracts usually sit in a ring-fenced company that borrows. The contract is assigned as security.
6
Delivery and drawdown
Funds are typically paid against delivered hardware, then repaid out of the contract revenue.
No contract yet?
Then the financing conversation is early. The faster route is to find the customer first. Put your capacity on the GPU capacity market, where compute buyers describe what they need, and we introduce the parties that fit. Once a contract is signed, the financing inquiry is a short step.
Open the GPU capacity marketAre you the customer?
If you buy GPU compute on a fixed-term contract, your agreement is what makes an operator financeable. Tell us what you need, and we introduce operators who can deliver it and who are set up to finance against your contract.
I need GPU computeFinanceGPU introduces borrowers to lenders. We do not lend, hold funds or advise on investments, and every credit decision is made by the lender.