The Margin Turned Before the Debt Did: Upgrading CoreWeave to Outperform on the Trigger We Set in May

Results vs. Consensus

Year-Over-Year Comparison

The year-over-year frame remains the least flattering way to read this company, and it should be shown anyway. Adjusted operating margin is still 1,150 basis points below where it stood a year ago, because the second quarter of 2025 was a harvesting quarter and this one is a building quarter. General and administrative expense growing 2.3% against 112% revenue growth is the second consecutive quarter in which the corporate cost base has decoupled from revenue, and it is the cleanest evidence in the table that scale is starting to work. Every other expense line still grew faster than revenue.

One item deserves flagging because it flatters the bottom line without touching operations. Other income of $125M against $6M a year ago is primarily an unrealised gain on investments, and $109M of it is removed again in the adjusted net loss reconciliation. The adjusted figure is the honest one.

Quarter-Over-Quarter Comparison

Two rows in that table changed direction. The interest-to-operating-profit ratio fell for the first time in this coverage, and operating expenses grew 5.8 percentage points slower than revenue, a gap that last pointed this way in the third quarter of 2025. Everything else is the familiar CoreWeave shape: debt up 41% in a quarter, capex up 38%, receivables rising with revenue, and deferred revenue up more than $2B as customers prepay for capacity they have contracted but not yet received.

The deferred revenue move is worth pausing on. A $2,169M sequential increase in the deferred revenue balance, against $2,575M of revenue recognised in the same quarter, means customer commitments funded a meaningful share of this quarter's build, though the cash flow statement puts the quarter's collected deferred-revenue contribution at $790M, with the balance of the increase billed rather than yet collected. It remains the cheapest capital in the structure and none of it appears in the debt line.

Quality of the Quarter

Revenue: A beat of 0.6% against consensus is the narrowest of this coverage, and it is the least important line in the release. Sequential growth of 23.9% landed in the upper half of a range set ten weeks earlier, and the composition improved: management raised list prices roughly 25% across SKUs in July, described Blackwell and Vera Rubin pricing and margins as setting new highs, and said prior-generation pricing sits at or above where it was years ago. Revenue that arrives with rising price is worth more than the same revenue bought with volume.

Margins: The line that changes the rating. Adjusted operating income of $128M against a $30–90M guide is a 42% beat on the ceiling, and it arrived in the quarter management had told us to watch. The mechanism put on the record in May was that a newly energised site carries one to two months of lease, power and depreciation cost before revenue arrives, with contribution margins normalising by month three. That claim was falsifiable, it has now been tested once, and it passed. What makes the beat harder rather than easier is the timing: nearly 500 MW was energised during the quarter, more than in any quarter of the company's history, and roughly 300 MW of it landed in June. The margin expanded while the drag was at its heaviest.

EPS: Adjusted loss per share of $(1.03) against $(1.20) expected, and against $(1.12) last quarter. The improvement is genuine but modest in context. The loss narrowed because operating profit rose $107M while interest expense rose $104M, so almost exactly the whole operating gain was consumed by the incremental cost of the debt that funded it. That is the arithmetic of this equity in one sentence, and it has not changed. What changed is the sign of the first term.

Business Drivers

Backlog: $104.2B, and the Growth Has Moved Out of the Reported Number

A $4.8B sequential increase looks like a sharp deceleration against the $32.6B added in the first quarter, and read alone it would be the bearish datapoint of the release. It is not, for a reason stated in a footnote to the press release and repeated on the call: more than $25B of net new customer commitments were signed in the early weeks of the third quarter and are excluded from the $104.2B. Add them and the two-quarter booking pace is roughly $30B, in line with the first quarter rather than a fraction of it.

The more useful disclosure is new. More than half the backlog is now attached to contracts where delivery has already commenced, and management expects that to exceed two-thirds by year end. Backlog quality has been the standing objection to this figure since we initiated: a very large number denominated in years, with no way to tell how much of it depended on capacity that did not yet exist. A delivery-commenced ratio is the first metric that speaks to it directly.

"Of the existing backlog, more than 50% is attached to a contract where customer delivery has commenced. We expect this figure to reach more than 2/3 of our Q2 backlog by the end of this year." — Nitin Agrawal, Chief Financial Officer

Assessment: Bookings did not slow; the reporting boundary moved. The delivery-commenced ratio is the single most useful new disclosure in the release, because it converts backlog from a promise into a schedule. We would like the same figure for the prior two quarters so it can be trended, which is the one thing it currently lacks.

Capacity: 1.5 Gigawatts Active, and a June That Was Bigger Than Any Prior Quarter

Active power roughly tripled year over year and grew by nearly 500 MW in a single quarter, more than the company had added in any prior quarter and, on management's characterisation, more than any single competing neocloud operates in total. The deployment ratio, which we have tracked since the third quarter of 2025 as the cleanest measure of whether contracted power is a real asset or a press release, reached 40.5% from roughly 20% four quarters ago.

"In fact, in June, we brought on more than 300 megawatts of active power, which makes June itself larger than any full quarter in our history." — Nitin Agrawal, Chief Financial Officer

The June concentration matters for the third quarter more than for the second. Capacity energised in the final month of a quarter contributes almost nothing to that quarter's revenue and all of its cost, which is the fit-out drag management quantified in May. Construction in progress rising to $11.9B from $9.6B says the same thing from the balance sheet side.

Assessment: This is the strongest capacity quarter of the coverage and it arrived with the margin expanding rather than contracting, which is the combination we have been waiting four quarters to see. The 40.5% deployment ratio also quietly answers a question from the third quarter of 2025, when the gap between contracted and active power was the bear case: the gap is closing, and it is closing while the numerator grows.

Customer Concentration Returns to the Disclosure Set, and the Answer Is Two-Sided

After three quarters in which the revenue-basis concentration figure was simply absent, it returned, and the headline is excellent: the largest customer fell from 71% of revenue to 36%. We flagged its absence in every one of those three reports, so it is worth saying plainly that the disclosure gap is closed and the direction is the right one.

The honest reading is more mixed. The top three customers now sum to 72% of revenue, which is one percentage point above where the single largest customer sat a year ago. What has happened is not a reduction in aggregate dependence but a redistribution of it across three counterparties instead of one. That is a genuine improvement in risk terms, because three simultaneous counterparty failures are far less likely than one, and because the named large customers now include investment-grade names. It is not the same thing as diversification of the revenue base, and the half-year column shows the same shape, with the disclosed customers alone summing to 63% (the third sits below the 10% disclosure threshold in that period).

The receivables split is the sharper improvement. A year-end position of one customer at 68% of receivables has become two customers at 32% each, which materially reduces the working-capital consequence of any single payment failure.

Assessment: Bear #2 improves but does not resolve. The correct summary is that CoreWeave has traded single-name risk for three-name risk on the same aggregate exposure, while raising the average credit quality of the names. We move the pillar's evidence base from backlog credit mix, which management volunteers, to the revenue table, which it does not, and we will now be tracking the top-three aggregate rather than the top-one figure.

Capital Structure: A Loan That Prices Renewal Risk

Read the ratings column quickly and this looks like a credit deterioration: A- and sub-6% in the first quarter, Ba2/BB+ at SOFR plus 550 in the third. Read the bottom two rows and it is the opposite, and it is the most important financing event of this coverage.

Every prior facility was structured so that the customer contracts securing it ran at least as long as the debt. The lender was therefore never exposed to what a used GPU is worth; it was exposed only to the counterparty paying the contract. DDTL 5.5 breaks that. The underlying contracts average roughly three years against a facility maturing in about five, which means roughly two years of the repayment depends on CoreWeave being able to re-lease or renew that capacity at a price that services the debt. A syndicate led by JPMorgan and MUFG underwrote that, oversubscribed, in what the CFO described as one of the most dislocated credit weeks of the year.

"Perhaps most importantly, the transaction demonstrated the credit market's growing conviction in long-term value of NVIDIA infrastructure running on CoreWeave cloud." — Nitin Agrawal, Chief Financial Officer

The wider spread is the price of that risk transfer, not evidence of a weaker company. A sub-investment-grade rating on a tranche where lenders take residual exposure, sitting alongside an A- rating on a tranche where they do not, is exactly what a functioning market should produce. The spread differential between the two is, for the first time, a market-quoted price for GPU residual value.

The aggregate cost of debt continued to fall regardless. Weighted average cost of debt is down almost 300 basis points over the past year, which management sized at roughly $1.1B of annualised interest savings against the end-of-quarter debt load. Total capital secured to date passed $32B, including an inaugural Eurobond and roughly $18B raised in the quarter across debt, convertibles and equity.

Assessment: This is the strongest single confirmation Bull #3 has received, and it comes from outside the company. Management has told us for four quarters that old GPUs hold value; this quarter a syndicated lending market put its own capital behind that claim and published a price for it. It also has a direct commercial consequence, discussed below: it unlocks the two-to-three-year contract lengths that most enterprise buyers actually want.

Key Topics & Management Commentary

Overall Management Tone: Management was more assured than at any point in this coverage, and for the first time the confidence was retrospective rather than prospective, resting on a delivered margin instead of a promised one. The prepared remarks led with operating leverage rather than with demand, which is a reversal of the past four calls. Analyst questioning was uniformly constructive and notably thin at six participants, so the call produced very little adversarial testing of the second-half ramp, which remains the least examined part of the story.

1. The Margin Inflection Management Had Promised for Three Quarters

Since the fourth quarter of 2025 the company has described a specific shape for 2026: a first-quarter trough, sequential expansion in the second and third quarters, and low double-digit margins by the fourth. The first quarter delivered the trough at 1.0%. This quarter delivered the first step, at 5.0%, above the range it was guided into.

"Perhaps most importantly, Q2 marked the quarter in which we saw margins inflect expanding sequentially as we had discussed over the past several quarters." — Nitin Agrawal, Chief Financial Officer

The mechanism management offers is scale arithmetic rather than cost cutting: each new build is a smaller proportion of a larger installed base, so the fit-out drag dilutes even as its absolute size grows.

"Each new deployment is landing against a much larger installed base than it was even 1 quarter ago. As that base grows, each new build becomes a smaller part of the whole, while contracted revenue from existing deployments remains in place. This is how we are transforming scale into operating leverage. It is why margins expanded in Q2 and why we expect them to continue expanding sequentially during Q3 and Q4." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: The claim is now supported by one observation under adverse conditions, since the quarter carried the largest fit-out drag in company history. One observation is not a curve, and the fourth quarter still requires roughly a tripling from the third-quarter guide. But this was the falsifiable claim, and it did not falsify.

2. Contribution Margins on New Contracts Step Up 5 to 10 Points

The most forward-looking number in the prepared remarks was not in the guidance. Contracts signed during the quarter carry expected contribution margins 5 to 10 percentage points above those signed in recent quarters, which is a statement about the profitability of revenue that has not yet been recognised.

"In Q2, the customer contracts we signed came with contribution margins we expect to be 5 to 10 percentage points above those added in recent quarters." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Management attributed much of the step-up to the Vera Rubin generation, and separately noted that the increase in the value of CoreWeave's output has outpaced the increase in the cost of its inputs, which is the direct answer to the component-inflation concern that dominated the previous call.

Assessment: If this holds, it is more valuable than the July price increase, because contribution margin on signed contracts flows into the 2027 and 2028 income statements automatically as those contracts deliver. It is also unverifiable from outside until it appears, and no dollar quantification was given.

3. A 25% Price Increase, Taken in July and Not Yet in Any Reported Number

List prices rose roughly 25% across SKUs in July, after the quarter closed. The margin expansion reported for the second quarter therefore contains none of it.

"And this operating margin improvement came before our July pricing changes which included an approximately 25% increase across SKUs in response to the current demand environment and the increasing ROI our customers are observing from their investments in the CoreWeave platform as they shift to inference. We are also passing through component price increases." — Nitin Agrawal, Chief Financial Officer

Assessment: A 25% list increase in a take-or-pay business affects new and renewing contracts only, so the revenue effect is slow and the margin effect slower. What it signals is more immediate: a company that is genuinely capacity-constrained can raise price, and one that is not, cannot. Read alongside the sold-out commentary, this is the strongest available evidence that demand exceeds supply at the current price.

4. The Recontracting Economics, Stated Plainly for the First Time

The CFO laid out the full asset-level economics in a way no prior call has: contracted revenue over the initial term repays the asset-level debt and generates the return, so recontracting is upside on an already-complete investment rather than a required assumption.

"So when an initial contract ends, the cluster no longer has any leverage, and we are free to recontract that cloud infrastructure or offer it to the market. We will have generated an attractive return even before the prospect of further monetizing the cloud infrastructure. Every resale or renewal is incremental on top of the returns already earned within the initial term." — Nitin Agrawal, Chief Financial Officer

This reframes the residual-value debate. Our initiation in August 2025 treated residual value as a condition of the investment case, because we could not tell whether the assets would outlive the loans. Management's position is that they do not need to: the loan is retired inside the contract.

Assessment: The framing is only as good as the underwriting behind it, and no contract-level return figures were disclosed to test it. But DDTL 5.5 is external corroboration that this structure is real, since a lender extending past contract maturity is implicitly agreeing that the asset outlives its own leverage.

5. An A100 Contracted to 2029, on a 2020 Architecture

The single most concrete residual-value datapoint of the coverage so far, and the fifth consecutive quarter of favourable evidence on this pillar.

"As an example, we recently signed an A100 contract that extends into 2029 at an attractive price. As a reminder, this SKU was introduced in 2020." — Nitin Agrawal, Chief Financial Officer

The prior evidence ran: a single H100 recontracted within 5% of original price (Q3 2025), portfolio H100 pricing within 10% across the year with A100 pricing up (Q4 2025), and sequential price increases across A100, H100, H200 and L40 (Q1 2026). This quarter adds duration to price. An architecture launched in 2020 under contract into 2029 implies a useful economic life approaching nine years against a depreciation schedule of roughly six.

Assessment: The residual-value question that capped our rating at initiation is now answered as completely as disclosure allows. No price was attached to this particular contract, which is the remaining gap, and "attractive" is management's word rather than a number.

6. Managed Inference Goes From $1M to More Than $100M of Booked ARR in One Quarter

Management described inference growth as constrained only by near-term capacity, and framed the roll-off of older fleet as the natural supply for it. That connects two threads that have been separate until now: what to do with GPUs coming off contract, and how to grow a higher-margin software and services business without diverting current-generation capacity from take-or-pay customers.

Assessment: These are the first quantified figures for the non-core-cloud businesses in this coverage, and roughly $500M of combined ARR against a $12.4–13.2B revenue year is small. The growth rate is the point, and so is the margin mix: management calls these lines margin-accretive, which makes them a lever on the very thing this rating turns on.

7. CoreWeave Omni Signs Its First Deal, and Still Has No Number

Omni deploys the CoreWeave software stack on customers' own infrastructure. We flagged it last quarter as the item most capable of changing the company's margin structure and least described.

"Through CoreWeave Omni, we are seeing significant interest from sovereign, enterprise and cloud customers alike. In the past few weeks, we signed our first deal, which will begin to scale in 2027." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: Progress from concept to signed contract in one quarter, with no size, no pricing model and no pipeline. It remains excluded from guidance and from our model, which is the conservative treatment and also the only possible one.

8. Vera Rubin Brought Up First, and Carrying the Margin Step

CoreWeave was the first cloud provider to bring up and validate NVIDIA's Vera Rubin NVL72, extending a first-to-market record that now spans several generations. More relevant to the rating is that the new generation is where the margin improvement is concentrated, which is unusual: new architectures normally carry their heaviest fit-out and yield costs at launch.

"Vera Rubin is a generation that is seeing the margin expansion right from the start. And so it's really exciting for us. The demand for the Vera Rubin platform is enormous and the pricing power that CoreWeave has been able to garner with its CoreWeave cloud, delivering that infrastructure really bodes well." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: Being first to a generation matters commercially only if it converts to price, and management says it has. This is the mechanism underneath the 5 to 10 point contribution margin claim, which makes the two disclosures mutually corroborating rather than independent.

9. Capital Intensity Rises Again: $35–39B for the Year

The full-year capex guide rose $4B at both ends, the second consecutive increase. Management attributed it to capacity being pulled forward rather than to cost inflation, and paired it with the raised revenue, run-rate and power targets, which is the pairing that was missing in the first quarter when capex rose against a soft revenue guide.

Second-quarter capex of $9.4B on the management basis exceeded the $7–9B guide, again on accelerated deliveries. The cash flow statement shows $6,422M of property and equipment purchases for the quarter, a different basis that should not be compared with the guide.

Assessment: A capex raise accompanied by a revenue raise, a run-rate raise and a power raise is a demand signal. A capex raise on its own is a margin problem. This quarter produced the first kind, which is precisely the distinction the market got wrong in May and right in August.

10. Power Reaches 4.2 Gigawatts Contracted, and Goes International

"Contracted power grew to 3.7 gigawatts in Q2, and since quarter end, we have added roughly 500 megawatts, bringing contracted power to 4.2 gigawatts as of today." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

More than 1 GW is now contracted outside the United States, including 360 MW in Indonesia that comes online in roughly 18 months. Adding the more than 1.5 GW of powered land and executed letters of intent takes the identified pipeline to roughly 6 GW against the 8 GW-by-2030 target, at the midpoint of 2026.

Assessment: The 8 GW target is now roughly three-quarters identified with four and a half years to run, which converts it from ambition to schedule. International expansion introduces a set of risks this coverage has not yet had to price, including currency, sovereign counterparties and longer build timelines.

11. Local Opposition to Data Centers Becomes a Named Risk

The first call of this coverage on which community moratoriums were raised as a constraint. Management's position is that opposition redirects where infrastructure is built rather than reducing how much gets built, and that none of the guided figures depend on sites facing pushback.

"we believe that the certain communities have moved forward with moratoriums. And we feel like moratoriums are -- they're not going to impact the demand for this infrastructure, they are going to impact where this infrastructure gets built." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: The answer is reasonable and unfalsifiable. Siting risk does not show up in a quarter; it shows up as a delay two years later, which is exactly how the third-quarter 2025 powered-shell delay behaved. We note it as an emerging watch item rather than a current impairment, and the 1.5 GW of powered-land optionality is the relevant mitigant.

Guidance & Outlook

Every guided line moved in the shareholder's favour except capex, and capex moved for a reason that was visible in the same sentence: more capacity delivered sooner. That combination is what was absent in May, when a capex increase arrived next to a revenue guide 6.1% below the Street, and it is the substance of the market's very different reaction this time.

Implied fourth quarter. Subtracting reported first-half results and the third-quarter guide from the full-year range gives fourth-quarter revenue of $4.15–5.10B (midpoint $4.62B) and adjusted operating income of $551–801M (midpoint $676M). That is a fourth-quarter adjusted operating margin of roughly 13% to 16%, consistent with the low-teens language, and it requires adjusted operating income to roughly triple from the third-quarter midpoint of $230M.

The second-half requirement, restated. In May the framework needed $919M of adjusted operating income in the second half against $81M implied in the first. The first half delivered $149M, nearly double the plan, and the second half now needs $811M–1,001M (midpoint $906M). The absolute task is essentially unchanged; the base underneath it is 84% larger, and one of the two remaining quarters is now guided with a mechanism that has been observed working once.

The sequential path. Adjusted operating income of $21M, then $128M, then a guided $200–260M, then an implied $551–801M. The first two steps are reported, the third is guided ten weeks out, and the fourth is arithmetic derived from a full-year range. The fourth step is where the risk sits and it is the largest one.

Street at: consensus stood at $3.42B for the third quarter and $12.63B for the full year. The guide midpoints are 3.1% and 1.3% above those figures respectively, which reverses the pattern of the last two quarters, when guidance landed below the Street on both occasions.

Guidance style: across the four quarters in which management has set a quarterly range, revenue has landed at or above it every time. Adjusted operating income has done the same in three of four; the exception is the fourth quarter of 2025, where the implied range of $110–140M produced $88M and the full-year profit guide was missed with it. The pattern is conservative on quarters and, until now, aggressive on years. Raising the annual profit guide on the strength of a single quarter's beat is a mild departure from that pattern, and it slightly increases the risk attached to the fourth quarter.

What the guide does not contain. CoreWeave Omni and the wider cloud-stack licensing remain excluded. The July price increase affects contracts signed after the quarter closed and therefore contributes little to 2026 revenue. Neither is in our model.

Analyst Q&A Highlights

Six participants asked eleven questions in a session that ran unusually short for a company of this size and controversy. The questions clustered almost entirely on the recontracting opportunity and the pricing environment, with no sustained challenge to the fourth-quarter margin step, the capex increase or the leverage.

How Much of the Fleet Comes Up for Renewal, and What It Renews At

The opening question of the call, and the one that most directly tests the pillar this coverage has always turned on. The questioner asked both for customer intent on contract length and for the magnitude of the fleet approaching renewal. The answer gave the first and largely withheld the second, characterising the near-term roll-off as a small part of the fleet without sizing it, while confirming that older-generation prices are at or above levels seen a year ago.

Q: "you did mention the renewal opportunity with shorter-term contracts as some of the older contracts come off expiration to leverage sort of the pricing that we are seeing in the market. Can you just help us think through as you engage in some of that discussion with customers, what you're finding in terms of typical customer intent in terms of contract period? And how much of your installed base of equipment is maybe up for renewal over the next few years, if you can get us -- give us a sense of how to think about the magnitude of that opportunity?"
— Samik Chatterjee, JPMorgan

A: "In terms of the capacity that's coming up for renewal, Samik, is a very limited part of our fleet. And the ASPs on the older generation remain higher or at levels that we've seen about a year ago."
— Nitin Agrawal, Chief Financial Officer

Assessment: The pricing half of the answer is strong and consistent with four prior quarters of evidence. The sizing half was not answered, and the phrase "a very limited part of our fleet" cuts both ways: it caps the near-term recontracting upside as much as it caps the near-term risk. An investor cannot yet model this opportunity, only believe in it.

What Is Actually Driving the Step-Up in Contribution Margins

The most valuable disclosure of the prepared remarks was the claim that newly signed contracts carry contribution margins 5 to 10 percentage points above recent vintages. The follow-up asked how much of that came from shorter contract duration versus competitive differentiation. The response declined to decompose it and attributed the improvement to platform value and to customers who are now monetising their own products and will therefore pay more for capacity.

Q: "So on the 5% to 10% better margin that you're seeing on recent deals that you're signing, can you help unpack the drivers how much is a function of shorter duration deals, versus strong competitive differentiation or other factors? And what are you seeing more broadly just out there in the market as it relates to pricing?"
— Brad Zelnick, Deutsche Bank

A: "So it's a combination of a lot of things, and it's difficult to deconstruct it. The infrastructure that we deliver through the CoreWeave cloud is more valuable to our customers than any other solution that they can encounter. The quality of the platform, the reliability of the infrastructure, the security, the TCO, all of those things contribute to customers coming back to us again and again and expanding their footprint within our cloud and infrastructure. And so there is a piece of it, which is they understand how much more valuable a given piece of infrastructure is delivered through us. The second piece of it is, many of our clients are monetizing their products. And so they are more aggressive about coming in and willing to pay us higher margins because they need access to the compute that will allow them to be successful."
— Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: A genuine non-answer on the decomposition, and the decomposition is what matters. If the step-up comes from shorter duration it is a mix effect that reverses when duration normalises; if it comes from differentiation it is durable. Management was later more specific in attributing much of it to the Vera Rubin generation, which points toward the durable explanation, but the direct question was not addressed.

Whether Local Opposition Threatens the Power Roadmap

The first appearance of siting and permitting risk as a substantive line of questioning in this coverage. The response separated demand from location, arguing that moratoriums redirect construction rather than reduce it, and stated that none of the guided capacity figures depend on contested sites. The supporting arithmetic offered was 4.2 GW contracted plus roughly 1.5 GW of powered land and letters of intent, against the 8 GW 2030 target.

Q: "I think the regulatory backdrop for data centers appears to be increasingly difficult. There have been reports of local opposition to data centers. With this in mind, can you talk about your confidence level in deploying more than 3 gigawatts of active power by the end of next year and kind of your road map to 8 gigawatts by the end of the decade?"
— Irvin Liu, Evercore ISI

A: "And so when we talk through the numbers with you guys, we -- basing our progress on where we are today and what we have guided here, none of those numbers will be impacted by the regulatory pushback as of today. We are comfortable with it."
— Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: The qualifier "as of today" is doing real work in that sentence and was not probed. Siting risk does not impair a quarter; it surfaces as a delay two years out, which is precisely how the powered-shell delay of the third quarter of 2025 behaved. Note also that the question's "more than 3 gigawatts by the end of next year" was the questioner's framing and was neither confirmed nor corrected.

Why 500 Megawatts of New Power Did Not Show Up in Sequential Revenue

The most analytically precise question of the call. Nearly 500 MW was energised while sequential revenue growth was broadly similar to the prior quarter, which appears inconsistent until the timing within the quarter is known. The answer supplied the timing: roughly 300 MW arrived in June, so most of the new capacity contributed cost in the second quarter and will contribute revenue in the third.

Q: "I realize there's likely some rounding here, but you added an impressive 500 megawatts of active power in the quarter. The revenue, if we're looking at the sequential adds is fairly consistent with last quarter. We can hear all the commentary around the uplift that's coming. So maybe help us think through the linearity of capacity added."
— Michael Turrin, Wells Fargo Securities

A: "300 of that alone was in the month of June, which is higher than any amount of power that we've added in any history of prior quarter for CoreWeave. So definitely, that power was back-end loaded in terms of Q2, which you would start seeing kind of come through in Q3 and Q4 in our business."
— Nitin Agrawal, Chief Financial Officer

Assessment: This is the exchange that most supports the third-quarter guide. Capacity energised in June is already built, already contracted and already depreciating, so the revenue attached to it is a scheduling matter rather than a selling matter. It also makes the second-quarter margin beat more impressive, since that beat absorbed the full cost of the June capacity with none of its revenue.

What the New Financing Structure Changes About Contract Length

A question connecting the new loan facility to commercial strategy rather than to the balance sheet. The response was the clearest statement of the quarter on why the financing matters: enterprise buyers want two-to-three-year terms, and until lenders would finance contracts shorter than the debt, that segment was effectively closed to CoreWeave.

Q: "How does the success of the DDTL 5.5 change the way you view on kind of target durations on a go-forward basis? Does -- is that an avenue you want to use more to maybe extract more margin in shorter duration contracts, given you know the useful life is there?"
— Brett Knoblauch, Cantor Fitzgerald

A: "When you're thinking about enterprise, enterprise tends to want to enter into contracts that are not 5 years in length. They tend to think in shorter cycles than that. And so by enabling the financing market to support the contracts in 5.5, we're able to go ahead and diversify our terms so that we're able to support additional contracts. It opens up whole new markets for us. These clients want to buy compute for 2 years or 3 years, and that's not a market that was easily accessible to us until we were able to bring DDTL 5.5 to market."
— Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: The most consequential answer on the call, and it reframes a financing event as a market-access event. It also implies a structural shift in the revenue mix toward shorter, higher-priced contracts, which raises reported growth and margin while shortening the average duration of the backlog. That is a trade-off, and nobody asked about the second half of it.

How Fleet Coming Off Contract Gets Allocated

A question about the operational decision rule for capacity rolling off its initial term: recontract, sell on spot, or feed the managed inference platform. The answer described an explicit optionality between term contracts and shorter-duration sales, with the decision turning on whether the economics justify locking in duration.

Q: "As you look at the GPU fleet that's maybe rolling off contract, can you talk about the cadence of how you guys look to either recontract that or kind of put it on spot or in your inference products? And how far in advance of the roll off of those contracts, would you look to kind of make that decision?"
— Brett Knoblauch, Cantor Fitzgerald

A: "Some of the infrastructure that comes off-line, we go ahead and we place back into a term contract. If we think the economics warrant putting it in. And the economics include both the term that we're able to garner as we think about the long-term stability of the company and the long-term obligations that we need to support as we continue to build and scale the company. But we also do recognize that in the short term, there is an opportunity to sell on shorter-term contracts and extract additional margin on this infrastructure as the world tries to catch up with what is a systemic disequilibrium that has really existed for several years now and will continue to exist for the foreseeable future."
— Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: The answer describes a decision framework without a single number attached: no cadence, no lead time, no clearing price. That is the persistent shape of this disclosure set, where the qualitative case for residual value is unusually strong and the quantitative case rests entirely on management characterisations. The lenders behind the new facility presumably saw contract-level data; public shareholders have not.

The Exit Run Rate Question That Was Asked About 2027 and Answered About 2026

A questioner asked whether, given upward pricing pressure from several directions, the $18–19B annualised run rate should still be regarded as the exit target for 2027. The answer restated the newly raised 2026 exit figure and did not address 2027 at all. The premise of the question was itself confused, since $18–19B was always the 2026 figure, but the exchange left the exit-2027 framework of more than $30B unmentioned on the entire call.

Q: "it sounds like there's upward pressure to pricing across multiple vectors, including the higher value you provide to your customers, the pass-through of higher component costs and the recontracting opportunity coming up. So with that in mind, should we still think of kind of the $18 billion to $19 billion in ARR as kind of the exit target for 2027?"
— Irvin Liu, Evercore ISI

A: "So we increased the exit ARR number that we provided in guidance to you folks right now at $18.5 billion to $19.5 billion for 2026, so that is baked in our guidance that we provided to you."
— Nitin Agrawal, Chief Financial Officer

Assessment: A muddled question produced a technically correct answer that left the more important figure untouched. The exit-2027 target of more than $30B, disclosed in May as more than 75% contracted, is the number our valuation uses as a terminal sanity check, and it went a full quarter without being reaffirmed or revised. That is not evidence it has changed. It is a gap in the record we would prefer not to have.

What They're NOT Saying

1. The size of the fleet coming off contract. Described as "a very limited part of our fleet" and left there. The recontracting opportunity is now central to the equity story, to the new financing structure and to the inference roadmap, and it has no denominator.
2. Any clearing price for spot, Flex or recontracted capacity. Third consecutive quarter without one. Prices are said to be at, above or well above prior levels, always relatively and never absolutely.
3. Utilization. Six quarters public, still no figure. With inference now a stated priority and capacity described as effectively sold out, a utilization rate would distinguish sold from used.
4. The exit-2027 run rate. Not reaffirmed, not revised, not mentioned. The nearest question was answered about 2026.
5. Free cash flow. Operating cash flow of $679M against $6,422M of property and equipment purchases in the quarter, and neither figure was set against the other by management or by any questioner.
6. Why operating cash flow fell 77% sequentially. From $2,984M to $679M, driven by a $656M outflow in payables and accrued expenses and a $435M build in receivables. On a call about operating leverage, the cash conversion went unmentioned.
7. Depreciation life. D&A of $1,393M is 54% of revenue. On a call presenting a six-year-old architecture contracted three further years forward, the depreciation schedule that assumes six years was not revisited.
8. What the fourth quarter requires. Management guided the third quarter explicitly and the fourth only through a full-year range that implies roughly $676M at the midpoint, nearly triple the third-quarter guide. Neither management nor any questioner stated the implied figure aloud.
9. The duration trade-off in the new financing. Shorter contracts at higher prices shorten the weighted average life of the backlog, which was extended to roughly five years only two quarters ago and presented then as a strength.
10. Sizing for CoreWeave Omni. First deal signed, scaling in 2027, and still no revenue model, pricing structure or pipeline. Second consecutive quarter on the "watch" list without a number.

Market Reaction

• Pre-print setup: close of $90.32 on August 11, up 26.1% year to date from $71.61 at the end of 2025, up 1.6% over the trailing 30 days from $88.88, and down 35.4% over the trailing twelve months from $139.78. The stock entered the print 39% below its 52-week closing high of $148.75 and 20.9% below the $114.15 at which we downgraded it in May.
• After-hours move: shares rose roughly 14% to 17% in extended trading on August 11, with prints reported around $104.30 to $104.50.
• Next-day session: opened at $108.70, a 20.3% gap up; traded a $105.44 to $111.83 range; closed at $107.73, up 19.3% and $17.41 on the day.
• Volume: 87.3M shares against a 30-day average of 26.5M, a 3.3 times multiple and the highest reaction-day volume multiple of the five prints we have covered.
• Benchmark: the S&P 500 rose 0.3% on the session and was up 12.9% year to date entering the print, so the move was almost entirely company-specific.

This is the first positive reaction day in five quarters of covering this stock. The previous four prints produced closes of negative 20.8%, negative 16.3%, negative 18.5% and negative 11.4%, every one of them on a quarter that beat consensus revenue. That pattern was the single most useful fact about CoreWeave as a security, and it broke here.

The proximate cause was the margin line, not the revenue line. A 0.6% revenue beat does not move a stock 19%. Adjusted operating income of $128M against a $30–90M guide, alongside a raised full-year profit guide, resolved the question the previous four sell-offs had all been about in one form or another. The market had been discounting the possibility that this business would never convert contracted growth into operating profit at all.

The setup did most of the work. Entering a print down 35% over twelve months, flat over thirty days and 21% below its level three months earlier, with the index up 12.9% year to date, is the mirror image of May, when the stock arrived up 80% year to date and a guide 6.1% below consensus was enough to take 11% off it. The same disclosures land very differently against those two setups.

The regular session extended the after-hours move rather than fading it. Reported after-hours gains of 14% to 17% became a 19.3% close. In two of the four prior quarters the reaction roughly doubled overnight in the other direction as investors worked through arithmetic the headline did not contain. That the extension happened upward this time, on 3.3 times normal volume, suggests the buying came from investors who read the full release rather than from a headline reflex.

Our read is that the reaction was proportionate but not complete. The margin inflection genuinely deserved a re-rating, and at $107.73 the stock has recovered only to 5.6% below where we downgraded it in May, while the index has gained roughly 5% over that stretch. What the move has not yet priced is the new financing structure, which received almost no attention in the coverage of the print and is, in our view, the more durable of the two developments.

Street Perspective

Debate: Does One Quarter of Margin Expansion Validate the Whole Second-Half Framework?

Bull view: The bull case on the Street holds that the hard part is done. The mechanism management described in May was tested in the worst possible quarter, one carrying the largest fit-out drag in company history, and it beat its own guide by 42%. Operating expenses grew 5.8 points slower than revenue, roughly 300 MW energised in June carries into the third quarter as almost pure margin, and the company raised the full-year profit guide rather than merely reaffirming it.

Bear view: The bear camp notes that the fourth quarter still needs roughly $676M of adjusted operating income against a third quarter guided to $230M, a near-tripling in one step, and that 86% of the year's profit still sits in the second half. It also notes that this company missed exactly this kind of implied fourth-quarter figure in 2025, when $110–140M produced $88M, and that the miss came for deployment-timing reasons that have not disappeared.

Our take: The bulls have the better of it on the mechanism and the bears on the magnitude. What changed is the category of the risk: in May the question was whether operating leverage existed at all, and now it is whether it scales fast enough in one specific quarter. That is a materially smaller question, and it is why our rating changes. We model the second half modestly below the guide midpoint and would treat a third quarter at the low end of $200–260M as a warning rather than a failure.

Debate: Is the New Loan a Credit Downgrade or a Residual-Value Endorsement?

Bull view: A growing view holds that DDTL 5.5 is the most important thing that happened in the quarter. A syndicated market agreed to lend for five years against contracts averaging three, at Ba2/BB+, oversubscribed, during a dislocated credit week. That is external capital underwriting the exact question that has capped this stock since its IPO, and it opens the enterprise segment that wants two-to-three-year terms at higher prices.

Bear view: The skeptical framing is simpler and reads the ratings table straight: the company financed at A- and under 6% in the first quarter and at sub-investment grade and SOFR plus 550 in the third, on a balance sheet that added $10.2B of debt in three months. Whatever the structure, the cost of the marginal dollar went up.

Our take: The bears are reading a price difference as a quality difference. The two facilities carry different risks, and a wider spread on the tranche where lenders take residual exposure is what a correctly functioning market produces. The evidence that the aggregate cost is falling rather than rising is that weighted average cost of debt declined almost 300 basis points over the past year. The bulls are right about the significance and, we think, early: this development got almost no attention in the coverage of the print.

Debate: Does the Improving Margin Outrun the Compounding Balance Sheet?

Bull view: The bull framing is that the inflection has begun and everything compounds from here: interest coverage improved for the first time, the exit run rate was raised, contribution margins on new business step up 5 to 10 points, list prices rose 25%, and the roughly $500M of combined managed inference and platform ARR is growing from nothing at very high margins.

Bear view: The bear framing is that the balance sheet compounds faster than the margin. Net debt rose 30.8% sequentially to $28.1B while adjusted operating income rose $107M and interest expense rose $104M, so the operating gain was almost exactly consumed. The full-year capex guide rose another $4B, and a single quarter of guided third-quarter interest expense, roughly $900M at the midpoint, is about 85% of the entire year's adjusted operating income guide midpoint.

Our take: That last comparison is the most arresting number in this report and it is correct. The bear case is arithmetically sound about 2026 and the bull case is about 2027 and beyond, which is when the assets built in this capex cycle produce revenue against debt that has stopped growing at this rate. Our valuation resolves it by using the exit-2026 run rate against an explicit exit-2026 net debt assumption, so that both effects are counted in the same number. On that basis the equity is worth more than it was in May, and it is priced lower.

Model Update & Valuation Framework

Valuation base unchanged in method. We continue to value on management's exit-2026 annualised run-rate revenue, because reported annual revenue systematically understates earning power when capacity energises continuously through the year, and we continue to state the net debt assumption explicitly because it does most of the work.

At the August 12 close of $107.73 and 551M shares, equity value is approximately $59.4B. Adding total debt of $35.07B and subtracting cash, restricted cash and marketable securities of $6.92B gives an enterprise value of roughly $87.5B, which is 6.8 times the midpoint of the FY2026 revenue guide, 4.6 times the midpoint of the exit-2026 run rate, and 2.9 times the exit-2027 target.

Why the multiple moves up a quarter-turn. Since initiation we have applied a discount for the possibility that operating leverage in this business was theoretical. That discount is smaller now. It is not zero, because one observation is not a curve and the fourth quarter is the larger test, so the uplift is a quarter-turn rather than a full one.

Twelve-month price target: $124 (base range $107–141). The target barely moves, and the composition of that non-move is the most honest summary of the quarter. The raised run rate and the higher multiple add roughly $14 per share; the $4B increase in the exit net debt assumption removes about $7.50, and 24M additional shares remove about $5.50. Growth created enterprise value and the balance sheet claimed most of it, which is the mechanism this thesis has described since initiation. A $124 target against $107.73 implies roughly 15% over twelve months.

What differentiates the scenarios. The bear case is not a demand failure. It is a fourth quarter in which the margin step arrives late, the exit run rate lands near $17B, the capex programme carries net debt to $44B, and the market applies an infrastructure multiple to an equity standing behind it. The bull case requires no acceleration beyond guidance, only that the run rate reaches the top of the guided range and that the demonstrated operating leverage plus the new financing structure earn a growth multiple. The tails remain unusually wide and turn, as at every prior report, on whether contracted revenue outlives the debt that funded the assets producing it. This quarter a lending syndicate took a position on that question with its own money.

Thesis Scorecard Post-Earnings

Grading Last Quarter's Commitments

Overall: Six of eleven commitments met or beaten, three partial or unaddressed, one not delivered, and one exceeded in a way that cuts both ways, since the capex overrun came from pulling deliveries forward. Most importantly, the item we singled out in May as the make-or-break was not merely met but beaten, in the quarter carrying the heaviest fit-out drag in the company's history. The thesis is stronger than it was three months ago on every pillar except leverage, where it is worse in level and better in direction.

Action: Upgrade to Outperform from Hold with a $124 twelve-month target. In May we set four conditions that would take us back to Outperform, and two fired cleanly. The first was a second-quarter adjusted operating margin that expanded as guided, "particularly if paired with a raise to the full-year profit guide": the margin beat the guide's ceiling and the full-year profit guide was raised. The second was a meaningful pullback with the trajectory intact: the stock fell 20.9% before the print. A third fired partially, with managed inference and platform ARR quantified for the first time even as Omni itself remains unsized. We upgrade on the triggers we set rather than on a re-derived view, which is the same discipline that governed the November 2025 upgrade.

Why conviction is moderate rather than high. The fourth quarter carries roughly $676M of implied adjusted operating income against a third quarter guided to $230M, and this company missed exactly that kind of implied fourth-quarter figure a year ago. Net debt is compounding at 30% a quarter and the capex guide rose another $4B. Position sizing should reflect a distribution running from $28 to $230 per share, which is not a normal equity distribution and should be treated as an option on a specific operational outcome.

What would take us to a higher conviction Outperform: a third quarter at or above the high end of $200–260M, which would make the fourth-quarter step arithmetic rather than aspirational; disclosure of the size of the fleet approaching recontracting, which would let the recontracting upside be modelled instead of believed; a spot or Flex clearing price; or a quantified CoreWeave Omni contribution, which remains outside guidance and outside our model.

What would take us back to Hold or lower: a third quarter below $200M of adjusted operating income, which would break the sequential framework in the first quarter after we upgraded on it; a full-year profit guide cut, which after this quarter's raise would carry more signal than an ordinary reduction; the first sequential decline in prior-generation GPU pricing; a counterparty credit event among the top three customers, which now represent 72% of revenue between them; or a financing that widens spreads on the tranches where lenders do not take residual risk, which would indicate the credit market has repriced the collateral rather than the structure.