Verdict: RMZ's combined $40 billion commitment to Andhra Pradesh and Maharashtra — for gigawatt-scale AI data centers, Global Capability Centres and industrial logistics parks — is one of the largest private infrastructure bets on India as a global AI compute destination. It also shows why local cloud capacity, sovereign data hosting and hardware engineering talent are becoming competitive advantages for Indian businesses.
Last verified: 2026-06-17 · Key deals: $10B in Andhra Pradesh (5–6 years) and up to $30B in Mumbai region (10 years) · Expected jobs: ~100,000 in AP; ~300,000 in Maharashtra · Capacity target: up to 1 GW hyperscale cluster in Visakhapatnam · Pricing/land/tenure figures are volatile and re-checked at each official update.
What did RMZ actually commit to?
At the World Economic Forum Annual Meeting 2026 in Davos, RMZ signed two separate state-level investment partnerships in under three weeks:
| Deal | State | Amount | Horizon | Core projects | Source |
|---|---|---|---|---|---|
| Andhra Pradesh | Andhra Pradesh | Up to $10 billion | 5–6 years | 10M sq ft GCC campus at Kapuluppada; up to 1 GW hyperscale data center cluster near Visakhapatnam; 1,000-acre industrial/logistics park at Tekulodu in Rayalaseema | ET Edge Insights |
| Maharashtra | Maharashtra (MMRDA + CIDCO) | Up to $30 billion | 10 years | Data centers and commercial projects in Navi Mumbai; real estate, logistics and infrastructure across the Mumbai Metropolitan Region | Re-mumbai |
The $30 billion Maharashtra figure is a facilitation ceiling — RMZ will lead capital mobilisation, master planning and project development across multiple asset classes, not a single cheque. The Andhra Pradesh deal is more project-specific, with a 1 GW data center cluster and a 10 million sq ft Global Capability Centre (GCC) campus as the headline anchors.
RMZ already owns and operates roughly 70 million sq ft of real assets valued at over $20 billion across six Indian cities, according to company statements cited by media outlets covering the Davos announcements.
Why are hyperscalers suddenly converging on India?
India is no longer just a destination for software outsourcing. It is becoming a compute destination. Google's separate $15 billion AI hub in Visakhapatnam, announced in October 2025 and confirmed alongside state government partners AdaniConneX and Airtel, is being built in the same city RMZ is targeting.
The drivers are structural:
- Data localization and sovereignty. India's Digital Personal Data Protection Act and broader push for sovereign cloud are forcing global cloud providers to build in-country capacity. We covered the wider shift in our HCLTech-Sarvam sovereign AI analysis.
- Inference demand is now larger than training demand globally. That shifts the economics toward distributed, lower-latency capacity — exactly what India can offer for domestic and regional workloads.
- Power and construction scale. India has a fast-growing renewable energy base and a construction ecosystem that can deliver large campuses quickly.
- Talent depth. The country produces a deep pipeline of software engineers; the new challenge is hardware engineering, cooling specialists and data center operations talent.
Google's project alone is expected to eventually exceed 1 GW across three linked campuses near Visakhapatnam, with a subsea cable gateway and supporting clean energy infrastructure.
What makes an AI-native data center different?
Traditional office parks and legacy data centers were not built for liquid-cooled GPU racks pulling over 1 MW per rack. RMZ's framing — "phase 2.0 of the IT story" — centers on designing for a future where GPU generations refresh every few years, but the concrete and power last 25–30 years.
Key design shifts include:
- Liquid cooling and direct-to-chip heat dissipation to support NVIDIA B300, GB300, Rubin and future architectures.
- Future-proof white space so hyperscalers can swap GPU generations without tearing down buildings.
- Low PUE targets to cut both power bills and the environmental criticism that data centers are power guzzlers.
- On-site or co-located renewable power to avoid simply siphoning electricity from the wider grid.
- Local supply chains — RMZ says its Mumbai and Chennai data centers already source over 90% of equipment locally.
How does this create jobs beyond the data center?
The transcript repeatedly stresses a "pyramid" model: power at the base, data centers as the backbone, and AI solutions at the top. That pyramid creates employment across the stack:
| Layer | Jobs created |
|---|---|
| Power and renewables | Project finance, EPC, grid engineers, power traders |
| Construction and mechanical | Civil, cooling, electrical, fire suppression, commissioning |
| Hardware and supply chain | Server assembly, cabling, racks, storage, networking |
| Data center operations | Facilities managers, NOC technicians, security, compliance |
| AI solutions and GCCs | ML engineers, data engineers, cybersecurity, product managers |
| Enterprise services | Finance, HR, legal, sales around every new campus |
RMZ has also flagged hardware engineering talent as a bottleneck. The company says it is working with academic and training partners to grow that pool rather than simply poaching from rivals — a plausible claim, but one worth tracking as projects move from announcement to ground-breaking.
What does this mean for small businesses and builders?
For most small businesses, the headline is: Indian cloud and AI infrastructure is about to get more local, more competitive and more sovereign. The broader shift is part of why sovereign AI is becoming non-negotiable in 2026.
Practical implications:
- Lower latency, lower FX risk. Running inference inside India avoids cross-border data-transfer complexity and dollar-denominated egress surprises.
- More local AI services. As capacity comes online, expect Indian SaaS and cloud providers to offer AI add-ons — customer support, bookkeeping, content, analytics — at prices tuned to local markets.
- New vendor options. Sovereign and private-cloud deployments become realistic for regulated sectors: finance, healthcare, government contractors.
- Talent pressure and opportunity. Roles in cloud operations, MLOps, data engineering and cybersecurity will grow. Small businesses that upskill early will hire more cheaply than those that wait.
- Real estate and services spillover. Large campuses create demand for transport, food services, security, fit-out, maintenance and local logistics.
If you are building a product that relies on AI inference, the India compute layer matters. Keep an eye on which providers end up actually operating in these new campuses, because announced capacity is not live capacity. The same principle applies when you pick AI tools — see our best AI tools for small business breakdown.
How do the numbers compare globally?
| Project / market | Location | Approx. commitment | Notes | Source |
|---|---|---|---|---|
| RMZ + Andhra Pradesh | Visakhapatnam region | Up to $10B over 5–6 years | 1 GW data center cluster + 10M sq ft GCC + 1,000-acre industrial park | ET Edge Insights |
| RMZ + MMRDA/CIDCO | Mumbai Metropolitan Region | Up to $30B over 10 years | Multi-asset facilitation; data centers + commercial + logistics | Re-mumbai |
| Google + AdaniConneX + Airtel | Visakhapatnam | $15B | AI hub, subsea cable gateway, clean energy | Data Center Dynamics |
| Colt DCS + RMZ | Navi Mumbai, Chennai | $1.7B | 50:50 JV; ~250 MW across phases | Colt DCS press release |
| India data center market | India | ~$5.1B in 2024 → ~$13.8B by 2034 (10.5% CAGR) | Revenue estimate; installed base was ~1.64 GW in 2025 | Zion Market Research, Black Ridge Research |
The India data center market was valued at roughly $5.1 billion in 2024 and is projected to reach about $13.8 billion by 2034, with a CAGR around 10.5%, according to Zion Market Research. Black Ridge Research puts the 2025 installed base at about 1.64 GW. Those forecasts are always sensitive to power availability, policy continuity and global capex cycles, so treat them as directional.
What are the real risks?
Every infrastructure super-cycle has gap between announcement and operation. The main watch-points:
- Power. India is adding renewables fast, but a gigawatt-scale cluster needs firm, clean power and transmission. Some estimates suggest 13 GW of data center load could require ~40 GW of renewable build-out.
- Water and cooling. Liquid-cooled facilities use less water than older designs, but at scale even reduced intensity adds up — especially in coastal or drought-prone states.
- Policy continuity. 10-year commitments span multiple election cycles; land, environmental and power clearances can slow or redirect projects.
- Supply chain localisation. "90% sourced locally" often starts as assembly; true component manufacturing takes years to scale.
- Demand timing. If AI capex cycles slow or shift to smaller edge deployments, some announced campuses may be phased more slowly than planned.
What this means for you
If you run a small business or are building a product: treat Indian AI infrastructure like a new utility layer coming online. You do not need to move everything today, but you should ask your cloud or SaaS vendors where their India compute lives, what their sovereign-cloud options cost, and whether your data residency requirements are covered. Our AI for small business guide covers how to evaluate vendors without over-engineering.
If you are a builder or service provider: the spillover demand in cooling, power, security, fit-out, training and operations support will be real, but lumpy. Lock in relationships with operators actually breaking ground, not just announcing MoUs.
Related reading
- 2026 US chip manufacturing shift
- India becomes a capital exporter in UK AI infrastructure
- Sovereign AI Race 2026: Why Anthropic's Fable 5 Ban Backfire
FAQ
Q: Is RMZ investing $40 billion in cash right now? A: No. The $40 billion is a combined commitment ceiling across two state deals — up to $10 billion in Andhra Pradesh over 5–6 years and up to $30 billion in Maharashtra over 10 years. Actual deployed capital will depend on approvals, demand and phased execution.
Q: What is a 1 GW data center cluster in simple terms? A: One gigawatt is 1,000 megawatts of power capacity. A large hyperscale data center campus is typically 50–150 MW, so 1 GW is roughly several very large campuses — enough to host hundreds of thousands of AI accelerators.
Q: Why does Google want to be in the same city as RMZ? A: Visakhapatnam offers land, a port, state incentives and a government pushing a multi-gigawatt digital ecosystem. Multiple large players in one city create a talent, supplier and connectivity cluster rather than a single-company project.
Q: Will this bring down cloud prices in India? A: Probably, but gradually. More local supply increases competition and cuts cross-border costs. However, power, cooling and financing costs remain high, and providers will price based on utilization, not announcements.
Q: Should small businesses care about sovereign AI infrastructure? A: Yes, if you handle customer data, health data, financial data or government contracts. Hosting in India can simplify compliance with the Digital Personal Data Protection Act and reduce jurisdictional uncertainty.
Discussion
0 comments