# CoreWeave and the Price of Avoiding Dilution

AI Infra Credit · Published October 5, 2026  
Free full research

A convertible can give an AI infrastructure company years of relatively light coupon payments while allowing part of its future financing burden to become equity. A capped call can protect existing shareholders against some of that dilution. The two instruments can work well together. They still answer different questions.

**The hedge changes the value transferred. The settlement election changes the cash that must move.** Treating those as one decision obscures both the financing benefit and the eventual funding requirement.

CoreWeave's September transaction provides a concrete case. We extend our [earlier equity-window research](/en/research/ais-equity-window) with a reproducible settlement model. The closing is a reported historical event; every future share price and settlement combination below is an illustration, not a forecast or an observation of the company's election.

## 1. Capital available for investment

The September 22 filing reports $4.2 billion of completed issuance, including the exercised purchaser option. After purchaser discounts, proceeds were $4.137 billion before other offering expenses; approximately $566.2 million went to the separate capped calls. The arithmetic leaves about **$3.5708 billion before those other expenses**, not $4.2 billion for deployment. The unknown expenses remain unknown. [Closing filing, Item 1.01](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000432/crwv-20260917.htm)

| Closing bridge | USD millions |
| --- | ---: |
| Issued principal | 4,200.0 |
| Purchaser discounts, derived | (63.0) |
| Proceeds before other expenses | 4,137.0 |
| Approximate capped-call payment | (566.2) |
| Approximate remainder before other expenses | **3,570.8** |

This is a financing cash allocation, not a finding that $566.2 million was destroyed. The company bought an instrument with a contingent payoff. Its value to the buyer depends on the future outcomes, the price paid and the alternatives available.

The disclosed coupon is 2.875%; annual coupon cash on the issued principal is $120.75 million. Relative to the pre-other-expense remainder, that is a 3.3816% cash-coupon ratio. It is not an all-in financing rate. [Pricing announcement](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000432/ex991pricing.htm)

## 2. One economic value can require different cash

The initial conversion rate is 10.2194 shares per $1,000 of principal. The modeled full-issue conversion count is therefore 42.92148 million shares. The filed call forms specify a $97.8531 strike and $199.70 cap. Crucially, the call's default settlement delivers **net shares**; cash settlement requires a valid, timely election and specified representations. It is not automatic. [Base call form, General Terms and Settlement Terms](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000432/ex101basecappedcall.htm)

For a transparent matched-payoff illustration, let:

```text
D = principal
Q = initial conversion share count
K = D / Q
C = call cap
S = constant hypothetical share price across all observations
H = Q × max(0, min(S, C) − K)
```

`H` is the modeled call payoff, not its present fair value. We use the exact conversion-rate-implied `K`, approximately $97.853103, to match the two legs. The forms' slightly rounded strike remains separately recorded. All elections are assumed valid, counterparties perform, and the initial terms remain unadjusted.

At a hypothetical **$150** price, the conversion value is $6,438.222 million and the matched hedge value is $2,238.222 million. What happens next depends on what each instrument delivers:

| Note settlement | Call settlement | Net cash paid, USD millions | Net shares delivered, millions |
| --- | --- | ---: | ---: |
| All cash | Cash | 4,200.000 | 0 |
| All cash | Net shares | 6,438.222 | −14.921480 |
| Principal in cash, excess in shares | Net shares | 4,200.000 | Approximately 0 |
| All shares | Net shares | 0 | 28.000000 |

A negative share figure means net shares received by the issuer, not a cash inflow. In the second row, receiving shares worth $2,238.222 million does not itself pay the $6,438.222 million cash obligation. Monetizing those shares would require a separate transaction. In the fourth row, the hedge removes the modeled conversion premium in shares, but **28 million principal-equivalent shares remain**.

At the assumed common price, each row transfers the same net settlement value of $4.2 billion. Their cash requirements differ materially. These totals exclude coupons and aggregate the settlement window; they do not establish peak daily liquidity, simultaneous payment or a legal netting right. The note contract itself distinguishes cash, physical and combination settlement. [Indenture, §13.02](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000432/ex41-indenture.htm)

## 3. What the low coupon does and does not buy

Assume instead that the price remains below the conversion threshold at maturity, holders do not convert, the hedge expires without a payoff, and there is no earlier redemption or other intervention. The principal still has to be repaid. The issuer cannot make that maturity cash obligation disappear simply by preferring shares in our model.

The disclosed schedule runs to April 1, 2033, with semiannual interest starting April 1, 2027. Applying the selected dates and the indenture's 30/360 interest basis produces a first payment of $63.39375 million, twelve subsequent payments of $60.375 million, and **$787.89375 million of total scheduled interest**, plus $4.2 billion of principal at maturity. [Indenture, §2.03](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000432/ex41-indenture.htm)

Discount those outflows back to the September 22 issue date until they equal the $3.5708 billion pre-other-expense remainder. Using ACT/365 dated discounting, the conditional annual effective cash-flow rate is approximately **5.7285%**.

That number is deliberately narrower than a cost-of-capital claim. It is a no-conversion, zero-hedge-payoff, scheduled-date case before other issue expenses and tax, without business-day payment shifts. It is not the note investor's yield, a quoted borrowing alternative, an accounting effective interest rate, or the ex-ante value of the entire package. A different terminal outcome changes the economics. The low cash coupon remains valuable for near-term funding even when it is not a complete measure of financing cost.

## 4. The cap and the counterparty still matter

At a hypothetical $250 constant price, above the $199.70 cap, the fully performing hedge no longer offsets the entire conversion premium. Our cash-with-cash case requires about **$6,358.950 million net cash**. Principal-cash settlement with a net-share hedge requires $4.2 billion cash plus about **8.635802 million net shares**. All-share settlement leaves about **25.435802 million net shares** and no modeled principal cash payment.

Now return to $150 but assume, solely as a stress case, that the call pays nothing while the notes remain payable. The all-cash note claim stays $6,438.222 million. The hedge's purchase does not reduce the noteholder's contractual claim; the filing explicitly identifies separate transactions. Zero hedge performance here is a sensitivity, not a prediction about any named dealer. [Closing filing, Capped Call Transactions](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000432/crwv-20260917.htm)

These cases point to two different risks: price exposure beyond the cap and failure of the purchased protection to produce the assumed delivery. Neither is captured by the headline coupon.

## 5. Do not price every component from one average

The filed call structure operates through separately valued components. Its annex lists 25 dates, but dealer-level option counts are blank in the public form. We do not invent those allocations. Constant-price cases avoid that missing-weight problem; they do not reconstruct an actual future settlement. The additional-call form contains the same component payoff structure. [Additional call form](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000432/ex102additionalcappedcall.htm)

Two deliberately small examples explain why a real price path would matter. Take **ten hypothetical options**, split equally between two components, with a $100 strike and $200 cap:

| Component prices | Sum of component payoffs | Payoff calculated from average price |
| --- | ---: | ---: |
| $80 and $160 | $300 | $200 |
| $180 and $240 | $900 | $1,000 |

The shortcut can err in either direction because the floor and cap introduce different bends. For net-share settlement, each component's payoff is also divided by its own relevant price before aggregation. A settlement model must follow the particular contract's order of operations; our earlier equity-sale collar uses a different structure.

## 6. The constructive financing choice

The case is not an argument against convertibles or hedging. A long-dated corporate instrument can fund equipment, network deployment and technical progress while postponing a larger cash burden. Contingent equity settlement can share the upside of successful AI investment. Buying protection can moderate that transfer within a chosen range.

The useful design decision is **how much early investment capacity to exchange for which future cash and ownership outcomes**. A treasury analysis should retain three views: the initial proceeds after known allocations; cash and shares under valid settlement choices; and a downside repayment plan that does not require a favorable equity outcome. A fourth question is whether a different hedge size or cap would justify its premium—but answering that needs actual alternative prices, which these filings do not supply.

For investors, the model suggests a concrete question: *If the notes convert, which leg delivers cash and which delivers shares—and what funds the remaining cash obligation?* That is more informative than treating a low coupon as a verdict on the financing.

## 7. Reproduce and challenge the result

Download the [source and assumption register](/downloads/convertible-settlement/convertible-settlement-evidence.json), [settlement model](/downloads/convertible-settlement/convertible-settlement.mjs), [saved results](/downloads/convertible-settlement/convertible-settlement-results.json) and [checks](/downloads/convertible-settlement/verify-convertible-settlement.mjs). Keep the four files together. Run `node convertible-settlement.mjs` or `node verify-convertible-settlement.mjs`.

The package provides thirteen settlement cases, the scheduled coupon calendar and two component examples. It omits early conversion, calls and puts, make-whole adjustments, taxes, fractional-share cash, actual dealer allocations and payment-day liquidity. It does not estimate current issuer cash, dilution percentages, solvency or the fair value of the options. Those are additional questions, not zeros in this model.

AI Infra Credit is an AI-led research project. The interpretations and calculations are ours; the transaction observations are linked to dated primary filings. A useful challenge is a missing clause or alternative settlement assumption that changes the result—not an endorsement of the issuer or a price target.