# Data centres and cloud: Follow one offering through the chain Server halls and cloud providers renting compute, storage and networks; heavy power users Source page: https://www.materialinsights.org/industries/data_centres/story/ #### Follow one offering through the chain **4** Example story · 6 steps · India ### An AI cloud in a Navi Mumbai tower Follow Yotta's NM1 campus at Panvel from land, capital and power to Nvidia GPUs rented by the hour. 1 / 6 · Capital, land and power #### Capital and land at Panvel **$173bn**India data-centre commitments since 2021 A hyperscale campus starts with a developer that can raise equity and debt, secure land near power lines and fibre, and wait years for rent. Yotta built NM1 in the Hiranandani Fortune City park at Panvel, part of a data-centre park planned to scale to 1 GW. Across India, some US$ 173 billion of data-centre investment has been committed since 2021, CBRE estimates.^[2](https://yotta.com/data-center/)^,^[1](https://cxotoday.com/hardware/cbre-h1-2026-report-india-data-center-capacity-hits-1-75-gw-set-to-cross-2-gw-by-year-end/) **Constraint** Powered land near Mumbai's cable landings is scarce, and Mumbai took 94% of India's new capacity in H1 2026. [Cem Cement](https://www.materialinsights.org/commodities/cement/)[St Steel](https://www.materialinsights.org/commodities/steel/)[Finance: banking, insurance, markets](https://www.materialinsights.org/industries/finance/)[Real estate](https://www.materialinsights.org/industries/real_estate/) 2 / 6 · Capital, land and power #### Tens of megawatts, day and night **52 MW**design IT load at NM1 NM1 is designed for 52 MW of IT load at a power usage effectiveness below 1.5, so running flat out it could draw up to about 78 MW, or some 680 GWh a year (calculated upper bound; real use is lower). Utilities and renewable developers supply that through high-voltage connections and long-term power purchase agreements, without a break.^[2](https://yotta.com/data-center/)^,^[3](https://www.iea.org/reports/energy-and-ai/executive-summary) **Constraint** The IEA expects world data-centre power use to more than double, from about 415 TWh in 2024 to 945 TWh by 2030. [Cu Copper](https://www.materialinsights.org/commodities/copper/)[Tc Thermal coal](https://www.materialinsights.org/energy/thermal_coal/)[PV Solar photovoltaics](https://www.materialinsights.org/energy/e_pv/)[HVAC AC transmission](https://www.materialinsights.org/energy/e_hvac/)[WOn Onshore wind](https://www.materialinsights.org/energy/e_onshore_wind/)[Electrical equipment](https://www.materialinsights.org/industries/electrical_equipment/)[Construction and infrastructure](https://www.materialinsights.org/industries/construction/) 3 / 6 · Building and facility equipment #### Transformers, batteries and chillers **4 years**longest US transformer lead times Every megawatt of IT load needs transformers, switchgear, uninterruptible power supplies with batteries, diesel generators and chillers, from firms such as Schneider Electric, Vertiv and ABB. Yotta specifies redundant power feeds, generator and UPS backup and N+1 or N+2 chillers. Equipment is now a bottleneck: in the US, Wood Mackenzie found, large transformers can take up to four years to arrive.^[2](https://yotta.com/data-center/)^,^[4](https://www.eenews.net/articles/data-center-surge-is-driving-up-transformer-costs/) **Constraint** US transformer lead times have doubled or tripled since 2019 to as long as four years (Wood Mackenzie), delaying campuses. [Cu Copper](https://www.materialinsights.org/commodities/copper/)[St Steel](https://www.materialinsights.org/commodities/steel/)[Pb Lead](https://www.materialinsights.org/commodities/lead/)[Li Lithium](https://www.materialinsights.org/commodities/lithium/)[HFC Fluorochemicals and refrigerants](https://www.materialinsights.org/chemicals/fluoro_intermediates/)[Electrical equipment](https://www.materialinsights.org/industries/electrical_equipment/)[Machinery and equipment](https://www.materialinsights.org/industries/machinery/) 4 / 6 · IT hardware #### Thousands of Nvidia GPUs **16,000+**H100 GPUs in Shakti Cloud, 2026 The costliest items are the accelerators. In 2023 Yotta planned a first cluster of 16,384 Nvidia H100 GPUs at NM1; by 2026 Nvidia said Shakti Cloud had more than 16,000 H100s. Nvidia booked US$ 193.7 billion of data-centre revenue in its year to January 2026, and nearly all leading AI chips come from one foundry.^[5](https://yotta.com/?p=10535)^,^[9](https://www.nvidia.com/en-us/case-studies/yotta-built-india-sovereign-ai-infrastructure-shakti-cloud/)^,^[6](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000019/q4fy26pr.htm) **Constraint** One designer dominates, supply hangs on a single Taiwanese foundry, and US export licences govern advanced chips. [Si Silicon](https://www.materialinsights.org/commodities/silicon/)[Cu Copper](https://www.materialinsights.org/commodities/copper/)[Au Gold](https://www.materialinsights.org/commodities/gold/)[ESG Electronic specialty gases](https://www.materialinsights.org/chemicals/electronic_gases/)[Semiconductors](https://www.materialinsights.org/industries/semiconductors/)[Computers and electronics](https://www.materialinsights.org/industries/computers_electronics/) 5 / 6 · Facility operation #### A Tier IV tower of 16 floors **7,000+**rack capacity at NM1 Yotta operates NM1 itself, running halls that house its own GPUs and rack space leased to banks, telecom firms and other clouds. The building has 820,000 sq ft over 16 floors, room for more than 7,000 racks and Uptime Tier IV certification for fault tolerance. Mumbai, with its subsea cable landings, took 94% of India's new capacity in H1 2026.^[2](https://yotta.com/data-center/)^,^[1](https://cxotoday.com/hardware/cbre-h1-2026-report-india-data-center-capacity-hits-1-75-gw-set-to-cross-2-gw-by-year-end/) **Constraint** A PUE near 1.5 means up to half a unit of extra power for cooling and losses per unit of IT load (calculated). [H₂O Water](https://www.materialinsights.org/commodities/water/)[WT Water treatment chemicals](https://www.materialinsights.org/chemicals/water_treatment/)[HP Heat pumps, air conditioning and refrigeration](https://www.materialinsights.org/energy/e_heat_pumps/)[ICE Internal combustion engines and gensets](https://www.materialinsights.org/energy/e_engines/)[Water and sanitation](https://www.materialinsights.org/industries/water_sanitation/)[Real estate](https://www.materialinsights.org/industries/real_estate/)[Electrical equipment](https://www.materialinsights.org/industries/electrical_equipment/) 6 / 6 · Cloud services #### GPUs rented by the hour **38,000+**GPUs on IndiaAI compute portal Finally the capacity is sold as cloud services. Yotta's Shakti Cloud rents GPU time to start-ups, researchers and government projects, and the IndiaAI Mission, with a ₹10,372 crore outlay, had onboarded more than 38,000 subsidised GPUs from Indian providers by March 2026. Worldwide, Amazon, Microsoft and Google took about 63% (calculated) of cloud infrastructure revenue in Q4 2025.^[7](https://telecomtalk.info/?p=1005751)^,^[8](https://www.techinsights.com/ja/node/61571) **Constraint** Three US clouds take about 63% of cloud revenue; Indian GPU clouds lean on subsidies and imported chips. [DC Data centres](https://www.materialinsights.org/energy/e_data_centres/)[PV Solar photovoltaics](https://www.materialinsights.org/energy/e_pv/)[Dist Distribution grids, mini-grids and interconnectors](https://www.materialinsights.org/energy/e_distribution/)[Computers and electronics](https://www.materialinsights.org/industries/computers_electronics/)[Software and IT services](https://www.materialinsights.org/industries/software_it/)[Telecommunications](https://www.materialinsights.org/industries/telecom/) #### What the chain shows India has the land, demand and capital for AI data centres; firm power, electrical gear and the GPUs themselves are the binding constraints. Sources 1. [CBRE H1 2026 report: India data centre capacity hits 1.75 GW, set to cross 2 GW by year-end](https://cxotoday.com/hardware/cbre-h1-2026-report-india-data-center-capacity-hits-1-75-gw-set-to-cross-2-gw-by-year-end/) CXOToday (CBRE data), 2026 2. [Yotta data centres: NM1 Navi Mumbai](https://yotta.com/data-center/) Yotta Data Services, 2026 3. [Energy and AI: executive summary](https://www.iea.org/reports/energy-and-ai/executive-summary) International Energy Agency, 2025 4. [Data center surge is driving up transformer costs (Wood Mackenzie data)](https://www.eenews.net/articles/data-center-surge-is-driving-up-transformer-costs/) E&E News, 2025 5. [Yotta Data Services collaborates with NVIDIA to catalyze India's AI transformation](https://yotta.com/?p=10535) Yotta Data Services, 2023 6. [NVIDIA announces financial results for fourth quarter and fiscal 2026 (Form 8-K exhibit)](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000019/q4fy26pr.htm) NVIDIA via US SEC, 2026 7. [Over 38,000 GPUs onboarded on AI compute portal under IndiaAI Mission (MeitY statement)](https://telecomtalk.info/?p=1005751) TelecomTalk, 2026 8. [GenAI helps drive quarterly cloud revenues to $119 billion as growth rate jumped yet again in Q4](https://www.techinsights.com/ja/node/61571) Synergy Research Group (TechInsights), 2026 9. [Yotta built India's sovereign AI infrastructure with Shakti Cloud (customer story)](https://www.nvidia.com/en-us/case-studies/yotta-built-india-sovereign-ai-infrastructure-shakti-cloud/) NVIDIA, 2026 ## Other sections - [Overview](https://www.materialinsights.org/industries/data_centres/index.md) - [Where ₹100 goes](https://www.materialinsights.org/industries/data_centres/cost-mix/index.md) - [Price links with suppliers](https://www.materialinsights.org/industries/data_centres/price-links/index.md) - [Linked industries](https://www.materialinsights.org/industries/data_centres/linked-industries/index.md) - [Value chain](https://www.materialinsights.org/industries/data_centres/value-chain/index.md) - [Materials drawn on](https://www.materialinsights.org/industries/data_centres/materials/index.md) - [Chokepoints and risks](https://www.materialinsights.org/industries/data_centres/risks/index.md) - [Sources](https://www.materialinsights.org/industries/data_centres/sources/index.md)