AI will not end the world of software, but it will reshape value allocation in cloud and enterprise software

Overview

The report focuses on whether AI will end the existing software world, with emphasis on the impact of AI on software and the cloud computing stack over the next five years or more, especially after 2030. Bernstein's core view is that AI will significantly change how software is built, deployed, and used, but will not make software disappear overall. Instead, AI will increase utilization intensity in cloud infrastructure, databases, platform software, and some enterprise applications, while putting greater exit pressure on vendors with slower innovation, weak domain moats, or those still in on-premise deployment models.

Core views

The report argues that IaaS/PaaS is the primary beneficiary layer of the AI revolution. AI inference, agents, database calls, storage, and enterprise workload migration all increase demand for GPU/ASIC, CPU, networking, storage, and PaaS software. On the SaaS side, AI will improve personalization, customization, and automation, with applications becoming the orchestration layer for complex AI agents and business processes. The authors reject the broad bearish thesis that AI will 'eat software,' contending that recent deceleration in some software growth is more due to matured cloud penetration and weakening cloud migration tailwinds than AI having fully destroyed software business models.

Analysis framework

The report uses three analytical lenses: technology cycles, cloud stack layering, and enterprise software moats. It first compares major cycles over the past 10 to 20 years—mainframe, minicomputer, PC/client-server, web, SaaS, cloud-native, and AI-native—then separates IaaS, PaaS, and SaaS layers and evaluates AI’s incremental effects on TAM, profitability, stickiness, data locality, development tools, databases, and enterprise application workflows, and finally compares relative beneficiaries and detractors across several software companies.

Methodology notes

• Technology cycle analysis10 to 20 year disruption cycles

  Major technology cycles typically last 10 to 20 years, and AI impacts should be assessed using 2030 and 2035 as key checkpoints.

  The report argues that SaaS and cloud computing did not scale overnight, and although generative AI is being adopted more quickly, the full impact on enterprise software still requires years to materialize.

• Industry chain layeringIaaS/PaaS/SaaS stack segmentation

  Split cloud and software stack into infrastructure, platform, and application layers, and assess incremental demand and substitution risk from AI at each layer.

  IaaS/PaaS benefit from demand for compute, databases, storage, and platform software; SaaS outcomes are more mixed and depend on innovation speed, process complexity, and domain semantic depth.

• Enterprise application moatComplex deterministic workflows and domain knowledge moat

  Enterprise software defensibility comes from complex deterministic business processes, client data, semantic knowledge, and customer relationships.

  Applications such as ERP, HCM, and full-suite CRM are less likely to be fully replaced by general-purpose AI tools if they rapidly embed AI and continue executing reliably.

Asset mapping & comparison

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

• IaaS/PaaS hyperscalers (Amazon AWS, Microsoft Azure, Google GCP, Oracle OCI, Alibaba)

  AI workloads, cloud migration, data locality, and agentic inference increase demand for infrastructure and platform capabilities.

  Strengths

  Scale, global footprint, customer base, platform software, data locality, and advanced datacenter operational capabilities form a moat.

  Weaknesses

  Decoupling trends may drive local or sovereign cloud competition, while AI infrastructure capex and pricing competition still require close monitoring.

  Comparison

  The report argues that AWS, Azure, and GCP together account for about two-thirds of the IaaS market, while Oracle is notable for sovereignty cloud, Alloy, and PaaS capability distribution.

  Risks

  If AI compute supply and demand normalize and prices decline quickly, or if new architectures reduce hyperscaler stickiness, margin and revenue growth could come under pressure.

• Oracle / ORCL

  Benefits from OCI, Oracle database cloud migration, PaaS capabilities, and enterprise application AI integration.

  Strengths

  The report views Oracle as relatively advantaged in white-label OCI, Alloy, sovereign cloud, and broad PaaS capabilities; migration of Oracle database workloads from on-premise to cloud also presents long-term opportunity.

  Weaknesses

  The valuation table shows negative net cash and weak free cash flow metrics at Oracle, while infrastructure spending and debt burden require attention.

  Comparison

  In SaaS, the report ranks SAP and Oracle as better positioned; in IaaS/PaaS, ORCL and MSFT are both beneficiaries within coverage.

  Risks

  If cloud migration pace slows, sovereign cloud partnerships underperform, or AI infrastructure economics are weaker than expected, growth elasticity could decline.

• Database vendors (MongoDB, Snowflake, Oracle database, etc.)

  AI applications, agent workflows, and enterprise data calling increase demand for databases, vector databases, and storage.

  Strengths

  Enterprise AI tends to run where data resides, giving databases and data platforms a strategic position.

  Weaknesses

  Some specialized database offerings may be re-architected, potentially altering competitive dynamics.

  Comparison

  The report notes that MDB and Snowflake may benefit to some extent from AI-driven incremental database usage.

  Risks

  If enterprise AI deployment is slower than expected, incremental database usage and storage growth could be delayed.

• Enterprise SaaS (SAP, Oracle, Workday, Salesforce, HubSpot, etc.)

  AI will make enterprise applications more personalized and automated, turning them into orchestration layers for agents and complex workflows.

  Strengths

  Complex deterministic processes, domain knowledge, semantic relationships, customer relationships, and existing data form defense moats.

  Weaknesses

  SaaS submarkets with slow innovation, simple functionality, or weak domain depth may be substituted by generative AI.

  Comparison

  The report sees SAP and Oracle as best positioned, HubSpot as relatively less exposed to AI headwinds, Workday as likely to remain stable, and Salesforce as facing slight headwinds with partial offset from Agentforce.

  Risks

  If customers shift to AI-native lightweight apps or incumbent vendors fail to embed AI quickly, valuation and growth may remain under pressure.

• Development tools

  AI coding, agent development platforms, and AI-augmented tools will reshape the development tools market.

  Strengths

  Total usage may rise because non-IT users may also use development tools to build agents.

  Weaknesses

  Tool switching is still frequent at present, and some development tool capabilities may become commoditized.

  Comparison

  The report notes that products like Claude, OpenAI Codex, and Cursor already reflect AI's impact on how code is written.

  Risks

  If AI-native platforms commoditize differentiation, traditional development tool revenue and stickiness may decline.

Key data

• Core observation windowAround 2030; full impact may extend through 2035The report takes 2023, when ChatGPT began impacting enterprise software, as the start point and sees 2030 as a reasonable checkpoint for judging the main impact.
• Historical technology cycle lengthAbout 10 to 20 yearsThe report uses major historical compute technology cycles to show that diffusion of enterprise software and cloud migration typically takes a long time.
• Cloud penetration in selected SaaS submarketsAbove 70%The report says cloud penetration in certain software submarkets such as CRM and collaboration is already high, and cloud migration tailwinds are weakening.
• AWS early scale referenceAWS revenue was about $3.1B in 2013, and Q1 revenue was only disclosed in 2015Used to illustrate that cloud computing also took years to generate meaningful revenue after inception.
• Salesforce early revenue referenceRevenue was about $176M in 2005Used to illustrate that SaaS commercialization and scaling were not completed instantly.
• IaaS market concentrationAWS, Azure, and GCP together account for about two-thirdsThe report cites IDC methodology stating the top three cloud vendors dominate the majority of the IaaS market.
• ORCL market priceUSD 163.00The investment implications table shows Oracle's stock price as of 2026-04-14.
• Oracle valuation metricsNTM revenue growth 28.0%, EV/NTM revenue 7.2x, P/FE 21.4xFrom the valuation comparison table, showing Oracle's growth and valuation position among the covered software companies.

Impact & implications

On the investment side, the report is more favorable toward hyperscale cloud vendors, database platforms, and enterprise software companies that can quickly embed AI into core workflows. MSFT and ORCL are explicitly identified as hyperscale beneficiaries within coverage, while database-related companies such as MongoDB and Snowflake may also benefit from incremental database usage driven by AI. In SaaS, SAP and Oracle are viewed as better positioned, HubSpot is relatively less exposed to AI headwinds, Workday is expected to remain broadly stable, and Salesforce may face slight headwinds that could be partially offset by Agentforce.

Risks

• AI infrastructure pricing and margins may decline once supply and demand normalize, affecting the economics of hyperscale cloud vendors.
• Generative AI may be disrupted by later technology waves such as reasoning, Agentic AI, First principles AI, and Quantum AI, and the current investment theme may not be the end-state.
• Slow-innovating SaaS vendors, pure on-premise vendors, and simplistic software submarkets may be substituted.
• Decoupling, data sovereignty, and local regulation may weaken global hyperscaler scale advantages in certain regions.
• The pace of enterprise AI adoption, data governance, security, and performance requirements may cause revenue realization to lag market expectations.

What to watch

• Whether, by 2030, enterprise AI inference and agentic workloads materially increase usage of CPU, GPU/ASIC, networking, and storage.
• Customer growth, regional expansion, and PaaS attach rates for Oracle OCI, Alloy, and sovereign cloud partnerships.
• Whether usage of databases, vector databases, and enterprise data platforms continues to rise as AI applications are deployed.
• After AI embedding, whether enterprise software vendors in ERP, HCM, and CRM deliver on pricing, retention, and automation value.
• Whether the development tools market moves from frequent switching to a more stable regime, and whether AI leads to tool commoditization.
• Whether SaaS submarkets with already high cloud penetration continue to experience growth deceleration as cloud tailwinds weaken.