IBM 연구

Most enterprises are embedding AI into core operations faster than they can control it. A new global study from the IBM Institute for Business Value, conducted in collaboration with Oxford Economics, reveals that the vast majority of organizations have dangerously little visibility into their own AI supply chains. The findings, published June 17, 2026, paint a stark picture of dependency, vulnerability, and a narrow path to resilience .

The Accountability Gap: 91% Lack Full Visibility

The study, titled "The Calculus of AI Sovereignty," surveyed 1,000 senior executives across 16 countries and 17 industries between February and April 2026 . Its central finding is what IBM calls the "Accountability Gap": 91% of executives report that their organization does not fully understand its AI dependencies across vendors, models, and infrastructure . This lack of visibility extends to switching costs: 71% of surveyed executives said it would be difficult to switch their primary AI vendor or model .

The Cost of Dependency: What a Seven-Day Outage Means

The research quantified the operational risk of this concentration. 81% of executives said a seven-day vendor outage would cause severe or critical disruption, effectively halting operations . This is not a theoretical concern—surveyed leaders reported an average of six AI-related disruptions over the past two years, largely driven by vendor services . Additionally, 68% said meeting data residency and sovereignty requirements across different geographies is challenging .

Only 7% of Organizations Have Advanced AI Control

Despite widespread AI adoption, very few organizations have achieved meaningful control over their AI systems. Only 7% of surveyed organizations operate at the most advanced level of AI control . The difference in outcomes is dramatic: these organizations protect 55% more operating profit from AI-driven disruptions compared to their peers .

The Flexibility Premium: 72% Would Pay More for Strategic Control

Executives are acutely aware of the lock-in risk. 72% of surveyed executives said they would accept a 20% cost increase to maintain their current AI vendors if it improved strategic flexibility . This finding suggests that the primary concern is not cost itself, but the ability to adapt and change direction.

Multi-Vendor Reality: Strategy or Accident?

While 73% of organizations describe their AI environments as intentionally multi-vendor, the drivers are often reactive rather than strategic. The leading reasons are independent business unit decisions (69%), geographic necessity (69%), and legacy complexity from mergers, acquisitions, and historical decisions (57%) . This suggests many enterprises end up with multiple vendors through circumstance, not design.

The Path Forward: Infrastructure and Architecture

IBM positions its concept of AI Sovereignty—running AI workloads in controlled, open hybrid-cloud environments built on Red Hat OpenShift—as the prescription for these control gaps . The study's findings align with several major strategic moves by the company:

• HashiCorp acquisition: Closed approximately 12 months before the study, being integrated with OpenShift to bolster infrastructure control .
• $11 billion Confluent acquisition: Announced in spring 2026, aiming to give enterprises real-time data streaming capabilities. This addresses the bottleneck that approximately 90% of corporate data remains unstructured and LLMs capture only 1% of it .
• watsonx Orchestrate: Being combined with Confluent's data fabric to let enterprises steer existing AI agents rather than build from scratch .
• Apptio: IBM's IT cost management unit launched "Conversational Insights" tools to help firms gain command of IT costs .

The study explicitly frames the "Accountability Gap" as a market opportunity: only 11% of organizations have deployed AI agents in production at scale, and IBM's platform-agnostic, multi-vendor approach aims to capture enterprises seeking strategic flexibility over lock-in .

Market Reaction

On the study's release date (June 17, 2026), IBM stock closed at $262.04, a decline of 3.24% on the day . The stock had rallied approximately 22% from a 52-week low of $212.34 in May 2026, but remained roughly 10-16% lower year-to-date in dollar terms depending on the data source . Its 52-week high was $332.46, meaning the stock was roughly 20% below that peak .