Add the names in [brackets] before you send it; the company name is set in the builder.
Lender presentation · October 2026
[Your company]
A 128-GPU B300 cluster in [Metro]: 16 B300 servers, $9.72M all in.
Seeking a $4.86M loan (50% of cost) over 48 months.
Prepared with the GPU Lenders neocloud builder. Illustrative figures at October 2026 ballpark prices, before taxes; not an offer of financing.
The opportunity
What B300 GPUs rent for, and what this cluster earns
Public on-demand list prices, per GPU-hour
| Lowest DeepInfra | $4.89 |
|---|---|
| Median of 23 providers | $8.38 |
| Hyperscalers AWS | $17.80 |
| This plan, on demand | $7.50 |
| This plan, 48-month contract | $4.60 |
Each provider's lowest on-demand list price for the B300, from 23 providers' pricing pages, checked October 17, 2026 (American Compute GPU cloud price comparison). Hyperscaler prices include the instance's CPUs and memory. Short-term rates may not last.
GPU-hours a year
1,121,280
128 GPUs, every hour of the year
Revenue, year 2 of the loan
$5.16M
Year 1: $3.87M, with 3 months before first revenue
EBITDA margin, first full month
63%
Customers and offtake
100% of the GPUs under contract for 48 months
The contract
| Status | LOI or term sheet |
|---|---|
| Customer | [Name] · Early-stage startup, recently funded |
| GPUs | 128 of 128 |
| Price | $4.60 per GPU-hour |
| Firm term | 48 months from first revenue |
| Contract value | $20.6M |
On demand
| GPUs | 0, plus the contract's 128 after month 51 |
|---|---|
| Price today | $7.50 per GPU-hour |
| Hours billed | 80% |
| Price fall | 10% a year |
Concentration: 100% of first-year revenue from the contract customer.
The cluster and the site
16 B300 servers, 248 kW in [Metro]
| Item | Qty | Unit | Total |
|---|---|---|---|
| B300 servers | 16 | $550K | $8.80M |
| Cluster networking | 16 | $55K | $880K |
| Shipping | 16 | $2.5K | $40K |
| Hardware, all in | $9.72M |
Ballpark prices, October 2026; the vendor quote replaces them.
| Power | 15.5 kW a server, 248.0 kW |
|---|---|
| Colocation | $280 per kW a month |
| A month | $69K |
| Site status | Colo quote |
[Facility, power confirmed]
Unit economics
$271K of EBITDA a month, 63% margin
Revenue
$430K
Colocation
-$69K
Running costs
-$89K
Loan payment
-$128K
Left over
$143K
Revenue per billed GPU-hour
$4.60
Cost per billed GPU-hour
$3.06
Colocation, staff and the loan payment
Contract revenue alone covers
2.12x
of the loan payment, after running costs
Running costs a month: staff $75K (4 full-time at $225K a year, loaded); insurance $2.9K (0.36% of the hardware a year); software and bandwidth $6K (Orchestration, monitoring, billing, IP transit, remote hands); sg&a $5K (Accounting, audit, legal, compliance, admin). First full month of revenue (month 4). Before depreciation and tax.
The financing ask
A $4.86M loan against $9.72M of GPUs
| Loan | $4.86M (50% of cost) |
|---|---|
| Rate | 12% |
| Term | 48 months, fully amortizing |
| Payment | $128K a month |
| Fee | 2% at closing |
| Reserve | 3 months of payments |
| Security | The GPUs and the contract (first lien) |
Uses
| Hardware, all in | $9.72M |
|---|---|
| Closing fee (2% of the loan) | $97K |
| Debt service reserve (3 months) | $384K |
| Colocation and running costs before first revenue (3 months) | $475K |
| Loan payments before first revenue (3 months) | $384K |
Sources
| Loan (50% of the hardware) | $4.86M |
|---|---|
| Your cash | $6.20M |
| Total | $11.1M |
Coverage and the downside case
Coverage against the 1.50x lender minimum
| Case | DSCR | Lowest |
|---|---|---|
| PlanThe inputs as set | 2.12x | 2.12x |
| DownsideOn-demand price 30% lower and 20 points fewer hours billed; the contract unchanged | 2.12x | 2.12x |
| Contract onlyNo on-demand revenue at all: how lenders size the loan | 2.12x | 2.12x |
Returns and resale
13% equity IRR, with resale at the conservative band
Cash in before first revenue
$6.20M
Payback from operations
Month 47
After debt service, before resale
Money back over money in
1.34x
Equity IRR
13%
The servers when the loan ends (2030)
| Conservative band, 12-21% of price | $1.06M-$1.85M |
|---|---|
| Moderate band, 21-43% | $1.85M-$3.78M |
Residual bands from American Compute’s GPU Residual Value report (June 2026). Networking and shipping are given no resale value.
Team
Who runs it
| Founder or CEO | [Name], [prior roles and deployments] | In place |
|---|---|---|
| Infrastructure engineering | [Hiring: profile and timing] | To hire |
| Platform engineering | [Hiring: profile and timing] | To hire |
| Sales | [Hiring: profile and timing] | To hire |
| Finance and vendors | [Hiring: profile and timing] | To hire |
Operating history: first cluster. [Past deployments the team ran elsewhere: nodes, sites, dates, uptime.]
Lenders look up the website and every leader’s LinkedIn profile first.
Next steps
What is in hand, and what comes next
In hand (3 of 15)
- A named customer for the capacity
- What the loan buys and how much
- The terms you want
Still to bring (12 of 15)
- Signed contract: share, price, firm term
- How the customer’s payments are secured
- Proof the customer can pay
- What you are buying, from a quote
- Colocation agreement with confirmed power
- Dated schedule to first revenue
- Who owns engineering, operations and finance
- Clusters you have run before
- +4 more
Timeline
| Closing | Month 0 |
|---|---|
| First revenue | Month 4 |
| Contract ends | Month 51 |
| Loan repaid | Month 48 |
The ask: a term sheet for a $4.86M loan over 48 months. Contact: [name, email, phone].