CoreWeave, Inc. (CRWV) Report Interpretation

Overview

Bernstein acknowledges that CoreWeave's 2Q26 report was its strongest execution result to date, with better revenue, margins, backlog conversion and power deployment. The institution raises its price target but maintains Underperform because the improved near-term outlook is accompanied by even faster CapEx growth, high financing costs and unresolved competitive risks beyond 2027.

Core views

CoreWeave's 2Q26 results provided the clearest evidence so far that the company can convert contracted AI-compute capacity into operating revenue. Revenue was USD 2.575B, versus Bernstein's USD 2.518B estimate and USD 2.563B consensus, rising 112.3% year on year and 24% sequentially. Backlog reached USD 104B, up 46% year on year, before more than USD 25B of net new commitments signed early in 3Q; including those commitments, backlog was approximately USD 129B. More than half of reported backlog was attached to contracts where delivery had begun, and management expected that proportion to exceed two-thirds by year-end. Bernstein therefore became marginally more confident that CoreWeave can meet its 2026 numbers, while stressing that continued execution is not assured. The quarter also showed the operating-leverage inflection management had signaled in 1Q. Adjusted EBITDA rose 100.5% year on year and roughly 30% sequentially to USD 1.510B, with margin increasing to 58.6% from approximately 56% in 1Q. Adjusted operating income rose to USD 128M and a 5.0% margin from USD 21M and 0.5% in 1Q, exceeding the company's prior USD 30M–USD 90M guidance range and consensus of USD 67M. Management attributed the improvement to a larger installed base, higher utilization and new contracts carrying contribution margins five to ten percentage points above those signed in recent quarters. CoreWeave also raised prices across SKUs by roughly 25% in July, so the 2Q margin improvement preceded any benefit from those increases. Physical capacity expanded rapidly. CoreWeave added nearly 500 MW of active power during 2Q and ended the quarter at 1.5 GW. Management raised its year-end active-power target to more than 1.85 GW from more than 1.7 GW. Contracted power reached 3.7 GW at quarter-end and approximately 4.2 GW by the earnings call, with another 1.5-plus GW potentially available through powered land, site-expansion options and executed letters of intent. These resources support management's goal of 8 GW of active power by 2030, but Bernstein argues that they also define the execution challenge: power, land, GPUs, networking, cooling and financing must be converted into operating clusters without repeating the pre-revenue cost drag that hurt margins earlier in the year. Managed inference is a possible route to broader and more durable demand. Booked annual recurring revenue for the offering grew from roughly USD 1M at launch to more than USD 100M within several months, and management targets at least USD 250M exiting 2026. Token-based and shorter-duration consumption could help place GPUs as long-term contracts expire and attract enterprises unwilling to sign conventional five-year take-or-pay agreements. Bernstein nevertheless views the business as early, capacity constrained and operationally more complex than CoreWeave's contracted-cloud model. Its margins, retention, utilization stability and additional capital requirements remain unproven, while the enterprise and storage examples cited by management are still small. Capital intensity remains the central counterweight to the operating improvement. CapEx was USD 9.352B in 2Q, above the high end of guidance, while construction in progress rose sequentially to USD 11.9B from USD 9.6B. Management lifted 2026 CapEx guidance to USD 35B–USD 39B from USD 31B–USD 35B and guided 3Q CapEx to USD 11.5B–USD 13.5B. The GAAP net loss improved to USD 626M from USD 740M in 1Q but was more than twice the prior-year loss, while net interest expense reached USD 640M. Bernstein expects 2026 interest expense of USD 3.3B against USD 7.4B of adjusted EBITDA. CoreWeave raised roughly USD 18B through debt, convertible securities and equity during 2Q and said it had secured more than USD 32B of debt and equity capital to date. Although its weighted-average debt cost has fallen by almost 300 basis points over the past year, Bernstein believes debt service and funding requirements are becoming increasingly consequential. Near-term guidance was stronger. CoreWeave guided 3Q revenue to USD 3.45B–USD 3.60B and adjusted operating income to USD 200M–USD 260M, implying continued sequential margin expansion and a low-teens adjusted operating margin by 4Q. Full-year revenue guidance increased to USD 12.4B–USD 13.2B from USD 12B–USD 13B, and the 2026 exit annualized revenue run-rate target rose to USD 18.5B–USD 19.5B. The narrative states full-year adjusted operating income guidance of USD 960M–USD 1.15B, while the accompanying guidance table reports a USD 1.23B midpoint. Bernstein expects the stronger guidance to drive upward near-term estimate revisions, although its own table still shows adjusted EPS estimates of negative USD 4.56 for 2026 and negative USD 4.77 for 2027, versus prior estimates of negative USD 4.20 and negative USD 2.41. Bernstein's Underperform thesis is primarily about what happens after the current capacity shortage. It does not expect the market to turn against CoreWeave immediately and anticipates another USD 45B of deals through the end of 2027. Its 2026 expectations are broadly aligned with consensus, but its forecasts begin to diverge in 2027 and its 2028 revenue estimate is approximately 15% below consensus, with the gap widening later. The institution expects data-center supply to become more readily available from 2028 and questions whether CoreWeave can continue signing very large hyperscale contracts under those conditions. It also doubts that CoreWeave's software differentiation can withstand hyperscalers' scale and spending power. Meta illustrates a customer that could become a competitor, while SpaceX is cited as a new entrant with major scale and cost advantages. Bernstein raises its price target to USD 74 from USD 67 but retains Underperform. The target uses a 25.5x forward EV/adjusted EBIT multiple, reduced from 28.4x, applied to USD 5.2B of NTM+1 adjusted operating income. The report gives an enterprise value of USD 82.338B. For context, CoreWeave's observed weekly EV/NTM adjusted EBIT multiple had ranged from 23.1x to 120.3x and averaged 52.1x, versus Bernstein's 25.5x valuation multiple. The higher target therefore reflects revised operating expectations despite a lower applied multiple; at the USD 90.32 closing price on 11 August 2026, the target implies 18% downside.

Analysis framework

Bernstein compares 2Q results with its estimates, consensus and prior guidance; tracks backlog conversion, power deployment and product development; tests the resulting operating leverage against CapEx, losses and financing costs; then contrasts its multi-year revenue forecast with consensus under an assumed easing of data-center supply. It converts the resulting earnings forecast into a price target using a forward EV/adjusted EBIT multiple.

Methodology notes

• Valuation methods

  Forward EV/Adjusted EBIT valuation

  Bernstein calculates enterprise value by applying a 25.5x multiple to USD 5.2B of NTM+1 adjusted operating income, then derives a USD 74 equity price target.

• Corporate Fundamentals and FinanceOperating and Financial Leverage Analysis

  Operating-leverage and financing-burden analysis

  The report evaluates whether deployment scale and utilization are lifting operating margins faster than CapEx, interest expense and debt-service requirements are growing.

• Industry AnalysisSupply-demand framework

  Data-center capacity supply and demand

  Bernstein attributes CoreWeave's current pricing power and contract growth partly to constrained data-center and power supply, then models weaker durability as industry capacity becomes more available from 2028.

• Event-Driven and Behavioral FinanceExpectation Gap and Expectation Management

  Results and guidance versus expectations

  The report measures quarterly results against Bernstein estimates, consensus and prior guidance, then contrasts Bernstein's longer-term revenue forecasts with the market's expectations.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• CoreWeave (CRWV.US)

  Primary subject and near-term beneficiary of constrained AI data-center capacity, but Bernstein expects greater pressure as supply expands.

  Strengths

  Large backlog, stronger deployment execution, improving utilization, expanding power capacity, better contract contribution margins and early managed-inference traction.

  Weaknesses

  Very high CapEx, substantial losses and interest costs, dependence on continued financing, customer concentration and unproven long-term pricing and GPU residual-value assumptions.

  Comparison

  Bernstein argues that hyperscalers possess greater scale and spending capacity; NBS, described as CoreWeave's closest comparable, has negative operating income.

  Risks

  Execution slippage, higher capital requirements, financing costs, weaker contract economics, declining pricing power and intensified hyperscaler competition.

• Meta Platforms (META.US)

  Cited as an example of a CoreWeave customer threatening to become a competitor.

  Strengths

  Hyperscaler scale and ability to invest heavily.

  Comparison

  Its potential entry illustrates Bernstein's concern that hyperscalers may compete directly with neoclouds for enterprise demand.

• SpaceX

  Cited as a new entrant that could increase competition for neocloud providers.

  Strengths

  Large scale and cost advantages.

  Comparison

  Presented as a competitively advantaged entrant relative to CoreWeave.

• NBS

  Identified as CoreWeave's closest comparable.

  Weaknesses

  Negative operating income.

  Comparison

  Used as the nearest neocloud valuation comparable, although its negative operating income limits direct multiple comparison.

Key data

• 2Q26 revenueUSD 2.575BUp 112.3% year on year and 24% sequentially; versus USD 2.518B Bernstein estimate and USD 2.563B consensus.
• Reported backlogUSD 104BUp 46% year on year; approximately USD 129B including more than USD 25B signed early in 3Q.
• Backlog under deliveryMore than halfManagement expects the proportion to exceed two-thirds by year-end.
• Adjusted EBITDAUSD 1.510BUp 100.5% year on year and approximately 30% sequentially, with a 58.6% margin.
• Adjusted operating incomeUSD 128MA 5.0% margin, versus USD 21M and 0.5% in 1Q and above prior guidance of USD 30M–USD 90M.
• July SKU price increasesApproximately 25%The reported 2Q margin improvement preceded these increases.
• Active power1.5 GWNearly 500 MW was added in 2Q; year-end target increased to more than 1.85 GW.
• Contracted and potential powerApproximately 4.2 GW contracted plus more than 1.5 GW potentialContracted power was 3.7 GW at quarter-end and approximately 4.2 GW by the call.
• Managed-inference booked ARRMore than USD 100MUp from roughly USD 1M at launch, with a target of at least USD 250M exiting 2026.
• 2Q26 CapExUSD 9.352BAbove the high end of guidance; construction in progress rose to USD 11.9B from USD 9.6B sequentially.
• 2026 CapEx guidanceUSD 35B–USD 39BRaised from USD 31B–USD 35B; 3Q guidance is USD 11.5B–USD 13.5B.
• GAAP net loss and interest expenseUSD 626M net loss; USD 640M net interest expenseThe net loss improved from USD 740M in 1Q but was more than double the prior-year loss.
• Capital securedMore than USD 32BCoreWeave raised roughly USD 18B during 2Q; weighted-average debt cost declined by almost 300 basis points over the past year.
• 3Q26 guidanceRevenue USD 3.45B–USD 3.60B; adjusted operating income USD 200M–USD 260MManagement expects continued sequential margin expansion.
• 2026 revenue guidanceUSD 12.4B–USD 13.2BRaised from USD 12B–USD 13B; exit annualized run-rate target increased to USD 18.5B–USD 19.5B.
• Long-term forecast gapApproximately 15% below consensus revenue in 2028Bernstein is broadly aligned with consensus for 2026 and begins to diverge in 2027.
• Valuation basis25.5x NTM+1 adjusted operating income of USD 5.2BThe multiple was reduced from 28.4x and supports the USD 74 price target.

Impact & implications

The quarter improves confidence in CoreWeave's near-term ability to deploy capacity, convert backlog and expand operating margins, and it supports higher near-term estimates. Bernstein argues, however, that the improvement does not settle whether returns will remain attractive after data-center supply eases, hyperscalers compete more directly and debt service rises alongside continued infrastructure investment.

Risks

• CoreWeave may fail to convert contracted power, sites, equipment and financing into operating clusters on schedule, recreating pre-revenue cost drag.
• Pricing durability, long-term contract economics and residual GPU values may weaken as data-center capacity becomes more available.
• Rising CapEx, interest expense and debt service could outweigh operating-margin improvement.
• Hyperscalers and new entrants may use their scale and cost advantages to compete directly for enterprise AI workloads.
• Managed inference remains early, with unproven margins, customer retention, utilization stability and capital requirements.
• Delays in hyperscaler capacity coming online could preserve CoreWeave's value proposition and represent an upside risk to Bernstein's Underperform thesis.
• A stronger specialized software platform could sustain demand for CoreWeave-operated data centers and challenge Bernstein's thesis.
• Continued shortages of power for data centers could sustain CoreWeave's pricing power longer than Bernstein expects.

What to watch

• Whether the share of backlog already under delivery exceeds two-thirds by year-end.
• Progress toward more than 1.85 GW of active power by year-end and 8 GW by 2030.
• Whether CoreWeave can convert approximately 4.2 GW of contracted power into operating capacity without renewed margin drag.
• Delivery against 3Q26 revenue, adjusted operating income and CapEx guidance.
• Progress toward at least USD 250M of managed-inference booked ARR exiting 2026, together with its margins and utilization.
• Interest expense, debt costs and funding requirements as 2026 CapEx rises to USD 35B–USD 39B.
• Whether CoreWeave signs the incremental USD 45B of deals Bernstein anticipates through the end of 2027.
• The pace of hyperscaler capacity additions and direct competition as data-center supply is expected to ease from 2028.