Beyond Contracts: How AI-Integrated Enterprise Ecosystems Generate Switching Costs and Vendor lock-in : A Qualitative Case Study of AI Infrastructure Platforms

Sammanfattning

Enterprise AI adoption is mediated by consolidated cloud platforms integrating compute infrastructure, foundation models, orchestration, and ecosystems. These platforms accelerate implementation but embed vendor dependence transcending contractual lock-in, reaching into technical and operational processes. The mechanisms through which AI platforms generate switching costs remain under-examined. This qualitative case study examines how enterprise AI platform ecosystems generate vendor dependence and how enterprises respond. The research addresses two questions: (1) What architectural design choices and ecosystem governance mechanisms create switching costs in enterprise AI infrastructure platforms? (2) How do enterprise firms navigate switching cost exposure in enterprise AI infrastructure platforms? The study employs a Yin Type 2 embedded single-case design with AWS and Azure as supply-side units. Supply-side analysis draws on documentary evidence and independent regulatory and analyst reports to investigate three propositions (P1 - P3). Demand-side analysis from four pilot interviews explores three propositions (P4 - P6); findings are treated as candidate patterns warranting validation. Documentary analysis identifies six switching cost mechanisms: cross-layer technical dependencies; ecosystem-level lock-in above individual services; ecosystem governance; value-creating features structurally inseparable from dependencies; learning effects compounding through usage; and compliance dependencies amplified by regulatory frameworks. The key finding is structural inseparability of value creation and dependency generation - features delivering highest value produce deepest lock-in. Three patterns emerge: architectural mitigation develops in response to experienced dependency rather than anticipation; multi-provider approaches function as negotiating instruments; and formal exit governance lags integration depth. Human capital and organisational knowledge accumulation may exceed technical switching costs in practical significance. Compliance frameworks intended to promote portability operationally amplify the structural switching costs they are designed to constrain. The study advances switching cost theory and platform economics by extending classical analysis to multi-layer AI infrastructure, characterizing lock-in as embedded in value creation rather than contractual restriction, and identifying structural incumbent advantages that current regulatory tools cannot address.