Loans against Nvidia GPU servers, for neoclouds.

Rackline lends exclusively against Nvidia GPU server fleets deployed by neoclouds — nothing else. One asset class, underwritten by people who track the secondary market for it every day.

Secondary index — H100 · 61.4%
% = today's resale price of the hardware, as a share of what it cost new.
Collateral cushion, year by year*
Hover a year to see the range.
Stretch case Base case Downside case Facility balance
Values are today's resale price of the hardware, as a percentage of what it cost new. *Illustrative example, not a real valuation or current market data. We size facilities to sit inside the downside case, not the base or stretch case — a single fitted curve hides how much hardware prices actually disperse.
Where we play

The compute market has four tiers. We work two of them.

Hyperscale and large-scale deals are corporate credit with GPUs attached. The middle market and micro-cluster tiers are where understanding the compute stack is the real edge — and that's where Rackline lends.

Largest deals

Hyperscale

Syndicated corporate debt. GPUs are incidental to the credit.

Large deals

Large scale

Underwritten to an investment-grade offtake or parent guarantee.

RACKLINE
Small & mid-sized deals

Middle market

Sophisticated neoclouds, thin lender coverage, requires real compute underwriting.

RACKLINE
Smallest deals

Micro clusters

Equipment-lease economics; hardware is the entire collateral package.

Rackline lends purely against the hardware, so we're active across the middle market and micro-cluster tiers — wherever the collateral and structure make sense.

Deal structure

Protections, not promises.

Every facility carries the same structural package, whether it's a first cluster or a fifth.

Bankruptcy-remote SPV

An isolated borrower

Each facility sits inside its own special-purpose vehicle with an independent manager, so the collateral is never entangled with the parent's broader balance sheet.

First-priority lien

Perfected, not just promised

A UCC-1 filing over the GPU servers plus a pledge of the SPV's equity — the security package a hyperscaler-backed deal gets, sized for a smaller cluster.

Step-in rights

Access to the metal

A lien waiver with the host colocation facility gives us the right to enter, operate, or remove hardware on default — not just a claim on paper.

Funded reserve account

A cushion before day one

One to three months of debt service sits in reserve at close, topped up from rent collections before anything flows to the sponsor.

Why compute expertise matters

Everyone can read a balance sheet. Few can read a GPU market.

Four things we underwrite around, that generalist lenders tend to miss:

Resale ≠ rental

GPU rental rates and secondary hardware prices move on different cycles entirely. Rental pricing reacts to whatever capacity just came online; resale pricing reflects the durable demand for owning the hardware outright. Underwriting off rental rates measures the wrong market.

Bands, not curves

A single depreciation curve hides dispersion. The same server configuration clears at very different prices in the same month depending on OEM, documentation, and buyer pool — so we underwrite to a conservative band, not a point estimate.

The financing market is being rebuilt in real time

The broader financing market is being reshaped by vendor-backed revenue guarantees, which are increasingly substituting for hyperscaler offtakes as the credit anchor behind neocloud lending. That's expanding the pool of neoclouds who can raise financing at all — and because Rackline's own facilities stand on the hardware itself, not on whether a borrower has one of these guarantees in place, we widen that pool even further.

The middle market is mispriced

Smaller, sophisticated neocloud borrowers can get the same structural protections as hyperscaler-backed facilities — bankruptcy-remote SPVs, perfected liens, funded reserves — but at a meaningfully wider spread, because most capital providers don't understand the collateral well enough to compete for them.

Eligible collateral

Nvidia server fleets only.

We don't underwrite AMD, custom silicon, or non-Nvidia accelerators. One vendor, one secondary market, tracked continuously.

Chip / platform Vintage Max advance Term
GB300 NVL72 2025–26 60%
36 mo
GB200 NVL72 2024–25 65%
36 mo
B300 2025–26 58%
30 mo
B200 2024–25 62%
30 mo
H200 2024–25 60%
30 mo
H100 2022–23 55%
24–36 mo
A100 2020–21 35%
18–24 mo
Click a row to select it in the chart above, or use the chart's tabs. *Illustrative example figures only — not current market terms, real pricing, or an offer of credit.
Get in touch

We're ready to partner with you — reach out directly.

Email works best right now, whether you're financing a fleet or looking to co-invest.

Building a neocloud?

Tell us about the Nvidia hardware you're financing or already run, and we'll walk through what a facility could look like.

Email deals@racklinecapital.com

Looking to co-invest or syndicate?

We're building out our credit facilities and investor relationships. Get in touch to hear about upcoming deals.

Email invest@racklinecapital.com