Presented by Google Cloud

For the last two decades, cloud has been the cornerstone of modernization for nearly every industry allowing enterprises to hand off infrastructure management, in order to focus on business innovation. But some data was never going to move. Regulatory, operational and business requirements have long kept certain workloads out of the public cloud, and enterprises need access to modern data technologies wherever their data lives.

AI is now making that need more urgent. As enterprises look to apply AI to their data, they need the same best-of-breed capabilities in their own data centers that they'd get in the cloud.

AI agents need intelligent access to data wherever it resides. That makes it increasingly important to bring modern AI and data capabilities across cloud, edge and private data centers rather than requiring every workload and dataset to move toward a single destination. That reality is already showing up in enterprise AI planning. An IDC survey of more than 1,300 AI decision-makers found that while respondents expected the majority of AI workloads to reside in the cloud, some would remain on-premises, particularly for systems interacting directly with the physical world. This is why today’s modernization effort is increasingly described in terms of “borderless” data architectures. It’s about activating your data wherever it lives. That reflects the reality of most enterprise data estates. Data already spans multiple clouds, on-premises systems and decades-old platforms. The emerging architectural challenge isn't to centralize all of it, but to give agents unified access to the data and metadata they need, across all those environments.

Business continuity now requires data continuity

Business continuity used to be simple: keep the app online and transactions moving. Public clouds have become extraordinarily good at providing the availability and multi-region resilience needed to meet those requirements for most enterprises. But large utilities, healthcare companies, banks and manufacturers have workloads that need to run no matter what happens, and they can’t tolerate a dependency on any single provider for any part of their mission-critical business.

Applications increasingly need both real-time and historical data to answer critical business questions, which has pushed enterprises to bring together operational and analytical systems that were long kept separate. AI agents raise the stakes, since they need both at once to reason and act. Keeping an application running isn't enough if it loses access to the data it depends on.

Consider a "stressed exit" scenario. A contractual issue or provider outage could force an enterprise to move a critical workload quickly. What happens if you have to make that decision overnight while maintaining continuity for your customers?

This changes what a successful fallback looks like. It isn't enough to restart an application somewhere else. Applications need to retain a unified view across their real-time and historical data, even as the underlying infrastructure changes. For mission-critical workloads, that can require a full replica running on-premises or with a secondary provider, with the same database implementation available in each environment. The goal is continuity without rebuilding the data layer in the middle of a crisis.

Agentic AI needs a ‘write once, run anywhere’ data layer upgrade

Enterprises rely on cloud providers for the heavy lifting of infrastructure management. But running across more environments shouldn't mean re-architecting for every site. Forcing developers to juggle different database engines across cloud and on-premises environments creates friction that slows agents down. Different implementations, dialects and toolchains force teams to spend time managing platform differences rather than business logic.

Developers should instead be able to design, test and debug using the same database kernel, transactional semantics and capabilities such as in-engine graph, full-text and vector search across cloud, private data centers and other environments.

That consistency also addresses a common objection to a borderless data architecture -- operational sprawl. Instead of adding a new management stack for every environment, teams can package this technology for infrastructure that enterprises already use, including Kubernetes, VMs and bare metal, while embracing open standards such as PostgreSQL.

Open standards give developers portability beyond any one provider. That means their investments can be used even outside a vendor's ecosystem, while maintaining a consistent technology stack and unified CI/CD pipelines. This simplifies toolchains, speeds up compliance reviews, and accelerates time-to-market across hybrid, sovereign, and multi-cloud environments.

Bring cloud-scale intelligence to data that can’t move

The unified view that continuity depends on is also what makes agentic reasoning work. Agents raise the bar further, often needing operational and analytical data together in a single transaction, regardless of location.

Those requirements depend on native database primitives like graph structures, vector indexing and full-text search. Graph capabilities, for example, allow agents to understand relationships between transactions, products and users. Bringing these capabilities together lets agents reason over fresh operational data without moving it among separate systems. New database engines can deliver analytical queries up to 100 times faster than standard PostgreSQL.

But these capabilities can’t always live wherever infrastructure teams prefer. Some agentic workloads have extreme latency sensitivity or must operate within geographic boundaries for regulatory reasons. For example, DORA, one of the EU’s regulations for the financial sector, can require financial workloads to remain within sovereign environments. Healthcare presents an even clearer example; electronic health records and patient telemetry may be unable to leave local clinical boundaries, requiring predictive diagnostics and localized AI inference to run within secure firewalls.

Modernization means delivering these same capabilities straight to where the workloads run.

Database intelligence is breaking free from the infrastructure beneath it

Just as virtualization decoupled applications from physical servers, database intelligence is now breaking free from underlying infrastructure.

Software now handles capabilities once locked to specialized infrastructure, like clock synchronization for global consistency and the storage layer itself. The result is cloud-scale database capabilities that run consistently across cloud, on-premises, edge and sovereign environments.

Decoupling database intelligence from physical location redefines modernization. Technology leaders can now bring cloud capabilities directly to where their data lives, rather than forcing workloads to move.

Explore how Google Cloud databases meet your data across on-premises, edge, and multi-cloud environments with in-engine vector search and grounded agentic AI.

Raj Pai is VP of Product Management, Cloud Databases at Google Cloud.

Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact sales@venturebeat.com.