美银 软件趋势：首批软件新品发布后的投资者反馈

Investorfeedbackfollowinginitialsoftwarelaunches 31March2026 Summary of recent investorfeedback United StatesSoftware initiation of Nebius,we engaged in extensive conversations with investors, includinglong-only funds, hedge funds, and credit-based investors: In this report we summarizethekey investor feedback, highlight the mostfrequent debates and points of pushback,and outlinehow investors evaluatetherelative positioning,risk profiles,and long-termprospects of thefour companies, Tal LiantResearch Analyst1646 855 5107tal.lianipbota.com Where we are in the workload cycle Kevin Nlederpruem1646 855-1540kevinniederpruen@botfacom A key thematic debate surfaced around the industry's transition from chatbot-centricLLM-driven workloads toward agentic solutions, where the bulk of long-term economicvalue is expected to accrue. LLM chatbots are already on a path toward Eden VacnichResearchAnalystBofAS16468551971eden.vachichaobofa.com commoditization,while agent-based workflows aretightly embedded into the enterpriseincreasinglyfocused onhowmarket sharedynamics evolve asadoptionofagentictoolsaccelerates.Forexample,investors expressedskepticismaround Microsoft'spositioning.noting slower-than-expected Copilot adoption.At the same time,we believe Microsoft'sinfrastructure could providemeaningful leverage as agentic workloadsbegin attachingto existing applications.We discuss the Microsoft debates in the following sections. LLM-large language model Neocloud valuationandvalueovertime Two themes consistently emerged around the Neoclouds: how to value them and canthey sustain the value proposition in the long run? EViSales is an imperfect frameworkforbusinesses that are both high-growth and highly capital-intensive, but limiteddisclosure around long-term margins leaves only a few altermatives in the near term.Webelieve these companies should ultimately be assessed on EV/EBIT or EV/EBITDA, andplan to address this in the future.Additionally.investors questioned the long-run viabilityof the Neocloud model beyond the current supply-constrained environment.If thesehyperscalers build sufficient in-house infrastructure,Nebius has a clear strategy with itsearlypush into enterprise-focused softwareand tooling.while CoreWeave needs tosharpen its message on how it could sustain its technical leadership in the future,andfocus on execution near term. Overall, long-term winners need to pair raw capacity withdifferentiated software, enterpriseadoption,and value-add services that extend beyondinfrastructurealone. Microsoft Competitivetractionand Copilotadoption Our discussion wascenteredaround 3key debates: 1）TheriskofMicrosoft's shareloss A recurring theme was the skepticism around Microsoft's Al leadership and the durabilityscaling of ChatGPT,Claude,Gemini,and a growingfield ofopen-sourcemodels,whichhas translated to the risk of diminishing value for Microsoft's Copilot.This skepticismhasbeen reinforced bymodest Copilotpenetration of-15mnpaid seats, or-3.5%ofpaid M365 commercial seats,an underwhelming number given Microsoft's extensiveinstalledbase. We believe this framing misses Microsoft's core strategy of focusing on tightly bundledfunctionality rather than best-of-breed solutions.In cybersecurity for example,Microsoftfastest-growing cybersecurity vendor. We see similar dynamics emerging in Al, whereMicrosoft'scontrolof enterpriseworkflows likeM365,Dynamics,Security.identityanddata gravity,creates meaningful advantages in agentic Al adoption.In this construct,Copilot will be a bundling driver of ARPU expansion, while the longer-term upside islikely driven byusage-based agent workloads and ecosystem monetization rather thanCopilot seat growth alone. 2)Azure constraint and capacity reallocation Lastquarter Microsoft noted it is diverting computefor internal capacity instead offocusing on revenue-generating Azure.This is a "narrative break',as Microsoft wasrewarded for the Azure growth outperformance in the last few quarters.The companynoted this last quarter that it:would have grown 4%+ YoY (in constant currency) butgrew 38% YoY due to these internal capacity shifts. The reallocations could limit upsidetogrowth expectations,and we are unsure of the rational given the company has means 3)OpenAlexposure Another issue was the risk of OpenAl not being able to grow revenues as committed.Managementnoted that half of its infrastructure is used for its own intemal activitiesand half to service others. Since the company works in an excess demand environment,any capacity shortfall of a single customer could be rerouted to others or to intemalneeds, although webelieve that any shortfall from OAl will be a negative driver for thestocks of all Cloud companies. We believe Microsoft stock is attractive and see twomain catalysts: orders converted torevenues as data center buildouts unlock some of the revenue backiog, and growth indemandforagentic Al,with MSFTbundlingagents with its own applications. Oracle Positioning and possi