2026年二季度软件财报的主题：AI推动支出改善

Themes into 2Q26 Software Earnings: AI Drives Improvement inSpending; Remain Selective Tyler RadkeAC+1-415-951-1660tyler.radke@citi.com CITI'S TAKE Ahead of software earnings and detailed company-specific previews, weare updating our views and top picks. Our CIO Survey, checks withpartners/customers, etc, IT Day and recent industry conferences point to animproved spending backdrop after a softer Q1. However, we believeinvestors should resist extrapolating recent improvements into a broad-based recovery across software. We continue to see spending improvementremaininghighly concentrated among a relatively narrow group ofbeneficiarieswithin our software universe,particularly cloud&datainfrastructure, and a select group of application vendors with strong AI-driven product cycle. We continue to see consolidation pressure andheightened ROI scrutiny across many traditional application softwarecategories. As a result, we remain selective in our positioning and continueto favor companies most directly levered to enterprise AI adoption andinfrastructure buildout. Our top picks are MDB, SNOW and PLTR. Yitchuin Wongyitchuin.wong@citi.com Peter Griffithpeter.griffith@citi.com Andrew Girard, CFAandrew.girard@citi.com See Overall Spending Improving QoQ But Consolidation Pressure Remains —Ourconversations at Snowflake Summit, Databricks events, AWS Summits and Cannesreinforced our view that modern data architecture is foundational for AI deploymentacross a variety of industries. At the same time, we continue to hear consolidationpressure and heightened ROI scrutiny across many traditional application softwarecategories, particularly those which lack an AI story and have increasingly relied onprice increases. We continue to believe the data management companies are bestpositioned to capture incremental enterprise AI spending. 5 Key Themes into 2Q26 Earnings —1. IT Spending Improved, But We Are NotCalling a Broad Software Recovery; 2. AI Spending Continues to Favor the DataLayer; 3. From Token Maximization to AI Optimization; 4. Capex and Returns Remainthe Next Major Debates; 5. Less Enthusiastic Earnings Setup for Neoclouds. Top Picks —Heading into 2Q earnings, we see improving spending trends relative to1Q but remain cautious about characterizing the environment as a broad softwarerecovery. Instead, we continue to see a highly bifurcated market in which AI-relatedspending is benefiting a relatively concentrated group of winners. Our preferredpositioning remains centered on MDB, SNOW, and PLTR within data infrastructureand AI enablement, MSFT among hyperscalers and AI infrastructure beneficiaries,and FIG and SHOP within application software. In our view, these companies remainamong the best positioned to capture incremental enterprise AI spending whilemany areas of software continue to face elevated competitive pressure and ongoingquestions around long-term business model durability. Key Themes into Earnings 1. IT Spending Improved, But Unlikely Enough for a Broad-Based Software Recovery Our latest CIO work suggests technology spending improved through the quarterfollowing a weaker 1Q, with signs that some delayed projects moved forward. Webelieve part of the improvement likely reflects the continued expansion of AI-related spending, while partner commentary also suggests some transactions thatslipped earlier in the year may have closed in 2Q. That said, we see limited evidencethat spending improvements are being distributed evenly across software.Incremental budget dollars continue to flow disproportionately toward AIinfrastructure, cloud platforms, and data modernization projects. 2. AI Spending Continues to Favor the Data Layer Perhaps the strongest takeaway from our recent work is that AI remains one of thefew areas where strategic prioritization continues to increase. Across CIOconversations, industry conferences, and customer discussions, enterprisesconsistently cite data modernization as one of the most important prerequisites forsuccessful AI deployment. Modern data platforms such as Snowflake, Databricks,MongoDB, Palantir, and the hyperscalers are increasingly occupying strategicpositions within enterprise AI architectures. As organizations move from experimentation toward broader deployment, webelieve data management, governance, and orchestration become increasinglyimportant parts of the technology stack. This remains a key reason why datainfrastructure continues to be our preferred segment within software. 3. From Token Maximization to AI Optimization One of the more notable shifts we observed during the quarter was the growingfocus on AI efficiency. Earlier phases of AI adoption prioritized access to leadingmodels and maximizing model performance. Increasingly, enterprises are shiftingattention toward managing costs, optimizing workloads across models, andimproving overall AI economics. At Cannes, many organizations discussed routing workloads across differentmodels based on cost and complexi