IBM Study: Limited Control and Rising Dependencies Leave Enterprises Exposed in the Age of AI

Details

IBM's Institute for Business Value released 'The Calculus of AI Sovereignty,' a study of 1,000 senior executives globally that quantifies just how locked in enterprises have become to AI vendors they cannot easily change. The top-line numbers are striking: 91% of respondents don't fully understand their dependencies across AI vendors, models, and infrastructure; 71% say switching their primary AI vendor would be difficult; and executives report an average of six AI-related disruptions in the past two years.\n\nThe operational risk is concrete: 81% say a seven-day vendor outage would cause severe or critical business disruption, effectively halting operations. Respondents cite unexpected price increases, usage restrictions, model deprecations, and performance degradation as recurring disruption sources — all events that have occurred across major AI providers in the past two years.\n\nThe study introduces 'AI sovereignty' as the business response: designing AI systems to adapt data, models, and infrastructure as conditions change. Organizations with the most advanced AI control capabilities protect more than half of their operating profit from AI-driven disruption, while those without are fully exposed to vendor decision-making they cannot influence.\n\nData residency adds a compliance layer: 68% find meeting data residency and sovereignty requirements across geographies challenging, which compounds vendor dependency with regulatory exposure.