CoreWeave Debt Maturities: When the $35.6 Billion Comes Due, Year by Year

CoreWeave owed $35.6 billion of debt principal at 30 June 2026, and its own 10-Q says exactly when each dollar falls due. Only $10.6 billion lands before the end of 2027. The heavier block, $14.9 billion, sits after 2030. Set the CoreWeave debt maturity schedule against the backlog the company says it will convert, and the stress point moves years later than most coverage puts it.

The short answer. CoreWeave's debt maturities run $4.4 billion in the rest of 2026, $6.2 billion in 2027, $4.4 billion in 2028, $2.4 billion in 2029, $3.2 billion in 2030 and $14.9 billion thereafter. Most of the early money amortizes from customer contracts. The refinancing question sits in 2031 to 2033, where bonds and convertibles mature.

The CoreWeave debt maturity ladder, year by year

A maturity ladder is the schedule of when borrowed principal must be repaid. It is a different question from how much debt a company holds, and for a lender it is often the more important one.

CoreWeave prints its ladder in the Debt note of each quarterly filing. The figures below come from the Q2 2026 10-Q, filed in August, which is the latest quarterly filing available on 27 October 2026. The third quarter report is due in November.

Principal amounts are the company's own, from the 10-Q maturity table. Shares and cumulative percentages are our arithmetic. Gross principal is higher than the $35.1 billion balance sheet carrying value because of $483 million of unamortized discount and issuance costs.

My read of this chart is simple. The 2027 bar is the one that gets quoted, and it is the least interesting bar on the page. The right-hand bar is where CoreWeave's dependence on the capital markets actually concentrates.

Two ladders hiding in one schedule

The single maturity table blends two kinds of debt that behave very differently. Separating them is the most useful thing you can do with this filing, and the 10-Q gives you enough to do it.

The Debt note lists every instrument and its gross principal at 30 June 2026. Grouping them by type produces two stacks of roughly equal size.

Instrument balances and effective rates as printed in the 10-Q schedule of debt. Grouping and subtotals are ours. The revolving credit facility had nothing drawn at 30 June 2026.

The amortizing stack repays itself, if the contracts hold

The delayed draw term loans, or DDTLs, are facilities a lender funds in stages as GPUs are delivered. Each is secured on specific infrastructure and the customer contracts running on it, and principal is repaid from the cash those contracts produce.

Take the bullet notes out of the yearly buckets and the pattern is stark. Of the $20.655 billion scheduled through the end of 2030, $2.0 billion is the 2030 notes. The other $18.655 billion is amortization, which is about 98.5% of the entire amortizing stack. These loans are designed to be gone by 2030.

That design is why the near-term buckets look less dangerous than their size. A DDTL payment is not a refinancing event. It is a contract paying down the debt that built it, the way the facility-by-facility breakdown of CoreWeave's secured lending describes.

The bullet stack needs a market

Senior notes and convertibles repay in one lump at maturity. Nothing in the business amortizes them along the way. They get repaid from cash on hand, from new borrowing, or, for the convertibles, by issuing shares.

Here the filing is unambiguous. Instruments named for 2031 and 2032 total $14.617 billion, which is 98% of the $14.896 billion thereafter bucket.

How the CoreWeave maturity ladder moved in 2026

One snapshot hides the direction of travel. The same table appears in the 2025 10-K and the Q1 2026 10-Q, and comparing all three shows what management did with the first half of the year.

Yearly figures from the XBRL data filed with each report. The 31 March 2026 thereafter figure is derived by subtracting the five tagged buckets from total principal, because the API returned no tagged value for it.

The 2027 bucket grew 44% in six months

Principal due in 2027 rose from $4.298 billion at year-end to $6.184 billion at 30 June. The 2028 bucket nearly doubled, from $2.393 billion to $4.416 billion. That is new amortizing debt being drawn to fund capacity, and each new DDTL adds a slice to every year of its life.

This is the number I would watch most closely in the next filing. A growing near-term bucket is fine while the matching contracts start paying on time. It becomes a problem only if capacity comes online late and the amortization schedule does not move with it.

The company pushed $10.5 billion past 2030 in one quarter

Between March and June, the thereafter bucket jumped from $4.387 billion to $14.896 billion. The Q2 schedule of debt shows where it came from: the 2031 9.75% notes, the 2032 9.625% notes, the EUR 2.0 billion 2032 notes and the $4.0 billion 2032 convertible, all new in 2026.

That is a classic terming-out move, and a sensible one. It also concentrates the refinancing need. In the first half the company raised $16.7 billion of new debt and repaid $5.2 billion. The ladder is being rolled, not retired.

CoreWeave maturities against backlog conversion

Principal due means little without the cash expected to meet it. CoreWeave does not forecast cash flow by year. It does disclose when it expects its contracted backlog to turn into revenue, and that is the closest public proxy.

The 10-Q states that the company had $103.7 billion of remaining performance obligations. It expects to recognize 41% of that in the 24 months to June 2028, 39% in months 25 to 48, and the rest in months 49 to 78.

The backlog windows run from July, and the maturity table runs by calendar year. To line them up, we spread each window's backlog evenly across its months and summed by year. That is a simplifying assumption, and real recognition will be lumpier as data centres switch on.

The working: window one is $42.5 billion over 24 months, about $1.77 billion a month. Window two is $40.4 billion, about $1.69 billion a month. Window three is $20.7 billion over 30 months, about $0.69 billion a month.

Backlog revenue is Zan Digital's estimate from the disclosed recognition windows. It is revenue, not cash available for debt service, and it counts only contracts signed by 30 June 2026. Principal due is the company's own figure.

What the cover multiple does and does not say

Between 2027 and 2030, contracted revenue already in hand runs between 3.4 and 8.4 times the principal due. On that view, the amortizing ladder is well covered by existing customers, which is consistent with how the secured lenders priced it.

The 2031 and 2032 pair is the outlier. Today's backlog would bring in roughly $16.6 billion across those two years, against $14.9 billion of principal. I would not call that a crisis. I would call it the point where the debt stops being matched to contracts and starts being matched to faith in the next round of contracts.

The real CoreWeave refinancing wall is 2031 to 2033

Put the pieces together and the wall has a date. It is not 2027. It is a three-year band of bullet maturities that starts in 2031 and, after September, runs into 2033.

At 30 June 2026 the band held two senior notes due 2031 at $4.5 billion combined and the 2031 convertibles at $2.588 billion. It also held the 2032 dollar and euro notes at $3.529 billion and the 2032 convertibles at $4.0 billion.

Then, after the quarter closed, CoreWeave priced $3.7 billion of 2.875% convertible senior notes due 2033, upsized from $3 billion, with an option for $500 million more. That deal is not in the June ladder. Add it, and the 2031 to 2033 bullet band reaches about $18.3 billion before the option.

Convertibles are cheap money with a condition attached

Convertibles make up $10.3 billion of that $18.3 billion. The 10-Q shows the 2031 and 2032 issues carrying an effective rate of about 2%, against roughly 10% on the straight notes. That gap is the whole appeal.

It also creates the main refinancing risk. If the share price is above the conversion price at maturity, holders take stock and no cash changes hands. The 2032 notes convert at $119.60 a share. If the stock sits below that, the company owes cash, and replacing roughly $6.6 billion of 2031 and 2032 convertibles at straight-debt rates near 10% would add about $527 million a year of interest. That last figure is our illustration, not a disclosure.

So the 2031 to 2033 question is really an equity question wearing a credit costume. The interest burden that a cash refinancing would layer on is the subject of the analysis of CoreWeave's interest coverage per revenue dollar.

What CoreWeave's 2026 and 2027 maturities actually need

None of this makes the near term trivial. It changes what you are checking for.

The $4.413 billion due in the second half of 2026 and the $6.184 billion due in 2027 are overwhelmingly amortization. The balance sheet held $5.5 billion of unrestricted cash at 30 June, and the revolving credit facility showed $2.0 billion of undrawn capacity.

The subsequent events note adds two things. The $2.6 billion DDTL 5.5 facility closed in August at SOFR plus 5.50%, and the filing records a $1.2 billion draw in the same month. Liquidity was being used, not hoarded.

The test for 2027 is operational. Each DDTL is sized to contracts that only pay once capacity is live, so a delayed site turns a self-funding loan into one the parent must carry. Delivery dates, not credit spreads, are the leading indicator for this bucket. Where those loans end up on lender balance sheets is covered in the piece on private credit's exposure to AI infrastructure.

Where this argument is weakest

The 1.1x figure is the headline of this post, and it is also its most fragile number. Here is the case against it.

Backlog that far out has simply not been signed yet

Every infrastructure company's contracted revenue thins out in later years, because customers sign three-to-five-year deals and renew them. CoreWeave's backlog more than tripled in the year to June 2026, rising 246% by the company's own count in its Q2 results. If even part of that pace continues, the 2031 to 2032 cover will look nothing like 1.1x by the time it matters.

That is the strongest objection, and I think it is mostly right. The honest reading is not that 2031 is underfunded. It is that 2031 is unfunded by contracts signed so far, which is a narrower claim.

Revenue is not cash available for debt

The cover multiples compare revenue to principal. Operating costs, interest, taxes and, above all, new capital spending come out first. On that basis every multiple in the table overstates true cover, the early years included. The ratio is useful for shape, not for level.

The method smooths what is lumpy

Spreading each backlog window evenly by month is an assumption we chose for transparency. Real recognition depends on when each site goes live and how each contract ramps, which the filing does not disclose. Treat the calendar split as illustrative, as the table caption says.

Finally, the instrument grouping relies on the maturity year in each note's name. The filing data we read did not print a maturity date beside every instrument, so a note named for 2031 is assumed to mature in 2031.

What to read in CoreWeave's next 10-Q

The third quarter report, due in November 2026, will be the first to include the September convertible and the full DDTL 5.5 draw. Three lines in the Debt note will tell you whether the shape described here still holds.

First, the thereafter bucket. It should rise by roughly the $3.7 billion convertible plus any option exercised. A bigger jump means more debt was termed out. Second, the 2027 bucket. Steady growth is normal; a jump without matching backlog growth is not. Third, the cover between backlog windows and principal, which you can rebuild with the method above in 10 minutes.

If you only have time for one comparison, set the change in backlog against the change in total principal. In the first half of 2026 both grew fast. The quarter they diverge is the one to read twice. For context on how other cloud vendors report the same backlog metric, see the comparison of cloud backlog and RPO disclosures.

Frequently asked questions

Where to start

Open the Debt note in the next CoreWeave 10-Q and copy the maturity table into the three-column comparison above. It takes less time than reading the earnings release and tells you more about the next five years.

If you buy GPU capacity from any neocloud, ask your account team one question before you renew: which facility finances the capacity you sit on, and does it mature before or after your contract ends? The answer tells you whose refinancing risk you are quietly underwriting.

Related reading