聚焦利润率与回报率：CRWV专题

CRWV.0Q United States Software Putting margins and returns under themicroscope: CRWV Edition Research Analyst+1-212-250-8563 Demand for Al infrastructure has driven a real inflection in public cloud revenueand backlog growth, enabling long-standing leaders to reaccelerate off anincreasingly largebaseand creating an openingfornew Neocloudsto enterwhathad been largely an oligopoly market at a significant scale.At the same time,however,margins acrosstheindustryhave comeunderincreased pressurefromrapid mix shift to what today is generally lower-margin, less IP-rich Alinfrastructure services and the upfront costs required to massively scale newcapacity in an attempt to meet demand. This is an issue we see managementteams across the space consistently talking about and working through, and adifficult oneforthe investment communityto wrestlewithas wetrytodissecttheeconomicsoftheseincreasinglylargeAlbusinesseswithinthebroaderpubliccloud market.To put it bluntly,despite allthe important questions we have aboutthe long-term competitive dynamics and structure of the industry,we often stillfind ourselves in conversations about whether contracts already generatingrevenue and/orbooked in backlog actually havepositive economic returns as abasic starting point. Research Analyst+1-212-250-6775 ResearchAssociate+1-415-2622041 ResearchAssociate+1-212-250-1203 Weareplantingourflagandaffirmingthatcurrent Al businesses acrossthecloudproviders we cover are indeed value-creative and, based on current economics inthemarket,constituteasustainablecommercialfoundation.Ourviewisbasedona detailed analysis of public filings, financial statements, presentations andcommentaryfromcompaniesacrosstheindustrythatleadustobelievethatthosewith more bearish views may be under-appreciating the impact of upfront coststo scale multi-year revenue-generating cloud capacity,the significant OpExleverage inthese businessesand thenatural risk mitigation embedded in manylong-term contracts. In what we intended will be the first in a series of reportsCoreWeave and its business model.Our conclusion isthat CoreWeave's activecontracts,and to the extent new contracts being added are underwritten to thesame standard asmanagement says,deliver very attractive economic returns toinvestorswhilelimitingrisksovertheirmulti-yeartermandbeforeanypotentialrecontracting upside thereafter. This analysis affirms mgmt's view of mid-20scontract-level contribution margins and reinforces our favorable view ofCoreWeave'sabilitytocapitalizeonthecurrentdemandforAlinfrastructurethatisprovidinga pathwayto build amature, scaled Al cloud platform.WemaintainourBuyratingand$135targetprice. Perhaps more than any company we cover,profitability remains acutely the number one area of pushback against accelerating revenue and capacityadditions and backlog that has swelled to near $100bn. Indeed, while there is littleoperating margins into single digits has led to growing questions around howmuchvalue surgingtop-linegrowth reallygenerates.CoreWeave's managementhas repeatedly highlighted the upfront cost headwind from scaling the businessahead ofwhatitassertsare very stronguniteconomics,but admittedlythisishardto see from public financial statements during the current period of heavy initialinvestment. Our detailed analysis in this report helps shed more light on the topicas we specifically:(1)examine CoreWeave's cost accounting,margins and howthey compare to peers; (2) attempt to map out the impact of scaling costs onhistorical margins; (3) overlay these costs with future active power and ARRguidance to frame the trajectory ahead; and (4) detail the underlying uniteconomics of CoreWeave's unique long-term contractstructure.Keytakeawaysareasfollows. Our work validates mid-2Os contribution profit margins for stabilizedcustomer contracts. Leveraging company reports and presentations, wecalculate and add back pre-revenue scaling costs and corporateoverheadtoarriveata~25%contract-level contributionmarginforCY25.To be clear, contribution margin is not a direct substitute for gross oroperating margin but being able toback-test and validate what we hearfrommanagement against actual results increases our confidence inunderlying assumptions. We also understand R&D, arguably a morescalable corporateexpense,isbeingallocated to contract-levelprofitability at a contract level could be understated. Looking ahead toCY26 and adding back our estimates for scaling costs and corporateOpEx to guidance suggests contribution margin just below the mid-20srange at 22% and an area for further diligence. As a high-level rule ofthumb for investors, our analysis suggests $0.7-0.8m of scaling costs inthe 1-2 months prior to the start of revenue generation per Mw ofcapacity currently. ARR & active power guidance support rebounding FY NGOM starting inCY27 as the impact of pre-revenue scaling costs declines and thepoised to sustain rapid, well above overall public cloud market growth.Ourwork and understanding of c