As organizations navigate the complexities of 2026, the transition from experimental AI to scalable, production-grade systems has become the defining challenge for the modern enterprise. Dominic Jainy, a seasoned expert in machine learning and blockchain architecture, has spent his career at the intersection of data and infrastructure, helping global firms move beyond the hype of pilot projects. Today, the focus has shifted toward hybrid architectures and sovereign AI, particularly in the rapidly evolving Asia-Pacific market. In this conversation, we explore how the shift toward modular, containerized platforms is redefining data strategy, the hidden economic pressures of “always-on” generative AI, and the critical role of data sovereignty in a world where data is the most valuable asset. We examine the move away from monolithic platforms, the rising costs of token consumption, and the strategic deployment of applied AI teams to ensure technology drives real business outcomes.
Many organizations are moving away from monolithic platforms toward modular, containerized architectures. How does this shift, specifically through foundations like the Taikun acquisition, change the way a company can “write once and deploy anywhere” across different environments?
The transition away from monolithic architectures is a fundamental liberation for the modern enterprise, as it removes the rigid infrastructure dependencies that have historically stifled innovation. By integrating specialized containerization layers, such as those derived from the Taikun acquisition, we have created a foundation where services are no longer shackled to a specific cloud provider or a single piece of hardware. This modularity means that an organization can develop an AI model or a data pipeline once and then seamlessly transport it from a localized edge device to a massive public cloud or even back to a private data center. It is about bringing the AI to the data rather than the other way around, which is a massive shift in gravity for the industry. This approach ensures that whether the data sits in a localized hub or a sprawling global network, the governance and security layers remain consistent and unbroken.
In the JAPAC region, we are seeing a distinct approach compared to Western markets regarding cloud adoption. Why are large financial institutions in this part of the world specifically reconsidering their cloud-first strategies in favor of hybrid architectures?
The landscape in Asia-Pacific is uniquely shaped by a combination of conservative regulatory environments and the sheer, staggering scale of the customer bases being served. If you look at one of the largest Indian banks, for example, their subscriber count is comparable to the entire population of several mid-sized countries, which creates a data weight that is difficult to manage on public infrastructure alone. These institutions are rearchitecting their systems because they realize that an all-out cloud-first approach doesn’t always align with the strict data sovereignty and security requirements they face daily. They want the agility of the cloud for certain experimental workloads, but they are increasingly building “private AI factories” within their own data centers to maintain absolute control over their most sensitive operations. This hybrid strategy allows them to balance high-speed innovation with the bedrock stability and compliance that their regional regulators demand.
As organizations move from AI pilots into full-scale production, what are the specific economic “shocks” they are encountering, particularly regarding the costs of running continuous generative AI workloads?
The transition from a proof-of-concept to a production environment is often a moment of harsh financial realization for many CFOs who see the “spinning meter” of cloud costs suddenly accelerate. During a pilot, the speed and ease of public cloud are intoxicating, but when an AI application is switched to “always-on” mode, the token consumption and inference costs can quickly spiral out of control. We are seeing a significant trend of “repatriation” for certain workloads, where companies move their AI processing back into the data center to achieve a more predictable and manageable cost profile. This isn’t necessarily about abandoning the cloud, but rather about optimizing the economics of scale—using private infrastructure for heavy-duty, continuous inference while leveraging the cloud for burst capacity. By embedding specialized microservices like Nvidia NIM into the local infrastructure, companies can maintain high performance without the unpredictable tax of third-party cloud consumption.
Sovereignty is no longer just a legal checkbox regarding data residency. How is the concept of a “sovereign AI factory” evolving to meet the needs of governments and highly regulated industries that require absolute control?
Sovereignty in 2026 has evolved into a comprehensive strategy that covers not just where data resides, but who has the keys to the kingdom and how that data is exposed during the AI training and inference process. For government entities and highly regulated sectors, the risk of data exposure in a shared cloud environment is often too high, which is why we are seeing a massive investment in air-gapped environments. These “sovereign AI factories” are essentially fortified data centers where the entire AI lifecycle—from data ingestion to model deployment—happens within a closed, highly secure loop. It involves more than residency; it is about the security guardrails and the lineage of the data, ensuring that every piece of information used by an agentic AI system is accounted for and protected from external eyes. We currently manage about 30 exabytes of data across our global customer base, and a significant portion of that remains in these private, sovereign environments because that is where the true value and security lie.
The shift toward “Applied AI” suggests that technology alone is no longer enough to satisfy stakeholders. How do specialist teams, like Forward-Deployed Engineers, bridge the gap between technical potential and actual business outcomes?
Technology is merely the engine, but “Applied AI” is the navigation system that ensures that engine actually takes the business where it needs to go. We recognized that many organizations have the data and the compute power, but they struggle with the complex change management and workflow redesign required to integrate agentic AI into their daily operations. By deploying specialist teams like Forward-Deployed Engineers directly into the customer’s environment, we can help them map specific use cases—such as automated loan processing or predictive maintenance—directly into their existing processes. This isn’t just about pre-sales support; it is about embedding experts who understand the nuances of data lineage, governance, and the specific architecture of the platform to accelerate results. This hands-on approach reduces the friction of adoption and ensures that the massive investments being made in AI today translate into measurable returns on investment rather than just more experimental clutter.
What is your forecast for the evolution of hybrid data platforms as agentic AI becomes the primary driver of enterprise automation?
I forecast that the next several years will see a total convergence of data management and agentic execution, where the platform itself becomes an intelligent “control plane” that manages workloads across any environment automatically. We will move away from manual infrastructure decisions, as AI agents will determine where a workload should run based on real-time factors like cost, latency, and sovereignty requirements. The focus will shift entirely to the data—the 30 exabytes of proprietary information that companies have spent decades collecting—as this becomes the primary fuel for specialized, agent-driven workflows. Organizations that successfully implement a “true hybrid” data platform with unified governance will be the ones that dominate their respective markets, as they will have the flexibility to pivot their entire AI strategy without being held hostage by a single infrastructure provider. Ultimately, the winners will be those who treat their data as a sovereign asset while maintaining the modularity to deploy it anywhere the business demands.