Most enterprise AI strategies are built on borrowed ground. Sergio Gago at Cloudera argues that organizations racing to deploy generative AI are optimizing the wrong layer, treating models as the whole product while their data, architecture, and governance sit on proprietary infrastructure someone else controls. With IDC projecting AI investment across Asia Pacific<\/a> hitting USD 175 billion by 2028, the foundation decisions made now will determine who actually owns the capability they think they’re building.<\/p>\n The organizations most exposed here aren’t the ones experimenting with AI, they’re the ones who’ve moved it into production without asking a deceptively simple question: can we keep running if our primary vendor changes its terms? Regulated industries, public sector bodies, and any enterprise with sensitive proprietary data sitting inside a hyperscaler’s managed AI service are already living this risk, often without realizing it. The IDC stat that only 7% of Asia Pacific enterprises are genuinely prepared on governance, risk, and compliance isn’t a lagging-market story. It’s a leading indicator of how many organizations have rented a capability they’ve mistaken for an asset.<\/p>\n The three-pillar framing Gago offers, interoperability, data sovereignty, and private AI, holds up analytically, even accounting for the fact that Cloudera sells hybrid data platforms and has a commercial interest in making proprietary clouds look fragile. The tilt worth watching is the implicit suggestion that open source solves the control problem by itself. It doesn’t. Open source models running on proprietary orchestration layers, closed APIs, or vendor-managed compute are still locked systems. The actual freedom condition is open formats all the way down, not just at the model layer. Apache Iceberg as a table format, open inference APIs, portable workload execution – these are the specifics that matter, and they require architectural discipline that most enterprises haven’t yet applied.<\/p>\n The decision this reframes isn’t a future procurement call. It’s the renewal sitting on the desk right now. Every managed AI service contract that doesn’t include data portability guarantees, audit rights, and workload migration paths is a sovereignty problem being deferred, not avoided. The CDO who treats that renewal as a routine cost negotiation rather than an architecture commitment is the one whose organization will discover, at the worst possible moment, that it bought permission rather than capability.<\/p>\n Data sovereignty means an organization retains legal and operational control over where its data lives, who can access it, and under what rules. It exists because data stored in another jurisdiction or inside a vendor’s proprietary system can become inaccessible through regulatory change, geopolitical friction, or a unilateral pricing decision. Think of it as the difference between owning a building and signing a long-term lease with a landlord who writes the maintenance rules. For AI systems built on proprietary data, sovereignty determines whether the system survives a vendor relationship ending.<\/p>\nWhat this means for your business<\/h2>\n
Concept deep-dive: Data sovereignty<\/h2>\n