As AI adoption accelerates across cloud, enterprise, data centre and connected device ecosystems, the conversation around infrastructure is shifting quickly. The real challenge is no longer just about building faster compute, but about creating the underlying semiconductor, networking, storage and security layers that can support data-intensive workloads at scale. That is where the role of purpose-built silicon is becoming increasingly important.
In this context, Naveen Bishnoi, Associate Vice President and India Country Manager at Marvell India Private Limited, points to a larger industry transformation. In his view, semiconductors are no longer just components inside a digital system. They are the foundation of the digital footprint itself, enabling data to be processed, moved, stored and protected across increasingly complex environments. His perspective reflects a broader shift in the industry, where infrastructure is now being designed around AI use cases, data movement efficiency and long-term scalability.
Purpose-built silicon is gaining ground
Bishnoi’s core argument is that generic, one-size-fits-all silicon can no longer meet the needs of modern workloads. AI, in particular, has created a wide range of use cases, from training to inference, and from enterprise deployments to hyperscaler-scale systems. Each of these requires different performance, power and cost characteristics, which is driving demand for custom ASICs and purpose-built semiconductor architectures.
He also highlights that the industry is entering a phase where performance alone is no longer the primary benchmark. Power consumption, total cost of ownership and system-level efficiency are now equally important. That is forcing semiconductor companies to design with a longer time horizon in mind, even as AI models and data requirements continue to evolve rapidly.
Networking and storage are the new pressure points
One of the most striking points in Bishnoi’s perspective is that data movement has become a major bottleneck in AI systems. Compute remains critical, but the ability to move data quickly and efficiently across large clusters is now just as important. This is why networking, including both electrical and optical technologies, is emerging as a central area of innovation.
Storage is also under pressure. AI systems need fast local access during processing, but they also need secure, scalable long-term storage. Bishnoi’s view is that future architectures will need to treat compute, network and storage as a tightly integrated whole rather than separate layers. That systems-level thinking will be essential if AI infrastructure is to keep pace with demand.
Security and India’s role in the ecosystem
Bishnoi also makes a strong case for building security into silicon from the start, rather than layering it on later. As AI systems become more distributed and more exposed, hardware-level trust, validation and supply chain security will be vital to keeping infrastructure resilient.
He also sees India moving into a more ambitious phase in the semiconductor journey, from consumer to creator. While manufacturing progress is visible, he believes the next step is to build a fuller ecosystem around design, IP creation, product development and system integration. In that sense, India’s semiconductor story is no longer just about capacity. It is about becoming a serious contributor to the global AI and digital infrastructure stack.
