India’s AI Data Centre Boom: Why Secondary Fluid Networks Are Becoming Critical to High-Density Cooling

India’s data centre landscape is entering a new phase. The country’s growing AI ambitions are driving demand for computing infrastructure at a scale that traditional data centre designs were not built to handle.

The numbers underline the shift. According to a recent report, the Ministry of Power expects AI data centres to add 26.3 GW of electricity load by 2031-32. That estimate is almost twice the earlier projection of 13.56 GW.

More computing power means more heat. That makes cooling a central engineering challenge.

AI is changing the cooling equation

Conventional data centres have largely depended on air cooling. It has worked well for standard enterprise workloads where rack power densities remain within manageable limits. AI servers are different.

Modern AI systems rely on clusters of powerful GPUs and accelerators that generate substantially more heat in a smaller physical footprint. The International Energy Agency’s 4E Data Centre Energy Efficiency report notes that rising heat density from AI workloads is a primary driver behind liquid cooling adoption. It identifies around 20 kW of rack power as a general point where air cooling can become inadequate, while AI applications can already operate well above that level.

At these densities, simply moving more air through a server room becomes increasingly difficult and energy intensive. Liquid can transfer heat far more effectively, allowing cooling to happen closer to the source.

India is already moving in this direction. The Ministry of Electronics and Information Technology has highlighted direct-to-chip liquid cooling, adiabatic cooling and immersion cooling as technologies being adopted to reduce water usage.

Where Secondary Fluid Networks fit in

The move to liquid cooling changes more than the server itself. It changes the infrastructure around the rack. This is where the Secondary Fluid Network, or SFN, becomes important.

An SFN distributes coolant between cooling distribution units and the equipment that ultimately removes heat from high-performance servers. It can include precision piping, flexible metallic hoses, fittings, expansion joints and other components designed to manage fluid flow safely across the facility.

As rack densities rise, these networks need to deliver consistent flow while accommodating thermal expansion, vibration and installation constraints. Reliability matters because a cooling failure can quickly become a compute failure.

The engineering requirements are also becoming more demanding. Liquid cooling infrastructure must support leak prevention, pressure control, maintainability and long service life while fitting into increasingly dense data centre environments. Recent industry reporting has identified SFN as a critical layer of modern liquid cooling infrastructure, particularly as AI and high-performance computing accelerate the adoption of high-density systems.

India’s next cooling challenge

The scale of India’s AI expansion makes these considerations increasingly relevant. Data

centre capacity grew from about 375 MW in 2020 to around 1.5 GW in 2025.

That growth will put pressure on power infrastructure and cooling systems at the same time. For developers, the lesson is straightforward: cooling can no longer be treated as an isolated mechanical system. It needs to be considered alongside power density, rack architecture, water strategy and long-term scalability.

The AI boom may be powered by processors but keeping those processors running will increasingly depend on what happens through the pipes around them.

Comfonomics brings this approach to life through its Secondary Fluid Network (SFN) capabilities, designed for the demands of high-density liquid cooling. Its solutions support efficient coolant distribution while addressing flexibility, thermal expansion and system reliability. As AI racks continue to push heat densities higher, robust SFN infrastructure can help data centres scale cooling capacity without compromising performance.