CoreWeave Stock: Redburn’s $54 Sell vs a $97.85…

The common read on Monday’s CoreWeave (NASDAQ: CRWV) Sell call is that the credit market has finally turned on the neoclouds, and the equity market simply has not noticed yet. CoreWeave’s own capital stack says something more awkward: in the same week, the credit market priced two very different CoreWeaves. On 18 September, investors lent the company $3.7 billion at a 2.875% coupon through convertible notes that only convert at about $97.85 a share. Three days later, Rothschild & Co Redburn initiated CoreWeave at Sell with a $54 price target, which sits 44.8% below that conversion price and 36.8% below Monday’s $85.43 close, per Nasdaq. The stock rose 5.0% on the day of the Sell call. The useful question is not whether credit is worried about CoreWeave. It is which slice of credit you are listening to, and what the answer means for anyone who rents, finances or trades against GPU capacity.

Here is the synthesis nobody has put side by side. CoreWeave’s straight, unsecured debt is priced like high yield: the company sold 9.625% senior notes due 2032 in June, and its August term loan pays Term SOFR plus 5.50%. Its convertible paper, by contrast, costs 2.875%. That 675 basis point gap is not a credit opinion at all; it is the price of the equity option buried in the convert, and convertible buyers typically hedge that option by shorting the stock. So the cheapest money CoreWeave raised this month is a volatility trade, not a vote of confidence, and the most expensive money is the honest credit signal. Redburn is right about which one matters. Having tracked CoreWeave’s financing filings since its IPO, I think the real tell is neither coupon: it is that interest is now running at roughly a quarter of revenue, and the midpoint of management’s own third-quarter guidance nudges that ratio higher, not lower.

What Redburn Is Actually Saying, and Why the Stock Rose Anyway

Redburn’s note is broader than a single-stock call. The firm started coverage on two neoclouds and six Bitcoin miners that have pivoted to AI hosting, putting Core Scientific, TeraWulf, Applied Digital, Cipher Mining, Hut 8 and IREN at Neutral, according to GuruFocus. Its central argument is that hyperscalers carry more economic leverage than their balance sheets show once off-balance-sheet commitments are counted, and that this could cap the next leg of AI infrastructure spending. The neoclouds sit at the end of that chain: they borrow to buy GPUs, then rent them to the same hyperscalers and AI labs.

The mechanics of the bear case are easiest to understand through an analogy from aircraft leasing. A lessor buys a jet with borrowed money and leases it to an airline for years. The deal works if the lease rate stays above the cost of the debt plus depreciation. It breaks if new jets flood the market, if airlines start buying their own fleets, or if refinancing gets more expensive. 24/7 Wall St. summarised Redburn’s three pressures in almost exactly those terms: falling GPU rental prices, hyperscalers scaling captive fleets, and rising financing costs on multi-billion-dollar buildouts.

Redburn sees more downside in Nebius than in CoreWeave. Its $84 Nebius target is 63.9% below Nebius’s $232.80 close on Monday, while the $54 CoreWeave target implies a smaller fall. The CoreWeave target is also 61% below the $139.37 consensus average compiled by MarketBeat, where 21 analysts rate the stock a buy, 10 a hold and 4 a sell.

Why did CoreWeave close up 5.0% on the day? Some of the move was sector flow: accounts on X tracking Leopold Aschenbrenner’s former positions flagged CoreWeave, Nebius, Bloom Energy and Micron all rising together on Monday. More of it was relief. The stock had already fallen 4.2% on 17 September, from $83.35 to $79.88, when CoreWeave announced the convert and a 35-million-share at-the-market programme on the same day. FinanceFeeds covered the earlier leg of this sell-off in its look at why CoreWeave and Nebius dropped about 20% in a month. A Sell call that restates known financing risks lands differently on a stock that has already absorbed a dilution announcement.

Redburn put its thesis in one line: “Credit markets are beginning to price risks that equities largely ignore,” the firm wrote, as quoted by GuruFocus.

How CoreWeave Is Responding: Pricing Power, Contract Length and Cheaper Debt

CoreWeave has not publicly commented on the Redburn note. Its filings, though, answer each of Redburn’s three pressures, and they are worth reading as a rebuttal.

On GPU rental prices, the company says its prices are going up. On the second-quarter call, CFO Nitin Agrawal said July pricing changes included “an approximately 25% increase across SKUs,” and the 17 September prospectus supplement adds a hard number: short-dated contracts signed since 30 June, with terms of three to six months, priced at about $40.0 million per megawatt of annualised revenue. The same document says CoreWeave added more than $25 billion of net new customer commitments early in the third quarter and lifted contracted power to about 4.2 gigawatts by 11 August, from 3.7 gigawatts at the end of June.

On hyperscaler captive fleets, CoreWeave’s answer is that hyperscalers are its customers. The 10-Q lists Microsoft and OpenAI as significant customers and discloses that in March 2026 Meta Platforms committed to pay up to approximately $21.0 billion for capacity running to December 2032. Nebius, the other name Redburn rated Sell, took a similar line last week when it told customers it would raise GPU rental rates, a move the market rewarded with a roughly 10% jump.

On financing costs, management says they are falling. Agrawal told analysts CoreWeave had cut its weighted average cost of debt “by almost 300 basis points, representing approximately $1.1 billion of annualized interest savings based on our end of Q2 debt load.” The August term loan also adds a new tool: CEO Michael Intrator said on the call that it lets CoreWeave sell compute on shorter terms “to extract additional margin.”

Intrator’s broader claim is the one the Sell call directly disputes. “Pricing and margins for our Blackwell and Vera Rubin SKUs are setting new highs, while pricing for prior generation SKUs is at or above where it was years ago,” he said on the second-quarter earnings call. Agrawal, closing the same argument, conceded where the risk sits: “The cost, primarily in the form of CapEx is front-loaded requiring a combination of debt, customer prepayments, and other corporate level capital to finance its build out.”

The Numbers: Interest Is Eating a Quarter of Revenue

Combine the income statement with management’s guidance and the pressure point becomes clear. Second-quarter interest expense of $640 million was 24.9% of $2.575 billion in revenue. The third-quarter guide of $860 million to $940 million in interest against $3.45 billion to $3.6 billion in revenue works out to 23.9% at best and 27.2% at worst, with a midpoint of about 25.5%. Revenue is growing fast, but debt is growing faster, and that is before the new $3.7 billion convert is counted.

Adjusted EBITDA covers the interest bill, but not by much. CoreWeave reported $1.51 billion of adjusted EBITDA in the second quarter, a 59% margin, which is about 2.4 times interest expense. On a GAAP basis, the company posted an operating loss of $49 million. The gap between those two figures is mostly depreciation on the GPU fleet, which is exactly the line Redburn is questioning.

Cash flow tells the same story. Operating cash flow was $3.66 billion in the first half, while purchases of property and equipment were $14.12 billion, according to the 10-Q. Management has guided full-year 2026 capex to $35 billion to $39 billion. That gap is funded by debt, prepayments and equity, which is why the ATM programme matters: 35 million shares at Monday’s close is roughly $3.0 billion, or about 7.6% of the 458.9 million Class A shares outstanding at 31 July.

The bull case sits in the backlog. Remaining performance obligations of $103.7 billion, with 41% expected within 24 months, imply about $42.5 billion of revenue over two years, or roughly $21 billion a year. That is well above the 2026 revenue guide of $12.4 billion to $13.2 billion, so the business is contracted to grow into its debt, provided the capacity arrives on time and customers pay.

The table shows where the Sell call and the convert disagree, and where they do not. Nobody is lending CoreWeave straight money cheaply. The low coupon only appears when lenders get equity upside, and about $498.8 million of the convert’s roughly $3.64 billion in net proceeds went on capped calls to limit dilution from that upside. The Real Investment Advice team noted on 4 September that CoreWeave’s five-year bonds yielded nearly 13%, and cited Fitch’s finding that 65% of first-quarter revenue came from two customers.

The pattern is familiar from other parts of the AI build-out. FinanceFeeds found the same split at Oracle, where Project Jupiter loans quoted near 90 cents are trading about where Oracle’s own bonds already sit. In both cases the headline says credit is cracking, while the filings show credit had been charging a risk premium all along.

Regulatory and Legal Pressure Points

The regulatory tension around CoreWeave runs through three channels: disclosure, power and derivatives.

Disclosure first. Both September financings were structured to raise money quickly: the convert was sold privately to qualified institutional buyers under Rule 144A, and the share sales run off a shelf registration filed in June. That speed has a cost in scrutiny. The 10-Q discloses a putative securities class action, Masaitis v. CoreWeave, filed in the District of New Jersey on 12 January 2026 under Sections 10(b) and 20(a) of the Exchange Act, plus related derivative suits. The company says the claims are without merit. For a company raising money every few weeks, the quality of its disclosures on contract terms, customer concentration and depreciation is becoming a legal exposure as well as an investor one.

Power is the second channel. Redburn’s argument partly depends on who ends up paying for the grid connections behind gigawatt-scale data centres. The House passed the Ratepayer Protection Act 417 to 3 on 16 September, a bill aimed at who bears data centre power costs, which FinanceFeeds examined in its look at who pays for data centre power. If that cost moves onto operators, neocloud unit economics get worse at exactly the point Redburn is focused on.

The third channel is market structure. CME Group plans to list futures on H100 and B200 rental prices on 5 October, subject to regulatory review. Those contracts settle against indices of on-demand neocloud rental rates. For the first time, the “falling GPU rental prices” in Redburn’s thesis and the “25% increase across SKUs” in CoreWeave’s will be checked against a public, exchange-traded forward curve. The 10-Q also flags U.S. export controls under the Export Administration Regulations as a supply risk, another reminder that the price of a GPU-hour is partly set in Washington.

What Happens Next

Three things should settle the argument between Redburn and CoreWeave within about two months.

1. The CME GPU curve. If the H100 and B200 futures launch on 5 October and the forward curve slopes down steeply, Redburn’s first pressure is confirmed by a market price rather than an analyst model. Because the indices track on-demand neocloud rates rather than CoreWeave’s long-term contracts, the first effect would be felt on short-dated deals like the $40 million-per-megawatt contracts CoreWeave disclosed in September. I expect the curve to show backwardation, meaning lower prices further out, for H100s, where the newer Blackwell supply is arriving, and much less for B200s.

2. Third-quarter interest versus revenue. CoreWeave’s Q3 results, expected in November, will show whether interest lands inside the $860–940 million range and revenue inside $3.45–3.6 billion. A ratio above 26% would support the Sell case. A ratio below 24% would support management’s claim that the cost of debt is falling faster than the debt is growing.

3. How much of the ATM gets used. Any shares sold under the 35-million-share programme will show up in the next 10-Q. Heavy use at prices near $80 would suggest the company prefers dilution to more 9%-plus debt, which would validate Redburn’s view of financing costs even if revenue keeps beating guidance.

My read: Redburn has correctly identified the pressure point, but a $54 target assumes the backlog converts into cash more slowly than the filings indicate. The stock’s next big move is more likely to come from the GPU futures curve and the Q3 interest line than from any analyst rating, including this one. For wider context on the AI trade, see FinanceFeeds’ analysis of what a deliberate AI slowdown would do to $700 billion in capex.

FAQ

Why did Rothschild & Co Redburn rate CoreWeave a Sell?

Redburn argues that falling GPU rental prices, hyperscalers building their own fleets and high financing costs will squeeze neocloud returns. It initiated CoreWeave at Sell with a $54 price target on 21 September 2026 and said credit markets are pricing risks that equities largely ignore.

What is CoreWeave’s convertible note conversion price?

CoreWeave’s $3.7 billion of 2.875% convertible notes due 2033 convert at about 10.2194 shares per $1,000 of principal, an initial conversion price of roughly $97.85. That was a 22.5% premium to the $79.88 close on 17 September 2026.

How much debt does CoreWeave have?

At 30 June 2026, CoreWeave reported about $31.4 billion of recourse debt and $3.7 billion of non-recourse debt, net of discounts, for a total of roughly $35.1 billion. Since then it has signed a $2.6 billion delayed-draw term loan and priced $3.7 billion of convertible notes.

Is CoreWeave profitable?

Not on a GAAP basis. In the second quarter of 2026, CoreWeave reported an operating loss of $49 million and a net loss of $626 million on $2.575 billion of revenue. Adjusted EBITDA was $1.51 billion, a 59% margin, but interest expense was $640 million.

What is CoreWeave’s revenue backlog?

CoreWeave reported $103.7 billion of remaining performance obligations at 30 June 2026, with 41% expected to be recognised within 24 months. The company also said it added more than $25 billion of net new customer commitments early in the third quarter.

How does the Redburn call compare with other analysts on CoreWeave stock?

Redburn’s $54 target is well below the $139.37 consensus average compiled by MarketBeat, where 21 analysts rate CoreWeave a buy, 10 a hold and 4 a sell. It is one of the most bearish targets on the stock.