Deloitte expects most future investments to shift towards AI-ready graphics processing unit (GPU) infrastructure, with financing increasingly moving towards an infrastructure-style debt model. Gupta said the industry could see debt financing rise from the current 55-60% to nearly 75% as the sector matures.
Gupta said, “I would expect this $100 billion to be funded,” citing strong interest from both domestic and international investors.
Kumar said AI adoption will reshape the composition of data centres, with GPU-based infrastructure replacing traditional CPU-based workloads over time.
“Today, what is happening is 30% of that is towards the GPU cloud, and 70% is towards the CPU cloud. Over the next five years, we expect that to reverse because almost all the workloads… will be AI workloads,” he said.
He explained that GPU data centres are significantly more expensive, with GPU cloud infrastructure costing $30-35 million per megawatt, compared with around $12 million per megawatt for CPU cloud infrastructure, excluding civil construction and mechanical, electrical and plumbing (MEP) costs.
Gupta said India’s competitive construction costs, lower energy costs, infrastructure status for data centres, and supportive government policies have made the country an attractive destination for investment. Land availability, power, water, fibre connectivity and state-level policy support remain key factors in deciding where new facilities are built.
Kumar added that future data centre development will increasingly be integrated with renewable power infrastructure, with developers seeking campuses that combine data centres, power generation and connectivity through undersea cable networks.
He also identified storage memory, rather than GPUs, as the emerging supply-chain constraint and said improvements in liquid and immersion cooling technologies could reduce water consumption over time, although water remains a key challenge for the industry.
For the full interview, watch the accompanying video
