CoreWeave Interest Coverage: 2.4x on EBITDA, 25 Cents a Dollar to Lenders

CoreWeave booked $640 million of net interest expense in the second quarter of 2026, against $2,575 million of revenue. That is 24.9 cents of every revenue dollar sent to lenders before tax, before capex and before any return to shareholders. CoreWeave interest coverage looks comfortable on adjusted EBITDA, at 2.4 times. It looks very different once depreciation is counted, and the gap between those two readings is the whole story.

The short answer

CoreWeave's interest coverage on adjusted EBITDA was 2.36 times in Q2 2026, meaning cash operating earnings before depreciation covered net interest a little over twice. Measured on adjusted operating income, coverage was 0.20 times. The company is covering interest from cash flow, not from earnings, and depends on depreciation being a cost it can defer.

CoreWeave interest coverage is a flow question, not a stock question

Most coverage of CoreWeave debt counts the stock: how many billions are outstanding, at what spread, maturing when. That work is done in our breakdown of the CoreWeave debt stack and its spread ladder, and this post does not repeat it. A stock of debt tells you what is owed. It does not tell you whether this quarter's business can carry it.

Interest coverage answers the second question. It divides some measure of earnings by interest expense for the same period. A ratio of 3 means earnings could pay the interest bill three times over. A ratio below 1 means earnings did not pay it at all, and the gap came from cash on hand, new borrowing or non-cash add-backs.

The choice of numerator decides the answer. EBITDA, operating income and cash flow each describe a different version of the business. For a company spending far more on equipment than it earns in revenue, those versions are not close to each other. I think the honest approach is to show all of them side by side and let the reader see the spread.

Every figure below comes from two primary documents. The first is the Form 10-Q for the quarter ended 30 June 2026. The second is the earnings release of 11 August 2026, which carries the adjusted EBITDA reconciliation. Every ratio is my arithmetic on their numbers, and the working is shown so you can check it.

Where one CoreWeave revenue dollar went in Q2 2026

Start with the simplest frame. Take one dollar of Q2 revenue and follow it down the income statement. Revenue was $2,575 million, so each line below is that line divided by $2,575 million.

Source: CoreWeave Form 10-Q, Q2 2026, condensed consolidated statements of operations. Cents are Zan Digital's arithmetic and are rounded to one decimal.

The striking row is the interest line. At 24.9 cents it is larger than marketing, administration and tax combined, and it is almost exactly the size of the net loss. Remove interest and CoreWeave would have been close to break-even on a GAAP basis in the quarter.

The row that does most of the work, though, is technology and infrastructure at 58.5 cents. Most of that line is depreciation. The earnings release reconciliation shows $1,393 million of depreciation and amortization in Q2 2026, which is 54.1 cents per revenue dollar on its own.

The same dollar a year earlier

In Q2 2025 CoreWeave reported $1,212 million of revenue and $267 million of net interest expense. That is 22.0 cents per dollar. In Q1 2026 it was $536 million against $2,078 million, or 25.8 cents. The share has moved within a narrow band while the absolute bill has more than doubled.

Revenue and interest are growing at roughly the same pace, because the debt is raised to buy the equipment that produces the revenue. The interest share has not exploded. It has also not come down, which is what lenders would eventually want to see.

Five ways to measure CoreWeave interest coverage

Here are the five ratios a credit analyst would compute, using the same quarter for numerator and denominator.

Sources: Form 10-Q Q2 2026 (interest, capitalized interest, operating income) and earnings release of 11 August 2026 (adjusted EBITDA, D&A, adjusted operating income). Q2 2025 uses $753M adjusted EBITDA, $559M D&A, $200M adjusted operating income, $267M interest and $23M capitalized interest. Ratios are Zan Digital's arithmetic.

Why EBITDA coverage is the number lenders watch

Lenders to an asset-heavy company care about cash available for debt service. EBITDA adds back depreciation because depreciation is a non-cash charge for equipment already paid for. On that reading, CoreWeave produced $1,510 million of cash-like earnings and owed $640 million, which is a serviceable ratio for a heavily indebted borrower.

The ratio is weakening, though. It fell from 2.82 times to 2.36 times over the year, and from 2.82 to 2.10 times if capitalized interest is counted. Interest grew 140% year on year while adjusted EBITDA grew 101%. The company's adjusted EBITDA margin slipped from 62% to 59% across the same period.

Why the operating-income ratios matter more to me

Depreciation is non-cash in the quarter it is booked, but the equipment it describes was bought with borrowed money and will need replacing. A GPU that wears out over six years has to be repaid inside those six years, or refinanced. That is why I treat EBITDA minus depreciation as the closer measure of what the business earns on a sustainable basis.

On that measure CoreWeave covered 18% of its interest bill in Q2 2026. The depreciation assumption carries the entire gap, and that assumption is examined in our analysis of GPU depreciation schedules and AI earnings. Shorten the assumed life of the fleet and the 0.18 times reading goes below zero.

Capitalized interest: the part of the CoreWeave interest bill you do not see

US accounting rules require companies that build long-lived assets to capitalize the interest on money borrowed to fund construction. That interest goes onto the balance sheet as part of the asset's cost, rather than through the income statement. It reappears later as depreciation once the asset is in service.

CoreWeave's 10-Q states that it capitalizes interest on data center construction and related equipment purchases. The amounts are disclosed in Note 5. They were $79 million in Q2 2026, against $23 million in Q2 2025. For the first half they were $176 million, against $36 million a year earlier.

That means the reported $640 million understates what CoreWeave's lenders were actually paid for the quarter. Add back the capitalized amount and the interest cost was $719 million, or 27.9 cents per revenue dollar. The debt note shows the same thing from the other side: contractual interest of $592 million plus $45 million of amortized discounts and fees, less $79 million capitalized.

Why capitalized interest matters for the Q3 read

Capitalized interest falls as construction finishes and sites go live. Q1 2026 capitalized $97 million, and Q2 capitalized $79 million. If that trend continues, more of the true interest cost moves onto the income statement each quarter, even if the debt balance were flat.

So part of the jump to an $860 million to $940 million Q3 guide could be accounting rather than new borrowing. The filing does not split the guide, so I would not assume either explanation yet.

Cash interest paid vs interest expense at CoreWeave

The cash flow statement carries one more interest figure, and it is the most flattering. For the six months to 30 June 2026, CoreWeave reported cash paid for interest, net of capitalized amounts, of $806 million. Interest expense, net, for the same six months was $1,176 million.

On cash, adjusted EBITDA of $2,667 million for the half covered interest paid 3.31 times. A year earlier the same ratio was 3.76 times, on $1,360 million of adjusted EBITDA and $362 million of cash interest. Cash coverage is higher than expense coverage in both years, and it has also fallen.

Reconciling the $370 million gap

The gap between $1,176 million of expense and $806 million of cash can be mostly reconstructed from the filing. The debt note attributes $985 million of first-half interest expense to debt obligations. The remaining $191 million of the income-statement line is not itemised in the portions of the filing I could read.

The $15 million residual is unexplained. Treat this table as a reconstruction that gets within 2%, not a filed reconciliation.

Accrued interest is the line to watch. CoreWeave issued its 9.75% senior notes in April 2026 and its 9.625% and euro senior notes in June 2026. Bonds of this type usually pay semiannual coupons, so the first payments on that new paper fall after the half-year closed. The cash ratio of 3.31 times therefore partly reflects timing, and it should converge towards the expense ratio once those coupons start.

The interest run-rate on CoreWeave's 30 June debt

A quarter's interest expense reflects the average balance during the quarter, not the balance at the end. CoreWeave borrowed heavily inside Q2: the cash flow statement shows $13,457 million of debt proceeds against $3,884 million of repayments. So the Q2 bill understates what the 30 June balance costs.

The debt note lists every facility with its principal and an effective interest rate, rounded to a whole percent. Multiplying each principal by its effective rate and summing gives an annual interest cost on the 30 June balance of roughly $2.99 billion. Gross principal was $35,551 million, so the blended effective rate is about 8.4%.

That works out to roughly $748 million a quarter before any interest is capitalized, against the $640 million booked in Q2. Excluding the two convertible issues, which carry effective rates near 2%, the blended rate on the rest of the stack is about 9.9%. For context on how that cost of money compares with other AI infrastructure borrowers, see our tally of high-yield data center bonds.

What each new billion costs

At a 9% marginal rate, each additional $1 billion drawn adds about $90 million a year, or $22.5 million a quarter. CoreWeave had undrawn commitments on several facilities at 30 June, including most of the $8.5 billion DDTL 4.0 facility, of which $2.8 billion was drawn. A delayed draw term loan, or DDTL, is a facility committed up front and drawn as equipment arrives.

The Q3 guide of $860 million to $940 million is $112 million to $192 million above the 30 June run-rate. At 9% that implies roughly $5 billion to $8.5 billion of extra average borrowing during the quarter, or a smaller amount plus less capitalized interest. That range is my inference, not company guidance.

Rerunning CoreWeave interest coverage when Q3 results land

CoreWeave has not confirmed its Q3 2026 reporting date as of 24 October 2026. Last year it reported Q3 on 10 November. When the numbers arrive, three inputs decide whether coverage improved: revenue, adjusted EBITDA and net interest expense. The guide gives two of them.

Q3 guidance is revenue of $3.45 billion to $3.6 billion, adjusted operating income of $200 million to $260 million and interest expense of $860 million to $940 million. CoreWeave does not guide adjusted EBITDA. So the useful exercise is to compute the EBITDA margin needed to hold Q2's coverage of 2.36 times.

Illustrative. Inputs are company guidance as reported by INDmoney (12 August 2026). Outputs are scenarios computed by Zan Digital, not forecasts.

My read is that Q3 coverage on EBITDA will land within a tenth of Q2's figure, because the guide was built that way. The more informative number will be adjusted operating income coverage. If it climbs from 0.20 times to the 0.26 times the guide implies, depreciation is starting to lag revenue. If it stalls, it is not.

When the Q3 2026 filing lands, fill in four numbers in the five-row coverage table above: interest expense, net; capitalized interest from the property and equipment note; adjusted EBITDA; and depreciation and amortization. The whole exercise takes about 20 minutes. If net interest is above $900 million and the adjusted EBITDA margin is below 60.3%, coverage fell for a second consecutive quarter.

Where this argument is weakest

The analysis above leans on ratios a bond analyst would use for a mature company. CoreWeave is not one, and there are three serious objections to reading it this way.

Coverage measured mid-build always looks bad

CoreWeave borrows for equipment months before that equipment earns revenue. Interest on the debt starts when the money is drawn, while revenue starts when capacity is delivered to the customer. A company growing revenue 112% a year will always carry interest on assets that are not yet earning. The ratio penalises growth itself.

The strongest version of the bull case is the backlog. CoreWeave reported $103.7 billion of remaining performance obligations at 30 June 2026, 41% of it expected within 24 months. If that converts on schedule, EBITDA grows into the interest bill. The coverage ratio is a snapshot of a company at the most indebted moment of each expansion cycle.

The debt is partly matched to contracts

Several of the delayed draw facilities are structured around specific customer contracts. The non-recourse DDTL 4.0 facility is the cheapest debt in the stack, at an effective 7%. Lenders on those facilities look at the contract cash flows, not at consolidated interest coverage. A consolidated ratio blends secured, contract-backed debt with unsecured notes that carry very different risk.

The ratios here use only two filings

I have not computed Q3 or Q4 2025, because those filings were not part of this research. A six-quarter series with two gaps can hide a turn. And the $191 million of interest expense outside the debt note, likely lease-related, is not explained here. A reader with the full 10-Q text should check that line before relying on my split.

None of this changes the core finding. It does mean the 0.18 times figure describes an accounting period, not a verdict on solvency. My position is narrower: CoreWeave pays its lenders from cash flow and from new borrowing, and the earnings to pay them on a depreciated basis do not exist yet.

What CoreWeave's ratios say about neocloud debt generally

CoreWeave is the only listed neocloud with several quarters of audited data at this scale. Its ratios are becoming the reference point for the category. Other GPU cloud operators raising debt or equity are compared against it by lenders and investors. That is visible in the coverage of the Firmus IPO and the neocloud listing window.

The lesson for the category is about the break-even utilisation of the hardware. A neocloud with 25 cents of interest per revenue dollar has very little margin for idle capacity. If utilisation slips, revenue falls while interest does not, and coverage on any measure drops quickly. The arithmetic of that trade is laid out in our GPU break-even utilisation model.

I would apply one rule to any neocloud credit. Ask for interest expense including capitalized interest, divided by revenue, for the last four quarters. If the company cannot produce that number, or the trend is rising, treat its EBITDA coverage as the most optimistic reading available rather than the base case.

Frequently asked questions

Related reading