Our Take
Capital is chasing capacity, not capability: India is becoming the factory floor for global AI compute, but the country still depends on U.S. technology and faces mounting pressure on power and water.
Why it matters
India's data center sector is attracting institutional money from tier-one investors, signaling confidence in the country as a stable, policy-friendly hub for AI workloads. But infrastructure alone does not close the AI model gap between India and the U.S.
Do this week
Infrastructure architects: audit your data residency and latency requirements now to determine whether India-based capacity can meet your SLA targets before capacity constraints force you into secondary markets.
Canada's largest pension fund backs Indian data center expansion
CPP Investments, managing roughly $20 billion in Indian assets as of March 2026, committed up to ₹70 billion ($741 million) to CtrlS Datacenters on Wednesday. The deal includes a ₹40 billion ($423 million) investment for an 8.2% equity stake and a ₹30 billion ($317 million) commitment to a joint venture focused on building hyperscale data center campuses tailored for AI workloads. CPP Investments will own 48% of the joint venture, CtrlS the remaining 52%.
CtrlS, founded in 2007, operates more than 15 data centers across India and has signaled plans to invest $2 billion over six years to expand capacity. The investment reflects accelerating foreign interest in Indian infrastructure: earlier this month, AirTrunk (backed by Blackstone) announced $30 billion in planned investment to build five gigawatts of capacity by 2030. Meta announced a partnership with Reliance Industries on a 168-megawatt AI data center in Gujarat last week.
New Delhi has incentivized foreign cloud infrastructure through tax exemptions on overseas-facing cloud services through 2047, provided workloads run from Indian data centers. Indian conglomerates including Adani Group and Tata Consultancy Services have unveiled competing data center projects to capitalize on the opportunity.
Scale, not sovereignty
India has become the primary destination for global AI compute buildout outside the United States. The reasons are straightforward: lower land and labor costs, regulatory support from New Delhi, and undersaturated power grids (relative to U.S. and European capacity constraints). CPP Investments' move signals that pension capital sees stable returns in the infrastructure layer.
But the investment exposes a structural asymmetry. India is building the factories. It is not building the intellectual property. While the country has a handful of startups developing indigenous AI models (Sarvam AI), the underlying technology stack remains U.S.-supplied. Amazon, Google, Microsoft, OpenAI, and Uber have all announced India investments in recent months, but those investments are predominantly in deploying and fine-tuning foreign models, not developing new capabilities. India's role in the AI value chain is production capacity, not innovation.
The rapid buildout also carries hidden costs. Hyperscale data centers consume vast quantities of electricity and fresh water. As capacity scales, competition for those resources will intensify, potentially creating bottlenecks that policy makers have not yet publicly addressed.
Evaluate India for latency-tolerant, regionally-isolated workloads
If your workload requires data residency in India or can tolerate 50-100ms cross-border latency, CtrlS and other emerging operators now offer viable alternatives to legacy on-premise infrastructure. Audit your SLA requirements against the operational maturity of Indian operators (power redundancy, disaster recovery, staff expertise) before committing to multi-year capacity agreements. If your workload requires sub-20ms latency to Indian users or demands tight integration with U.S.-based model serving, the cost of cross-border routing may outweigh the per-compute savings.