Reimagining data centre infrastructure with AI

We live in an era of unprecedented data-driven transformation, with Artificial Intelligence (AI) leading the charge as one of the most groundbreaking innovations of our time. In just two years, 200 million ChatGPT users are generating a staggering 1.6 billion AI queries each month—and by 2025, this number is expected to surpass a billion users.

What once seemed futuristic has quickly become part of everyday conversations. The rise of AI and the semiconductor boom that is fuelling it have taken centre stage in both tech circles and living rooms. And now, here the world has in its hand GPT-4o—an AI capable of seamlessly interacting through audio, video, and text, all in real-time.

As one looks toward the 2030s, it is easy to imagine humanoid chatbots taking shape, powered by AI systems that communicate as naturally as humans. The future of AI is not just arriving—it is unfolding at lightning speed.

This technological revolution is driven not only by AI models but also by the critical infrastructure that supports them, such as AI-powered data centres, which rely heavily on optical fibre connectivity. As these data centres transition to GPU-based servers, the demand for high-performance fibre optic infrastructure, both within and between the centres, will significantly increase.

By 2025, India’s fibre infrastructure will empower even remote areas, enabling an AI-driven economy across industries like healthcare and manufacturing

Over the last decade, data centres have undergone constant advancements. They have become larger, accommodating more power-demanding, high-density servers and improving efficiency and sustainability.

Accelerated Compute: Driving Digital Society

As AI-driven applications and workloads expand, data centres must evolve significantly to support the growing computational power and storage demand. Traditional data centre infrastructure, which primarily relied on general-purpose servers, will be replaced by highly specialised hardware designed to handle the parallel processing requirements of AI workloads. These changes will include integrating more powerful processors and adopting AI-specific technologies such as accelerated compute and custom-built chips, capable of processing vast amounts of data at high speeds.

By deploying accelerated compute, data centres will be able to run complex AI algorithms faster, reduce latency, and support a more comprehensive array of AI-driven applications. This shift is already underway, and select market players, with their strong focus on future-proof technology infrastructure, are positioned to support the widespread adoption of accelerated compute-enabled data centres in India and beyond.

Optical Fibre: Enabling High-Performance AI Workloads

With the rise of AI applications, the demand for faster, more reliable data transmission is reaching new heights, making optical fibre a critical component of next-generation data centre infrastructure. As the demand for AI computing increases, optical fibre giants are driving the development of advanced fibre optic networks that will form the backbone of these next-gen data centres, ensuring they can manage the high-speed, data-heavy workloads associated with AI technologies.

India has become one of the biggest hubs for data centre ecosystems. As per initial estimates, the country is expected to boost its GPU-based server capacity in these AI-led data centres to nearly 5.2 lakh GPUs by 2026. These GPU-heavy data centres will require 36x more fibre than CPU racks, driven by increased server density and higher bandwidth requirements. It will also require more compact, high-density optical fibre cables (~70% more fibre) than traditional data centres.

The increasing demand for AI capabilities will push India to accelerate the deployment of fibre optic infrastructure. With AI data centres requiring robust, low-latency connectivity, expanding fibre networks to Tier-2 and Tier-3 cities will be a priority.

Fibre optic cables will form the critical foundation for the AI-driven economy, supporting industries ranging from healthcare and retail to manufacturing and autonomous transportation. By 2025, India’s fibre infrastructure will be instrumental in democratising access to AI technologies, empowering businesses and individuals in even the most remote parts of the country.

No wonder companies are developing sophisticated and integrated product portfolios to cater to the unique cabling needs of GPU-dense data centres. Optical fibre cabling in AI data centres must also deliver high bandwidth, low latency, and high density to meet the volume and complexity of AI workloads.

AI-led data centres in India will house nearly 5.2 lakh GPUs by 2026, demanding 36x more fibre than traditional setups to support evolving technologies.

As India prepares for 2025 and beyond, integrating AI with fibre technology will become critical to meeting the demands of today’s digital world and shaping the intelligent, interconnected ecosystems of the future. By prioritising innovation and adaptability, the country can lay the foundation for a truly connected and data-driven global economy.

By Dr Badri Gomatam

The author is the Group Chief Technology Officer at STL.

feedbackvnd@cybermedia.co.in