CoreWeave · unit economics and capital structure Debt snapshot: September 30, 2026 · Form 10-Q

CoreWeave: A pro forma

A multi-dimensional sensitivity analysis, inspired by the phase diagram of physics.

  • Server economics. One GPU, bought with a loan and rented by the hour.
  • Debt stack. What CoreWeave owes, at each interest rate.
  • Rental prices. What a GPU-hour actually costs to rent.
  • Other neoclouds. Two listed rivals on the same map.
  • Questions and insights. What we found, and what is still open.
  • Role-play the CFO. Make your own moves and try to reach green.
  • Method and sources. Numbers come from SEC filings and other public sources.
  • For agents. The model, the data and test cases, for your AI to take and rework.
State ·
Loan cover at the tightest year
Rent banked so far divided by loan payments due so far. Below 1.00 means rent alone does not cover the loan.
Owners get back per $1 invested
Loan rate
Fed rate plus the lender’s margin
Owners’ target return
Loan rate plus the owners’ premium
Fig. 1 · One GPU, one loan, rented by the hour

Phase diagram

Move the levers and watch which of three states the GPU lands in. Like water, it changes state when squeezed: price is the pressure, the cost of money is the heat.

Melt path: the same deal signed one, two, three years later

Swipe sideways to see the whole chart →

Solid
Profitable: rent repays the loan and owners beat their target return.
Liquid
Lenders paid, owners short: the loan is repaid, but owners earn less than their target.
Anxious
Needs new financing: rent banked falls behind the loan payments. New capital or a refinancing has to fill the gap.

Click or drag on the diagram to move the two levers on its axes. All other levers stay where the sliders are. The three small diagrams are the same GPU seen along each pair of the three big levers. The melt path shows on price against rentable life; CoreWeave’s debt shows on the two Fed-rate views.

Fig. 2 · Rental economics of one GPU-hour

Where each rental dollar goes

Which GPU? CoreWeave’s fleet average, not one make or model.
  • Rented like an Nvidia H100: an hour when new.
  • Costs to buy and install: a blend of H100s (about $37,000) and newer Blackwell GPUs ($60,000 to $77,000).

Rent per GPU-hour, averaged over the GPU’s rentable life. Percentages are shares of the first bar.

Swipe sideways to see the whole chart →

Every step as a table
Step Per GPU-hour Rent per GPU over its rentable life
Fig. 3 · Which lever matters most

Longer bar, stronger lever

Lever 10% better Lever 10% worse Repayment line: loan cover 1.00× Loan cover now

Each row moves one lever alone, by 10% of its value now, and holds the rest at the slider values. “Better” means a higher price, a fuller fleet or a longer life, and a lower cost or rate. Loan cover is rent banked divided by loan payments due, at the tightest year. Loan share, loan length and the owners’ required return are left out: they do not push the result in one direction.

Fig. 4

The race against the loan

Cumulative dollars per GPU. New financing is needed the first year rent banked ends below loan payments due.

Table 1

Breaking points

Each row moves one lever alone and holds the rest at the slider values.

Lever Now Owners miss their target when it is New financing is needed when it is