New Enterprise Shift: AI, Digital Sovereignty, and the Rise of Enterprise Agents

Constellation analyst Holger Mueller recently sat down with Raju Vegesna, Chief Evangelist at Zoho, to discuss digital sovereignty, GPU gravity, and where enterprise AI adoption is really headed. Several points from the conversation deserve a closer look.

Bottom line up front: most sovereignty regulation today addresses where data lives, not where knowledge goes. That gap will matter more as AI systems scale.

Sovereignty Is a Spectrum, Not a Switch

Digital sovereignty has become the hottest topic in enterprise technology, and for good reason. Amazon's investment in a sovereign cloud region in Brandenburg, Germany, operated exclusively by EU passport holders, signals that sovereignty has moved from concept to capital commitment. But Mueller's conversation with Vegesna surfaced a more nuanced reality: sovereignty is not a binary state. Organizations sit somewhere on a spectrum across energy, trade, finance, physical security, and now AI. Few have mapped where they actually stand on each dimension.

The Book Analogy Exposes a Regulatory Blind Spot

Vegesna offered a framing worth remembering: data is a book. Data residency laws require the book to stay inside a country's borders. But if someone walks in, reads the book, and leaves, the book stays behind while the knowledge walks out. Regulators are answering the data residency question. They are not yet answering the question of knowledge sovereignty.

This distinction matters more with every AI deployment. Models trained or fine-tuned on in-country data can still export the value of that data the moment the model itself, or its outputs, cross a border. Judicial sovereignty compounds the problem: if the actors involved sit outside a country's jurisdiction, legal recourse is limited even when a violation is provable, and AI's black-box nature makes that violation difficult to prove in the first place.

GPU Gravity Is Becoming the New Data Gravity

Enterprises are familiar with data gravity, the tendency for data to attract more data and the applications built around it. Mueller and Vegesna's conversation points to a parallel force emerging around compute. As agent-to-agent transactions become more common, the speed and cost of the underlying GPU infrastructure will start to determine outcomes. An agent running on faster, more current architecture will simply out-negotiate one running on older hardware. This is a new variable for sovereignty planning, one that pulls organizations toward proximity with the fastest available compute, sometimes in tension with data localization requirements.

The Pendulum Swings Back Toward On-Premise

Enterprise infrastructure has cycled between centralization and decentralization for decades: mainframes, PCs, cloud, and now a partial return to on-premise. What is different this time is the driver. Agentic AI workloads are largely about capturing intent, generating a query, and running it against existing systems and infrastructure. That pattern favors on-premise architecture, giving on-prem a second life it would not have had otherwise.

Verifiability, Not Capability, Is the Real Adoption Gate

Perhaps the most practical insight from the conversation: AI is penetrating industries fastest where output can be verified quickly. Code either compiles or it does not, which is why coding use cases have advanced furthest. Industries with longer verification cycles, such as healthcare or pharmaceuticals, will take longer to see AI adoption at scale, regardless of how capable the underlying models become. Enterprises evaluating AI investment should weight verifiability as heavily as raw model performance.

Gold Rush or Real Revolution

Mueller and Vegesna also addressed the elephant in the room: is the current scale of AI investment justified, or is this the modern equivalent of the 1849 gold rush? Vegesna's answer split the difference. Yes, there is exuberance, and yes, some investments will not pay off. But unlike the gold rush, today's AI capital is concentrated among a small number of companies with resources that now rival the GDP of entire nations. That concentration is itself fueling the sovereignty conversation, as governments grow uneasy about companies operating at a scale historically reserved for nation-states.

What This Means for Enterprises

Organizations building AI strategy in 2026 should treat sovereignty as a multidimensional planning exercise, not a checkbox. Data residency compliance is necessary but insufficient. Enterprises need to account for where knowledge, not just data, ultimately resides, and they need to weigh GPU access and verifiability as seriously as they weigh model selection.

Agents are becoming the new office suite. The organizations that get sovereignty, infrastructure, and trust right will be the ones that capture the advantage as that shift accelerates.