Major cloud providers are pouring billions of dollars into infrastructure initiatives aimed at giving enterprises tighter oversight over data residency and access permissions. Amazon Web Services has earmarked upwards of 7.8 billion euros for its German-based European Sovereign Cloud, while enterprise software giant SAP has committed more than $20 billion toward sovereign cloud developments across Europe. Despite these massive capital injections, such spending alone fails to guarantee operational continuity if a vendor experiences an outage or cuts off access to critical services.

Fresh research from IBM investigates a vulnerable blind spot that cannot be resolved merely through infrastructure budgets. The tech conglomerate's newly published Digital Sovereignty Index assesses whether corporations can empirically validate their technological safeguards in real-world scenarios—including their capability to migrate between vendors without fracturing business operations.

While an overwhelming 82% of surveyed corporate decision-makers express robust confidence in their capacity to furnish proof of control, a mere 26% of businesses actually satisfy the index’s highest evidentiary benchmark. This widening chasm grows increasingly perilous as artificial intelligence assumes greater operational responsibilities, binding companies ever closer to third-party technology providers whose proprietary systems they may find difficult to monitor, explain, or govern during a crisis.

For Gary Cohn, IBM’s vice chairman, these mounting technological dependencies necessitate rigorous oversight from corporate boardrooms.

“Boards no longer have a choice. They are responsible for overseeing all the risks within an organization,” he notes in an exclusive discussion. “For some companies, the greater concern may be financial or reputational risk. But for every company today, digital sovereignty management is essential.”

True digital sovereignty transcends simply storing data within geographic borders; it requires preserving command over critical technology when operational dynamics shift. Organizations must maintain transparent audit trails detailing how AI algorithms formulate decisions, alongside viable contingencies if their primary tech suppliers become inaccessible.

The Divergence Between Assumed Confidence and Actual Control

IBM polled 2,003 executive decision-makers spanning seven nations, cross-referencing their responses with regulatory insights and customer case studies. The resulting index appraises control evidence across four fundamental domains—artificial intelligence, general technology, operational procedures, and data management—by measuring six specific criteria: accountability, traceability, safeguard efficacy, intervention capability, vendor-switching agility, and the proficiency to justify automated decisions under regulatory scrutiny.

Among the participating organizations, 94% disclosed facing a sovereignty-related complication over the preceding three years, with 81% of those entities legally or operationally mandated to provide formal proof of control. Yet, only 30% successfully navigated their most severe incident free of material operational disruption, and 53% encountered substantial roadblocks retrieving crucial data records held by external vendors.

It is worth noting that IBM derived its scoring metrics from self-reported descriptions provided by participants rather than conducting objective, independent audits.

Out of the enterprises falling into IBM’s elite evidence tier, 52% managed to resolve their premier sovereignty crisis without operational fallout, contrasted with just 22% in the intermediate tier. Nevertheless, the study stops short of establishing a direct causal link connecting rigorous control evidence to superior business outcomes, leaving respondents to define material disruption on their own terms.

“The first question I would ask a CFO who is questioning whether digital sovereignty justifies the investment is how much it would cost the business if one or more of its systems were compromised. I believe companies need to approach digital sovereignty much like an insurance policy, recognizing the importance of understanding how their systems operate, identifying potential vulnerabilities and continuously addressing them,” Cohn elaborated.

Cloud Titans Vie to Establish Sovereignty Standards

IBM's analytical stance conveniently bolsters its commercial offerings, notably Sovereign Core, which reached general availability on May 5. The platform delivers a customer-managed computing landscape featuring automated evidence generation and persistent compliance tracking, aligning neatly with the index's core pillars. Mistral integrated as the system's inaugural certified model provider, coupling IBM's control architecture with an independent AI vendor.

This philosophy mirrors IBM’s broader Red Hat strategy, which prioritizes open software architectures and seamless workload portability across diverse computing ecosystems. Cohn anticipates that customers will eventually procure idle compute capacity from competing suppliers based on pricing advantages, thereby dissolving risky reliance on isolated proprietary facilities.

“Just as the world has developed hybrid cloud, I believe it will also develop hybrid compute. This is where containerization becomes important, and where our work with Red Hat comes into play. You can package your data and formulas into containers and send them to a computing facility,” he explains. “Because everything remains packaged together, if something happens to that facility, you can move the workload to another computing facility in real time.”

While encapsulating software simplifies migration in theory, it does not guarantee that an entire corporate AI operation can be relocated instantaneously. Proprietary backend integrations often complicate transitions, and substitute AI models frequently yield disparate outputs.

Jonathan Adashek, IBM’s senior vice president and chief global affairs officer, insists that technology vendors must offer more than empty contractual assurances.

“You cannot genuinely claim to have control if the evidence needed to demonstrate it resides entirely with someone else. At IBM, we see this as a shared responsibility, with providers giving clients the visibility and evidence needed to verify their controls, while enterprises must understand their dependencies and establish accountability from the outset,” he asserts.

Industry rivals pursue vastly different models for dividing this operational responsibility. AWS runs an isolated European sovereign region staffed exclusively by EU citizens. During a stress test in October, Amazon temporarily severed the region from its global network backbone, routing operations exclusively through European internet service providers. The exercise proved the regional hub could function during a brief isolation phase, though it did not confer permanent independence or formal certification.

Similarly, Microsoft delivers tiered control mechanisms. External key management enables clients to safeguard their own root cryptographic keys, while its Data Guardian program introduces regional partner oversight whenever Microsoft engineers access EU and EFTA customer environments remotely—though this oversight fails to block remote access outright. Microsoft also features a Production Touch Ledger for tamper-evident logging alongside an upcoming Sovereign Control Panel preview.

Local operator models present yet another alternative. Google intends to transition operations and cryptographic keys for its German sovereign cloud to a Thales-affiliated enterprise, targeting general availability by year's end. Meanwhile, Delos—an SAP subsidiary—assumed direct operational control of its Microsoft-anchored German government platform in October, stripping Microsoft of administrative access ahead of a planned 2027 rollout. These arrangements delegate day-to-day oversight to domestic entities, even if the underlying infrastructure remains tethered to original vendor technology.

Oracle’s Alloy leverages a partner-managed framework, exemplified by SoftBank’s services in Japan. Snowflake has likewise disclosed partnerships with Orange Business to build a French sovereign service hosted on Bleu infrastructure, utilizing Orange for contracting and local European teams for administration, though the project remains in the proposal phase.

The race for digital sovereignty is rapidly infiltrating enterprise application layers as well. SAP has consolidated its products under the EU AI Cloud banner, and Salesforce has cleared select tools like MuleSoft and Tableau for deployment via French sovereign cloud partners. Concurrently, Databricks has pledged expanded European regions sans definitive deployment timelines. Consequently, enterprises cannot safely assume sovereign deployment models apply universally across their entire software stack.

Client-operated infrastructures afford the highest degree of operational command. Cisco grants customers absolute authority over its air-gapped hardware portfolios, which function completely detached from external networks, while HPE markets a disconnected Private Cloud AI solution. Still, running physical hardware locally does not eradicate every reliance. European cloud lobbying group CISPE recently targeted Broadcom over sovereignty concerns, noting that Broadcom remains the sole supplier of essential patches and software updates for its VCF cloud architecture.

While IBM’s index draws valuable attention to these hidden dependencies, it does not empirically prove that IBM's proprietary solutions mitigate those vulnerabilities any better than competing alternatives.

The High Financial Toll of Cloud Migration

Market research firm Gartner estimates global enterprise spending on sovereign cloud infrastructure will surge to $80 billion by 2026, marking a 35.6% jump over 2025 expenditures. As capital flows into the sector, rival firms are rushing to establish benchmarks for sovereignty readiness. Capgemini released a competing Digital Sovereignty Index in September, revealing that 36% of audited organizations would require upwards of 12 months to extricate themselves from a failing critical provider. Furthermore, Kyndryl—IBM’s former managed infrastructure spin-off—reported on November 10 that 68% of surveyed executives would willingly pay a 10% price premium to secure absolute operational control.

Despite heavy focus on documentation, IBM’s study highlights that only 33% of enterprises regularly or continuously test their ability to produce evidence of control.

Philip Guido, executive vice president and chief commercial officer at AMD, argues that enterprises should prioritize flexibility and vendor choice over absolute technological isolation. “It also provides a stronger foundation for infrastructure decisions driven by business outcomes rather than technological independence as an end in itself,” he notes. “Open standards and interoperable software frameworks can make it easier to move workloads between environments, although migrating AI systems may still require adaptation and testing.”

Because AMD stands to gain market share when businesses enjoy greater freedom to swap computing architectures, Guido's candid assessment regarding the friction of AI workload migration carries significant weight. Meaningful investment in digital sovereignty must ultimately identify which specific workloads can be relocated, the technical complexity involved, and the associated financial costs.

Traditional Governance Fails to Manage Autonomous AI Agents

AI-related complications plagued 44% of study participants, making them the most widespread sovereignty challenge cited in IBM's report. Artificial intelligence also registered the lowest evidence score among the index's four foundational evaluation pillars. Adashek warns that legacy corporate governance frameworks are entirely obsolete as autonomous AI agents assume greater decision-making authority.

“Most governance frameworks were designed for people and relatively static software. AI agents can now make decisions and take actions across systems and models that businesses may not own, which means governance must become continuous, with controls and oversight applied as agents operate rather than simply reviewing their actions afterward,” he emphasizes. “However, relying on a third-party model does not absolve an enterprise of accountability for the agents it deploys. It must still be able to trace their actions and intervene when necessary.”

Nathalie Barrera, global policy lead for trust and security at Palo Alto Networks, cautions that organizations routinely miscalculate the vast permissions they hand over to automated artificial intelligence agents.

“Most organizations believe they understand their access boundaries, but modern AI integrations frequently bypass traditional visibility layers,” Barrera concludes.