As geopolitical tensions, cyber threats and data sovereignty concerns reshape enterprise technology strategies, organizations are increasingly looking for AI infrastructure that combines intelligence with resilience. In this conversation with Tech Achieve Media (TAM), Angad Ahluwalia, Chief Operating Officer at Arinox, discusses why traditional centralized architectures are becoming vulnerable, the growing importance of sovereign AI, and how solutions such as air-gapped, on-premises AI systems can help enterprises retain control over their data and decision-making capabilities. He also shares his perspective on India’s AI infrastructure roadmap, public-private collaboration, and the strategic investments needed to achieve long-term technological independence.
Also read: Geopolitical Crisis Playbooks: Why IT Must Plan for Location-Based Disruptions
TAM: With rising geopolitical tensions exposing centralized data centres as strategic risks, how does CommandCORE redefine enterprise security architecture?
Angad Ahluwalia: When data centers were damaged in UAE and Bahrain, banking apps went dark, payment rails froze and many organizations lost their decision-making capability. This is the kind of vulnerability that Arinox solves. We believe in bringing AI to your data, not your data to the AI. Existing systems need sensitive data to move outside the four walls of the organization for processing, creating network dependency, and jurisdictional risk, like in the recent conflict.
Arinox’s CommandCore flips this model. It is a fully air-gapped, private AI system that runs entirely on-prem, bringing the intelligence layer inside the enterprise. When external pressure peaks, CommandCore keeps running. The entire security architecture from trust to the network is solved because data never leaves the system, and more importantly, decision-making capability stays intact. That’s what defines resilient security architecture. It gives businesses a fundamentally different posture, especially for critical infrastructure organizations.
TAM: How should governments rethink data sovereignty in an era of cross-border cyber threats and infrastructure dependencies?
Angad Ahluwalia: Governments have long understood that controlling physical infrastructure is fundamental to national security. Data sovereignty follows the same logic and it’s a great first step towards cyber defence. But in today’s threat environment, true sovereignty comes not just from where your data is located, but also with how your data is processed, interpreted and acted upon at every layer of the AI stack.
True sovereignty isn’t just about where data sits, it’s about who controls the AI systems that operate on it. For government and defence, that control is non-negotiable. For regulated enterprises in banking, healthcare, and manufacturing, it is fast becoming equally important, which is why the focus must shift to deploying self-contained AI infrastructure that operates within the four walls of the business.
TAM: What does it practically mean to “shrink” a data centre, and how does this shift impact latency, cost efficiency, and operational control?
Angad Ahluwalia: Traditional data centers offer proven reliability, deep customization and ease of doing business. These are benefits many businesses rely on. Shrinking the data center, and localizing it, is not challenging this existing model but offering a complementary path for organisations that need enterprise-grade AI capability without compromising on security. Already large enterprises in banking, healthcare, manufacturing etc house on-prem data centers. Arinox’s Sovereign AI-In-A-Box, coupled with high performance computing, provides the agentic layer that ensures data and intelligence doesn’t leave the four walls of the organization.
What it does is condense advanced compute and an agentic operating system into a compact plug-and-play appliance that can be deployed on-prem or at the edge. For regulated industries this ensures data stays within the defined boundaries and faster deployment with lower overhead. Latency drops , costs are predictable too because capacity is scaled with modular micro-deployments rather than large upfront builds. This also delivers full operational control as AI-In-A-Box or in-house AI Factories can be air-gapped and work within the perimeter an organization defines.
TAM: How can India design AI infrastructure that aligns with its scale, diversity, and regulatory landscape instead of mirroring Western or Chinese models?
Angad Ahluwalia: The right approach for India is a hybrid, distributed architecture. For non-sensitive workloads, cloud-based AI provides speed and lower entry friction. But for sensitive sectors like defense, banking, healthcare, and governance, organizations need to own their intelligence stack, including compute, models, and orchestration. Scale comes from standardizing the hardware and software stack nationally.
This can be done using open-source models deployed within controlled infrastructure.
India’s own models will take time to mature. In the meantime, the practical approach is to use global models, air-gapped where needed, build on best-in-class compute but run them within sovereign environments while implementing strong application-layer intelligence locally.
TAM: How can leaders foster collaboration between government, startups, and enterprises to accelerate sovereign AI adoption?
Angad Ahluwalia: The honest answer is that governments and businesses must shift from pilots to implementation. Right now, the government sets the policy, enterprises wait for procurement frameworks, and startups pitch into a slow-moving cycle. Ongoing geopolitical conflicts have accelerated the timeline for everyone. At a government level, each ministry, be it Union or State-level, must establish a Center of Excellence for AI Operations, which is an evolution of the traditional CTO function.
This is where AI agentic systems are built, infra is housed, and employees are empowered with intelligence systems. Startups need to bring working systems into these environments, and enterprises and governments need to integrate them into real operations instead of isolated test cases. That’s how adoption accelerates and India can move from evaluation to execution.
TAM: What bold bets should India’s leaders be willing to take today to secure long-term technological independence?
Angad Ahluwalia: Currently a lot of investment is going into building the foundational models. The first big bet will be investing in building the AI application layer across industries. While foundational models are important, real-world value comes from the application layer where AI interacts with users, integrates with systems and drives outcomes. This means standardizing infrastructure that speeds up deployment of agents and workflows where intelligence is embedded directly into operations.
The second bet is investing in inference sovereignty. This means running AI workloads in controlled environments, while simultaneously building long-term capability in models without slowing adoption. The truly promising prospect is that India has the talent and the data diversity to build AI applications that actually understand the linguistic, legal, and cultural context of Indian institutions. Betting on that now is what would enable strategic independence. Technological independence won’t come from models alone but it will come from how effectively AI is operationalized across India’s economy.
