When more than half of enterprises already have autonomous AI agents running in production, the question isn't whether to deploy AI — it's who controls it. A new MIT Technology Review Insights report, based on a survey of 2,000+ senior executives across 13 countries, delivers a clear answer: enterprises "deeply committed" to controlling their data, infrastructure, and governance achieve 5x the ROI on AI initiatives compared to peers relying on third-party platforms.
The research reveals a 0.93 correlation between sovereignty commitment and AI success outcomes — the strongest single driver identified in the study. For CIOs evaluating infrastructure strategy and CFOs measuring AI returns, the message is blunt: data control is now the best predictor of whether your AI investments will pay off.
The New Bargain: Capability Now, Control Never
When generative AI first moved from research labs into business applications, enterprises made a tacit trade: feed proprietary data into third-party models, get powerful results, but give up control over where that data lives and how it's governed.
That deal worked when AI was experimental. But now, with 72% of enterprises running AI agents in production and IDC projecting 1 billion active agents by 2029 executing 217 billion actions per day, the terms have changed.
Kevin Dallas, CEO of EnterpriseDB (EDB), frames the stakes: "Data is really a new currency; it's the IP for many companies. The big concern is, if you're deploying an AI-infused application with a cloud-based large language model, are you losing your IP? Are you losing your competitive position?"
For business leaders, this isn't a technical question — it's strategic. 70% of global executives now believe they need a sovereign data and AI platform to be successful, according to EDB's internal data.
Why Sovereignty Delivers 5x Returns
The MIT Technology Review study identified three economic drivers behind the 5x ROI multiplier:
1. Governance Scales With Control
When you own the infrastructure, you set the policies. Devin Pratt, research director at IDC, explains: "Sovereignty defines which agents can touch the data, in which region, under which policies, and how all of that is monitored and audited. Rather than being a brake on agentic AI, sovereignty sets the safe operating boundaries that allow organizations to scale with confidence."
Translation for CFOs: Fewer compliance failures, faster audits, lower legal risk. Every avoided breach or regulatory fine flows directly to the bottom line.
2. IP Protection Unlocks High-Value Use Cases
Enterprises holding back proprietary data from third-party models are limiting AI to generic tasks. Sovereignty leaders feed their most sensitive data into self-hosted models, unlocking differentiated use cases competitors can't replicate.
Mukesh Chandak, director of Strategy and Innovation at Thales, argues sovereignty "unlocks the highest-value AI use cases rather than constraining them."
Translation for CIOs: Your competitive moat isn't the model — it's the proprietary data you can safely deploy. Sovereignty removes the blocker.
3. Hybrid Models Dominate, But Ownership Matters
The study reveals hybrid environments are the dominant operating model — 95% of organizations plan to establish their own AI and data platforms within the next three years, even if they continue using cloud services selectively.
But there's a difference between "hybrid" and "sovereign hybrid." The 5x ROI gap shows up when enterprises retain ownership and governance control, not just when they mix cloud and on-prem infrastructure.
The Agentic Era Changes The Rules
Traditional AI governance was built for static data estates. Agentic AI changes the game: these systems don't just generate outputs — they take actions, trigger workflows, and make real-time decisions on operational data, often with minimal human oversight.
Michael Schrage, research fellow at the MIT Sloan School's Initiative on the Digital Economy, warns that sovereignty can't be a one-time exercise when models and multi-agent workflows evolve weekly.
The report surfaces three top drivers of sovereignty efforts:
- Security and resilience: 85%
- Data localization: 74%
- Ownership and control: 72%
Notice what's missing from that list: cost savings. Sovereignty isn't a cost play — it's a strategic advantage play. The 5x ROI comes from enabling use cases competitors can't touch, not from cutting infrastructure spend.
What "Deeply Committed" Actually Means
The study categorized enterprises into three tiers based on sovereignty commitment. The 5x ROI gap appeared between "Deeply Committed" organizations and everyone else.
"Deeply Committed" enterprises:
- Own their data infrastructure (not just access it)
- Set their own governance policies (not inherit vendor defaults)
- Control model deployment (not rely on third-party APIs exclusively)
- Audit autonomously (not depend on vendor transparency)
This isn't an all-or-nothing choice. You can use OpenAI's API for non-sensitive workflows while running sovereign models for proprietary data. The difference is who controls the stack when it matters.
The Leadership Gap Slowing Response
Here's the problem: while 95% of organizations plan to establish sovereign platforms within three years, many lack a clear ownership model for AI governance.
The report flags a leadership gap: CIOs understand infrastructure control, CFOs understand ROI, but few organizations have aligned both perspectives into a unified sovereignty strategy.
Result? Delayed decisions, fragmented ownership, and continued reliance on third-party platforms even when the ROI data says otherwise.
NVIDIA's Davos Warning: Sovereignty Goes National
This isn't just an enterprise conversation. At the World Economic Forum's annual meeting in Davos (January 2026), NVIDIA CEO Jensen Huang argued for national AI sovereignty:
"I really believe that every country should get involved to build AI infrastructure, build your own AI, take advantage of your fundamental natural resource — which is your language and culture — develop your AI, continue to refine it, and have your national intelligence be part of your ecosystem."
For enterprises, the takeaway is clear: If nations are prioritizing sovereign AI infrastructure, your third-party model provider's policies could change overnight based on geopolitical shifts you don't control.
Sovereignty isn't just about ROI — it's about operational resilience in an environment where AI infrastructure is becoming national infrastructure.
What This Means For Decision-Makers
For CIOs and CTOs:
Stop treating sovereignty as a compliance tax. The MIT data shows it's the strongest predictor of AI ROI. If you're evaluating vendor lock-in vs. self-hosted infrastructure, the economic case just shifted.
Action: Map which AI use cases require proprietary data. For those workflows, evaluate sovereign deployment options (on-prem, private cloud, hybrid with ownership control). Use third-party APIs for generic tasks only.
For CFOs and Business Leaders:
Ask your CIO one question: "Who controls the data and governance for our highest-value AI use cases?" If the answer is "AWS" or "OpenAI," you're leaving 5x ROI on the table according to this study.
Action: Require ROI projections to include sovereignty scenarios. Model the cost of self-hosted infrastructure vs. the revenue upside of enabling proprietary use cases competitors can't replicate.
For Legal and Compliance Teams:
Sovereignty sets the boundaries that allow scaling. Every policy debate about what data can feed AI models is a sovereignty question in disguise.
Action: Work with IT to define sovereignty tiers: which data requires full ownership, which tolerates third-party processing under contract, and which can live anywhere. Make this a procurement requirement for new AI vendors.
The Bottom Line
If you're running AI agents in production (and 72% of enterprises are), sovereignty isn't a nice-to-have — it's the difference between 1x and 5x ROI.
The MIT study's 0.93 correlation is unusually strong for enterprise research. In most cases, ROI depends on execution, market timing, and a dozen other variables. Here, one variable dominates: who controls the infrastructure.
As enterprises race toward 1 billion active agents by 2029, the winners won't be the ones who deployed AI first. They'll be the ones who deployed it on infrastructure they control.
The question for your organization: Are you building on rented land, or do you own the foundation?
Continue Reading
- Anthropic Beat OpenAI in Enterprise AI — The $2K Problem
- AI Governance Frameworks Every Enterprise Needs in 2026
- Why 60% of AI Projects Fail — And How to Fix Yours
About THE D*AI*LY BRIEF
Enterprise AI insights for technical and business leaders. Subscribe at beri.net or follow Rajesh Beri on LinkedIn, Twitter/X, and Facebook.
