AI Will Not End the Software World, but It Will Reshape the Value Distribution Across Cloud and Enterprise Applications
AI summary card
AI Will Not End the Software World, but It Will Reshape the Value Distribution Across Cloud and Enterprise Applications
Bernstein believes that by around 2030, generative AI will boost demand for IaaS/PaaS, databases, and cloud infrastructure, while driving differentiation within SaaS, rather than simply letting AI "eat software."
- AI workloads will simultaneously drive demand for GPUs/ASICs, CPUs, networking, storage, and database usage, so the total addressable market for the cloud infrastructure layer is likely to expand.
- Enterprise AI is more likely to run at the cloud providers where the applications and data already reside, so the data, platform software, and operational capabilities of large cloud providers still form a moat.
- SaaS will not be disrupted as a whole; complex, deterministic workflow areas such as ERP, HCM, and full-suite CRM are more defensive, but vendors that innovate slowly and pure on-premises providers face higher risk.
- Slower cloud growth is not entirely driven by AI disruption; part of it reflects more mature enterprise application cloud migration and cloud penetration approaching higher levels.
- Generative AI itself could also be reshaped by the next wave of AI technology, so investors should not view the current AI stack as the final end state.
Report interpretation
Overview
This report discusses the medium- to long-term impact of generative AI on the global software and cloud technology stack, focusing on how IaaS/PaaS, hyperscalers, SaaS, enterprise applications, databases, and developer tools may evolve around 2030. The report rejects the simplistic bear-case narrative that AI will completely absorb software, arguing instead that AI will change how software is built, used, and monetized, but will not eliminate the importance of enterprise software, platform software, or cloud infrastructure.
Core views
The core view is that AI is broadly positive for IaaS/PaaS and the large cloud providers, because enterprise AI needs to run where the data, applications, and platform capabilities already reside, thereby increasing demand for GPUs/ASICs, CPUs, databases, storage, and platform services. The SaaS layer will diverge: incumbent vendors that can quickly embed AI, understand complex workflows, possess domain expertise, and maintain customer relationships can still survive and benefit; by contrast, software companies with weak innovation, simple functionality, or purely on-premise deployments are more likely to be displaced. The report also emphasizes that cloud migration remains in an early stage at the IaaS/PaaS layer, and this structural tailwind could persist beyond 2030.
Analysis framework
The report uses a technology-stack layering framework to examine IaaS, PaaS, SaaS, and sub-sectors within enterprise applications, and combines historical technology cycles, cloud migration maturity, AI workload characteristics, data residency requirements, platform-software dependence, and the trend toward sovereign cloud deployments to assess potential winners and risks. The timeline starts with the software implications triggered by ChatGPT in 2023, uses 2030 as the main observation point, and notes that the full impact may extend to around 2035.
Methodology notes
Major historical technology cycles usually take many years to fully reshape the software industry.
The report uses the history of SaaS and cloud computing as a reference, noting that it took years from the early commercialization of Salesforce and AWS to the emergence of meaningful revenue scale, so AI's full impact on enterprise software should not be assumed to happen instantaneously.
Different cloud and software layers have different sensitivity to AI, benefit paths, and disruption risks.
IaaS/PaaS benefit more directly from growth in demand for compute, databases, storage, and platform services, while SaaS depends on application complexity, workflow determinism, data semantics, customer relationships, and vendor innovation speed.
Enterprise AI is better suited to run where the data and core applications already are.
Because of performance, security, governance, and complexity requirements, AI inference and agent workloads are more likely to remain within existing cloud, database, and application platforms, reinforcing cloud providers with data and platform-software capabilities.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- IaaS/PaaS hyperscalersAI workloads expand demand for cloud infrastructure and platform services
- Strengths
- They have global scale, data center operations capabilities, platform software, customer data, and enterprise-grade security and governance.
- Weaknesses
- Capital expenditure pressure, volatility in AI hardware supply and demand, and sovereign-cloud or deglobalization trends may increase architectural complexity.
- Comparison
- Compared with smaller regional cloud providers, the large hyperscalers have stronger platform software depth, ecosystem, and customer base; however, regional and government-backed clouds may gain share in certain markets.
- Risks
- If AI infrastructure becomes commoditized, prices fall quickly, or customers migrate workloads to lower-cost suppliers, profitability could come under pressure.
- ORACLE CORPThe report views it as a potential beneficiary of cloud, databases, OCI, Oracle Alloy, and sovereign cloud trends
- Strengths
- It has an enterprise database base, PaaS capabilities, OCI hardware, and white-label/sovereign cloud solutions.
- Weaknesses
- It still needs to prove that it can continue to gain share and maintain execution in global hyperscaler competition.
- Comparison
- The report believes Oracle is relatively well positioned in sovereign cloud and white-label OCI capabilities, creating differentiation versus large cloud providers and regional partners.
- Risks
- Cloud migration speed, competitive pressure, AI infrastructure margins, and customer adoption pace could affect the degree of realization.
- SaaS incumbentsAI creates opportunities for product redesign and automation, but also increases disruption risk
- Strengths
- Customer relationships, domain expertise, complex business processes, semantic data, and embedded workflows provide defensiveness.
- Weaknesses
- Vendors with simple functionality, slow innovation, or limited AI integration capability are more vulnerable to replacement.
- Comparison
- ERP is considered the most protected, followed by HCM and full-suite CRM; differentiation across SaaS submarkets is significant.
- Risks
- AI-native tools lower the barriers to development, customers re-evaluate their application stacks, and cloud migration maturity in some submarkets has already slowed growth.
- Database and PaaS vendorsAI agents and enterprise data usage increase demand for databases, vector databases, and related platform services
- Strengths
- AI needs access to, governance over, and processing of enterprise data, so databases and platform software remain key foundations.
- Weaknesses
- Some PaaS functions may be replaced or redefined by AI-native platforms.
- Comparison
- The report cites MongoDB, Snowflake, and ServiceNow as potential beneficiaries from incremental database or platform usage, though the degree of benefit differs.
- Risks
- The emergence of new specialized databases, PaaS function reconfiguration, and competition from built-in cloud-provider services could alter the profit pool.
- Development toolsAI coding and agentic development platforms will replace some traditional development tools while expanding overall usage scenarios
- Strengths
- Non-IT staff may also use development tools to create agents or automation flows, expanding the user base.
- Weaknesses
- Enterprises are currently switching AI development tools frequently, so market stability is still limited.
- Comparison
- Traditional development tools face replacement by AI/AI-enabled platforms, but total usage may rise as agentic development becomes more widespread.
- Risks
- Commoditization, intensifying competition, and delayed customer standardization could suppress revenue growth.
Key data
- Main Time WindowAround 2030; the full impact may extend to around 2035The report uses the impact of generative AI on software beginning in 2023 and uses 2030 as the key point for assessing the main effects.
- Historical Technology Cycle10 to 20 yearsThe report argues that major technology disruptions usually last for many years rather than being completed instantly.
- AWS Revenue Disclosure ReferenceAWS revenue was about $3.1B in 2013, and it began separate disclosure in Q1 2015Used to illustrate that cloud computing took a long time to go from emergence to visible commercial scale.
- SaaS Early ReferenceSalesforce was founded in 1999, and revenue was about $176M in 2005Used to illustrate that SaaS commercialization also took years of buildup.
- Cloud PenetrationCloud penetration in some software submarkets has reached 70%+The report believes slowing growth in some enterprise application software submarkets is related to more mature cloud migration.
- Major HyperscalersAmazon, Microsoft, Google, Oracle, AlibabaThe report views these companies as leaders in cloud infrastructure in terms of revenue, global footprint, customer base, and influence.
- Example BeneficiariesMSFT, ORCL, MDB, SNow, SAP, HubSpot, Workday, SalesforceThe report mentions MSFT and ORCL as hyperscaler beneficiaries within coverage, while database and SaaS vendors may benefit or be affected to different degrees.
Impact & implications
The investment implication is that the market's fears about AI disrupting software may be too one-size-fits-all. Cloud infrastructure, PaaS, databases, and data platforms that can host enterprise AI agents may see incremental demand; within SaaS, investors should focus more on product complexity, AI embedding speed, workflow moats, and execution, rather than simply re-rating the entire software sector lower. Oracle is mentioned repeatedly in the report, and its database cloud migration, OCI, Alloy, and sovereign cloud capabilities are viewed as potential beneficiaries.
Risks
- Rising price competition in AI infrastructure could leave hyperscaler margins below expectations.
- Model companies or AI-native companies may expand across the full technology stack, changing the distribution of cloud and software value.
- Some SaaS submarkets could be replaced by generative AI or agentic platforms, especially applications with simpler functionality and less complex workflows.
- Deglobalization and sovereign-cloud requirements could support local and regional cloud providers, weakening the expansion efficiency of global cloud providers in some markets.
- The generative AI stack itself could be reshaped by new technology waves such as inference, agentic AI, first-principles AI, or quantum AI.
- Enterprise AI adoption may be slower than expected, delaying the realization of demand for AI-related cloud, database, and PaaS services.
What to watch
- The pace of enterprise migration of key workloads to the cloud before 2030, especially databases and core applications.
- The actual change in CPU, GPU/ASIC, storage, networking, and database usage after agentic AI becomes more widespread.
- Customer expansion and revenue performance for Oracle OCI, Oracle Alloy, and sovereign cloud partnerships.
- The speed and monetization ability with which SaaS vendors deeply embed AI into ERP, HCM, CRM, and vertical applications.
- Whether the AI development tools market moves from frequent switching toward stable standardization.
- Gross margins, pricing trends, and customer stickiness for AI workloads in cloud infrastructure.
- Whether the next generation of AI technology starts a new disruption cycle before generative AI fully permeates software.