Compute IQ
Agentic underwriting for GPU financing
Compute IQ reviews the customer buying compute, the offtake contract and the operator behind the cluster.
Compute IQ is the underwriting system for the compute economy. Built on more than $25 billion of GPU financing data points as of October 8, 2026, it scores every deal against published underwriting standards. Every deal reviewed, funded or not, improves the model.
Every deal, scored the same way
Compute IQ places each GPU loan on one 100-point scale. The score determines financeability and the pricing tier. Fixed and floating rates are available.**
Illustrative: 46 deals in the order they were scored, each on one 100-point scale. Scores below 60 are not financeable; 60 to 75 prices in the 12–15% tier, 76 to 85 in the 10–13% tier and above 85 in the 9–11% tier. 28 of the deals funded; the rest were screened.
| Band | Scores | Funded | Screened |
|---|---|---|---|
| 9–11% | Above 85 | 14 | 2 |
| 10–13% | 76–85 | 8 | 2 |
| 12–15% | 60–75 | 6 | 1 |
| Not financeable | Below 60 | 0 | 13 |
Compute IQ scores the customer, the contract, and the operator behind the cluster, not just the hardware.
Customer and contract carry most of the review
The customer's credit and the terms of the offtake contract carry most of the review, because contracted revenue repays the loan.
A first deployment can still qualify
A newer operator with a strong customer and a well-drafted contract can be financed. Track record counts, but it isn't the whole score.
The same rubric for every operator
Every deal is scored on the same 100-point scale, so the outcome rests on the submitted evidence.
The Compute IQ score sets the price and structure, without case-by-case negotiation.
How the score sets loan terms
The score sets the loan size, pricing tier and structure together, establishing terms before a term sheet is issued.
A stronger score, a lower pricing tier
Better customer credit and stronger contract terms place a deal in a lower pricing tier. The tier remains the same over the loan term; the rate may be fixed or floating.
One structure for every loan
Every qualifying loan uses the same package: a ring-fenced SPV, escrow until install, a debt-service reserve and telemetry monitoring.
One score for every GPU loan
Compute IQ scores every financing originated through USD.AI on the same three inputs: the operator, the customer and the contract. That score sets the loan size, price and structure, and every lender reviews the same assessment.
The operator
Experience, financial position and ability to deliver the deployment.
The customer
The credit quality of the business buying the compute.
The contract
The terms behind the revenue, the payments and the financing period.
- Loan sizing
- Pricing
- Structure
See how a deal earns its score
Follow a sample GPU loan as Compute IQ scores the customer, the contract and the operator.
Sample deal
256 NVIDIA B200 GPUs for an AI lab
A neocloud finances a new cluster that an AI lab has contracted to use for 36 months.
Score this deal one area at a time, starting with who pays for the compute, and see where it lands.
Who pays for the compute
+34
- FinancialsAudited, revenue growing+9
- BackersEstablished venture investors+9
- Cash runway30 months+10
- LeverageLittle existing debt+6
The offtake contract behind the loan
+36
- Term36 months, the loan's full term+14
- PaymentsFixed monthly, take-or-pay+12
- CoverageRevenue covers debt service with headroom+10
The business running the cluster
+18
- Track recordThree years running GPU clusters+8
- DeploymentsTwo clusters in operation+6
- Equity25% of the hardware cost+4
Compute IQ score
0 / 100
Not scored yet
Same rubric, every deal
The score becomes the GPU loan
Financing is originated through USD.AI, and capital releases only once an independent third party verifies that the servers are delivered and installed on site.
Common questions about GPU loan underwriting
Compute IQ scores three areas: the operator, the customer's credit and the customer contract. The customer contract is the agreement under which the customer pays for GPU capacity. Because every deal goes through the same review, mid-market GPU deals are straightforward to evaluate and easy to compare.
GPU loan underwriting starts with an approved hardware purchase order, proof of the equity contribution and an executed colocation (data center space) agreement. Applicants also submit an executed customer contract or evidence of on-demand revenue. Diligence covers a corporate org chart showing owners of 10% or more, trailing 12 months of financial statements and KYC/AML (know-your-customer and anti-money-laundering) checks. See the underwriting and risk standards.
Compute IQ is USD.AI's agentic underwriting tool: software that applies patterns seen across hundreds of deals to each new one, against published underwriting standards. Every operator is scored on the same rubric, and final credit decisions follow documented review and approval. When a deal falls short, Compute IQ identifies the gaps that need to be addressed.
The Compute IQ score maps to financeability and one of three pricing tiers. The tier remains the same over the loan term; rates may be fixed or floating. Every qualifying financing through USD.AI carries the same structure: an SPV (a separate legal entity holding the GPUs and contracts), controlled accounts and a debt-service reserve (cash held for loan payments). It also carries residual-value insurance on the hardware collateral and telemetry monitoring.
On-demand or spot revenue does not disqualify a deal with USD.AI. On-demand revenue comes from customers renting GPU capacity as needed, and spot revenue comes from compute sold at the going market price. Both affect loan pricing because that revenue is not contracted.
As of October 8, 2026, Compute IQ is built on more than $25 billion of GPU financing data points. Every deal reviewed, funded or not, improves the model, so each score reflects how the market actually trades. Each deployment is assessed against the evidence accumulated across that financing activity.
