What makes a neocloud MSA financeable to institutional lenders

Institutional lenders do not underwrite GPUs as hardware — they underwrite the contracted cashflow those GPUs are supposed to produce. A neocloud MSA is financeable only when that cashflow is stable, assignable, and hard to kill: the customer can't easily walk, the economics are must-pay, the off-taker is creditworthy, the GPUs and contracts sit in a bankruptcy-remote SPV, and the future receivables are already assigned to the financier.

A financeable MSA is a neocloud's master services agreement written so that a bank or private-credit fund will lend against the revenue it produces, at a rate the operator can live with. The distinction that governs everything below: "The chips are the collateral on paper; the contracted revenue is the collateral that actually gets underwritten" (Peony, 2026).

This article is written for neocloud operators and founders raising debt against their sales contracts. It walks the five things a credit committee looks for, in the order they will look for them, with the contract language that turns a services agreement into bankable offtake.

Key terms, defined

• A neocloud is a specialist GPU cloud that rents AI compute at scale — bare-metal GPU servers, clusters, or compute-as-a-service — rather than general-purpose cloud.
• An MSA (master services agreement) sets the general terms between the neocloud and its customer; the commercial specifics (capacity, term, price) usually sit in order forms underneath it (American Compute, 2026).
• An off-taker is the customer that has contracted to buy the compute — the party whose payments repay the loan.
• Take-or-pay means the customer pays for reserved capacity even if it goes unused (Law Insider).
• A hell-or-high-water clause requires that "payments must continue irrespective of any difficulties" (Wikipedia).
• A bankruptcy-remote SPV (special purpose vehicle) is a separate entity that holds the assets and contracts so their cashflow is isolated from the operating company's insolvency (Appleby).
• An assignment of receivables is a present transfer of the right to the MSA's future payments to the financier, effective now rather than on a later trigger.

What do lenders actually underwrite in a neocloud deal?

Lenders underwrite the customer's ability to pay, not the operator's hardware. GPUs depreciate fast and resell into a thin, volatile market, so a financier sizing a facility looks first at the contracted revenue — how much, for how long, and how hard it is to stop. In practice the debt has to amortize inside the contracted-revenue window, so the loan matures in step with the contract and the fleet's useful life (Peony, 2026).

That is why "with offtake agreements in-hand, operators have the revenue visibility to raise GPU-backed debt" (American Compute, 2026). The MSA is the asset being financed. The five sections below are the five places a credit committee decides whether that asset is real.

Can the off-taker walk away? Termination for convenience and cause

The first hole in future cashflow is a customer that can leave. A financeable MSA closes it: termination for convenience should be gone or punitive, and termination for cause should be narrow, specific, and curable.

Termination for convenience — the right to walk for no reason on notice — is the enemy of financeability. Ideally it is deleted. Where a large customer insists on it, it should be pushed out and made expensive: a lock-up period plus a make-whole or remaining-term payment, so leaving early still pays the loan. Reported deals show the shape of this: one $1.3 billion USD deployment "prohibits either party from terminating for convenience during the first 36 months" (American Compute, 2026).

Termination for cause should be a scalpel, not a kill switch. A customer that can exit on any missed metric hands the lender an unstable cashflow. The financeable version narrows cause to defined, material, uncured breaches and routes ordinary performance misses to service credits instead — the customer gets compensated, but the contract (and the payment stream) survives. The financier wants it as hard as possible for the customer to stop paying.

Whose credit matters — the neocloud's or the customer's?

Who writes the check matters more than who owns the cluster. A neocloud's own balance sheet is usually thin; the off-taker's credit is what a lender is really buying, and it moves the cost of capital more than anything else in the deal.

The benchmark is CoreWeave's investment-grade facility, which priced at SOFR plus 2.25% and was rated A3 by Moody's and A (low) by DBRS, maturing in March 2032 — because the rating agencies underwrote the hyperscaler customer, not the GPU hardware, treating long-term take-or-pay agreements from counterparties like Meta and Microsoft as "near-sovereign credit support" (Peony, 2026). Weaker off-takers pay for the difference: before investment-grade pricing was available, private-credit structures on GPU deals carried risk premiums "reported in the 10-15% range" (Peony, 2026).

Directionally, that is the whole game. A neocloud selling into hyperscaler-grade credit can reach financing in the mid-single digits; a weaker off-taker can push the cost into the low-to-mid teens — or make the deal unfinanceable at any price. The practical consequence for an operator: a thin balance sheet with a strong customer is still financeable, and a stronger-looking operator with a weak customer often is not. If you can choose which contract to anchor a facility on, anchor it on your most creditworthy off-taker.

Who pays if the GPUs sit idle? Take-or-pay and hell-or-high-water

Someone has to pay no matter what — that is what turns a services contract into offtake. Two clauses do the work.

Take-or-pay obligates the customer to pay for reserved capacity even when it is not used (Law Insider). In compute, this shows up as reserved-instance commitments backed by real money up front: typically deposits run from roughly 20% to 60% of contract value, and in one arrangement a customer paid $10,293,350 USD before a 36-month, $34.3 million USD contract became effective - a 30% prepayment (American Compute, 2026).

Hell-or-high-water keeps the money flowing through trouble: the clause provides that the payments must continue irrespective of any difficulties. In a neocloud MSA, the financeable version keeps the customer paying through site issues, equipment problems, and most operator-side disruption, and pairs it with no-offset / no-withholding language so a service dispute cannot be used to switch off payment (American Compute, 2026).

The one place this genuinely gets negotiated is force majeure, and reported MSAs land all over the map — from obligations that are "delayed but not excused" to a customer right to exit after 30 consecutive days of disruption (American Compute, 2026). For financeability, the payment obligation should survive as much of that as the customer will accept; every excused-payment carve-out is a hole the credit committee will find. (For where these must-pay clauses meet uptime obligations, see our piece on what's market in data-center uptime SLAs.)

Where should the GPUs and MSAs sit? The bankruptcy-remote SPV

Do not leave the boxes or the contracts in the operating company. A financeable structure puts the GPUs and the MSAs in a bankruptcy-remote SPV, and the financier takes security over that SPV.

A bankruptcy-remote SPV is a legally separate entity built so that its assets and cashflow stay "beyond reach of its creditors" if the originator becomes insolvent, using structural features like a true sale of the assets into the SPV, limited-recourse obligations, and non-petition covenants (Appleby). In a neocloud deal, that means the SPV owns the GPUs and is party to the customer MSAs, and "the associated customer contracts and their receivables are assigned into the same entity, so the revenue that repays the loan flows to the SPV, not through the operating company" (Peony, 2026).

The result is a clean waterfall the lender can see end to end: customer → SPV → debt service → residual to sponsor. MSAs should therefore permit assignment to an affiliate SPV up front, so order forms and their receivables can be moved into the financing entity without re-consenting each customer later (American Compute, 2026). Leaving the hardware or the contracts in the opco collapses that waterfall — the moment the operating company has other creditors, the cashflow is no longer clean.

What happens to the cashflow if the neocloud folds?

This is the clause that decides whether a lender is protected on the day it actually matters. The financier needs a present assignment of the MSA receivables — not a promise to assign later — that survives the operator's collapse.

In a financeable MSA, the off-taker acknowledges the assignment, and on the operator's default or insolvency the customer pays the financier (or an account the financier controls) directly. The assignment must be effective now, so it is not clawed into the operator's bankruptcy estate. Alongside it, lenders want step-in rights and a workable handoff to a replacement operator. Reported deals already contain the mechanics: CoreWeave's $11.9 billion USD agreement with OpenAI includes an "alternate operator" provision enabling a forced handoff within "two business days" of a material breach.

Step-in and assignment do different jobs, and a financier wants both. Step-in lets the lender take over and run the cluster. Assignment lets the lender take the future cash even before — or without — operating the site. The first keeps the service alive; the second captures the receivables directly. The combination is what makes the cashflow genuinely "hard to kill."

The financeable-MSA package, at a glance

Financeability lever	The hole it closes	Financeable position
Termination for convenience/cause	Customer walks away from the cashflow	Convenience gone or punitive (make-whole); cause narrow, material, curable; misses routed to credits
Off-taker credit	Weak payer raises the cost of capital or kills the deal	Anchor the facility on the most creditworthy off-taker; strong credit compresses the spread
Take-or-pay + hell-or-high-water	Customer stops paying for idle or disrupted capacity	Pay-for-reserved-capacity, keep-paying-through-trouble, no-offset, tight force-majeure carve-outs
Bankruptcy-remote SPV	Opco creditors reach the cashflow	GPUs and MSAs in the SPV; security over the SPV; clean customer → SPV → debt → sponsor waterfall
Present assignment of receivables	Cashflow dies with the operator	Present assignment, off-taker acknowledgment, direct pay on default, plus step-in rights

Key takeaways

• Lenders underwrite the contract, not the GPUs. The chips are collateral on paper; the contracted revenue is what actually gets underwritten, so the MSA is the asset being financed.
• Close the walk-away holes. Delete or heavily penalize termination for convenience, and narrow cause to material, uncured breach — route ordinary misses to service credits.
• The off-taker's credit sets the price. Hyperscaler-grade credit has reached investment-grade pricing (CoreWeave: SOFR + 2.25%); weaker off-takers have seen 10-15% risk premiums or no deal at all.
• Make the economics must-pay. Take-or-pay plus hell-or-high-water, backed by real prepayment and no-offset language, turns a services contract into recognizable offtake.
• Isolate and pre-assign the cashflow. Put the GPUs and MSAs in a bankruptcy-remote SPV, and give the financier a present assignment of receivables — with step-in rights — that survives the operator's insolvency.

Make your sales contracts financeable before you go to the market

The package a credit committee wants is specific: a creditworthy off-taker, almost no walk rights, must-pay economics, a bankruptcy-remote SPV, and an already-assigned right to future MSA cashflow if the neocloud dies. Getting there is a drafting problem, and every idle GPU hour spent waiting on redlines has a cost.

LegalBooks is a tech-native law firm built for compute infrastructure. In 2026 we have negotiated more than $266 million USD in client contract value, roughly 99% of it data-center deals, with redlines back within six hours, guaranteed. We work both sides of these deals — operators, off-takers, and the capital behind them — as technical peers, licensed in the US (California and Nevada) and Canada (Ontario).

Book a call with counsel to pressure-test whether your MSAs are financeable — before a lender does it for you.

This article is general information, not legal advice. For advice on your specific situation, consult a qualified attorney. LegalBooks provides counsel for compute infrastructure deals across the US and Canada.