Data Center / AI Infrastructure

The compute layer - racks of servers and GPU/accelerator nodes that run the AI and cloud workloads a data centre exists to serve.

This is where the economic value of a data centre is realised - everything else (power, cooling, building) exists to keep this layer running.

Hardware distributors and system integrators capture margin on every refresh cycle; the chips themselves are mostly imported, so India's exposure is concentrated in distribution, assembly and HPC/AI server integration rather than silicon.

• AI training/inference demand
• Cloud and enterprise workload migration
• Hyperscaler capex cycles
• Domestic HPC/AI server assembly incentives

• Global GPU/accelerator supply and allocation, largely outside India's control
• Import dependence on chips and high-end components
• Power and cooling capacity gating how much compute a site can actually host

• Order book / backlog growth
• Revenue per server shipped
• Gross margin trend (distribution vs. integration mix)