If Neocloud Debt Stays Near 1,000 bp, Where Does Cash Stick?

ANALYSIS · MEMBERS CONTINUE · 5 min free / 14 min full · Data as at 16 Oct 2026

This week the market priced CoreWeave's unsecured debt like distressed paper: five-year default swaps near 1,038 basis points and a 2031 bond yielding 15.78%. Two weeks earlier, Lambda borrowed US$1 billion at 6.78%, and in March CoreWeave itself borrowed at about 5.9%. Same industry, same chips, a gap of roughly nine percentage points. If neocloud debt stays near 1,000 basis points, where does the cash stick?

In short

• The question: if neocloud debt stays priced near distress, where does cash stick?
• The answer: Scenario B. The cost of money now splits by customer. Capacity sold to investment-grade buyers still borrows near 6–7% from insurers; everything else pays about 15–16% or sells shares. Cash sticks first with the chip sellers paid at delivery, then with lenders secured on rated contracts, then with the asset-light sellers of compute hours. Neocloud shareholders, unsecured bondholders and AI labs without an investment-grade rating wait
• Why: CoreWeave's five-year default swaps reached about 1,038 basis points on 15 October 2026, which Reuters said implies close to a 63% chance of default within five years. Its 9.75% bond due 2031 yielded 15.78%. Lambda's US$1.008 billion loan, backed by two investment-grade customers, priced at a 6.78% fixed rate on 1 October
• The size of the book: CoreWeave carries about US$52 billion of debt and leases against a US$104.2 billion backlog. Gartner says AI data centre owners need US$8.5 trillion of revenue through 2030 and will see no more than US$3.6 trillion
• Behind the wall: who gets paid, in order · the price of money in plain numbers · three scenarios, arithmetic shown · what a rating is worth on a US$1 billion cluster · six practical takeaways · Members' FAQ
• What would break it: the secured market closing to all but the biggest buyers (A), or third-quarter results and lab ratings pulling unsecured yields back toward 10% (C)
• Nothing here is a recommendation. This maps what this week's prices and filings already assume, with the working shown

What's in this piece: the numbers that matter · why the price of money became the clock · what the map assumes in one line · members from here · who gets paid, in order · three scenarios, arithmetic shown · our read · what would change the view, with dates · practical takeaways · FAQ · references

The numbers that matter

Chart 1 — What neoclouds pay to borrow, by who stands behind the loan. Company releases and CNBC; gap is Second Order arithmetic. Education only.

Why did the price of money become the clock?

A neocloud buys GPUs, fills a hall and sells the computing hours to AI labs and big tech firms. The Neoclouds layer shows where that sits in the chain, and our GPU financing explainer covers how the chips get paid for. The short version: the chips are bought years before the hours pay them back, so the business runs on borrowed money.

That money comes in two kinds. The first is a loan secured on one cluster and one customer contract, held in a separate company so lenders can take the chips and the contract if things go wrong. When the customer is investment grade, that loan can be rated investment grade too. Below, "rated" means rated investment grade. CoreWeave's March facility, tied to a Meta contract, was the first; Lambda's October loan, tied to two rated customers and sold to insurers, followed. The second kind is unsecured company bonds and shares, which get paid from whatever cash is left.

This week the second kind broke away from the first. A Gartner note on 14 October forecast that more than 80% of neoclouds will stop operating independently by 2030. CoreWeave fell 15.5% that day and its default swaps went through 1,000 basis points. Our Wire on the swaps covered the move. The same week, Modal raised US$750 million of equity to sell hours it does not own the chips for, and Nscale's flagship Norwegian site was reported six months late ahead of its listing. So the money question is not whether demand for AI computing is real. It is who can still borrow cheaply to serve it, and that now depends on the customer, not the chip.

What does the map already assume?

Three answers compete:

1. The money binds harder. Unsecured yields stay near 16%, listings stall, and even secured lenders narrow to the very biggest buyers. Builds slip and landlords wait
2. The money splits by customer. Rated contracts keep clearing near 6–7% with insurers. Everything else pays double-digit rates or sells shares. Growth follows the rated book
3. The bind eases. Third-quarter results show cash arriving, AI labs win ratings, and unsecured yields fall back toward 10%

A tip sheet would call a bottom in one stock. Analysis asks which answer this week's prices already need.

Our read, in one line: the map already assumes Scenario B. The price of neocloud money now splits by customer: capacity sold to investment-grade buyers borrows near 6–7%, everything else pays about 15–16% or sells shares. Cash sticks first with chip sellers paid at delivery, then with lenders secured on rated contracts, then with asset-light sellers of hours. Neocloud shareholders, unsecured bondholders and AI labs without an investment-grade rating wait. The stuck step is the customer's credit, not the chip. The scenarios and the arithmetic are below.

The free preview ends here.

—— Still to read · ~9 of 14 minutes ——

MEMBERS CONTINUE HERE

• Who gets paid, in order, and who waits
• The price of money in plain numbers
• Three scenarios with the arithmetic
• What a rating is worth on a US$1 billion cluster
• Six practical takeaways
• Members' FAQ