Data centres and cloud: Value chain, from upstream to the customer
Server halls and cloud providers renting compute, storage and networks; heavy power users
Value chain, from upstream to the customer
Criticality to the chain: 5 · Chokepoint 4 · Critical 3 · Important 2 · Supporting 1 · Commodity service Stage 1 Capital, land and powerRaise equity and debt, buy land with power access and develop powered shells or full campuses that are leased to clouds and colocation tenants.4
Why important: Capital is plentiful from many funds, but the scale is extreme (about US$ 173 billion committed in India since 2021 alone) and a pull-back would slow builds. Technologies Powered-land banking Build-to-suit leases Project finance and green bonds Site selection for power, water and fibre Depends on Industries Finance: banking, insurance, marketsReal estateExamples: Blackstone (QTS), Brookfield, KKR, AdaniConneXConnect campuses to the grid and sign long-term power purchase agreements, increasingly for dedicated solar, wind, gas or nuclear supply.1
Why chokepoint: Grid connection is now the binding constraint: the IEA projects data-centre use more than doubling to about 945 TWh by 2030, new substations take years, and there is no substitute for firm power. Technologies High-voltage substations Corporate renewable PPAs Round-the-clock hybrid renewables with storage Gas turbines and small modular reactors (planned) Depends on Commodities Cu CopperGas Natural gasU UraniumSi Silicon Energy HVAC AC transmissionPV Solar photovoltaicsWOn Onshore windCCGT Gas turbines and combined cycle Industries Electrical equipmentConstruction and infrastructureExamples: NTPC, Tata Power, Adani Green Energy, NextEra Energy, Constellation EnergyDesign and build the shells, electrical rooms, cooling plant and fit-out, often on fast-track contracts of 12 to 24 months.
Why supporting: Many general and specialist contractors compete, though skilled electricians and commissioning engineers are scarce in hot markets. Technologies Modular and prefabricated construction Building information modelling Raised-floor and slab designs for liquid cooling Commissioning and integrated systems testing Depends on Commodities Cem CementSt SteelCu CopperAl Aluminium Chemicals Con Construction chemicalsFR Flame retardants Energy Oil Crude oil Industries Construction and infrastructureBuilding materialsExamples: Larsen & Toubro, Turner Construction, DPR Construction, Sterling and WilsonSupply transformers, switchgear, UPS systems, backup generators, chillers and liquid-cooling units for each megawatt of IT load.5
Why critical: A few global firms supply most gear, and US transformer lead times have doubled or tripled since 2019 to as long as four years (Wood Mackenzie), so equipment shortages can delay campuses by a year or more. Technologies Uninterruptible power supplies with lithium batteries Medium-voltage switchgear and busways Diesel and gas backup generators Chillers and evaporative cooling Direct-to-chip and immersion liquid cooling Depends on Commodities Cu CopperSt SteelLi LithiumPb LeadAl Aluminium Chemicals HFC Fluorochemicals and refrigerantsMEG Monoethylene glycolPTFE Fluoropolymers (PTFE, PVDF) Energy PE Transformers and power electronicsHP Heat pumps, air conditioning and refrigerationICE Internal combustion engines and gensetsLFP Lithium iron phosphate batteries Industries Electrical equipmentMachinery and equipmentExamples: Schneider Electric, Vertiv, Eaton, ABB, CumminsDesign the GPUs, custom AI chips and server CPUs that are now the largest single cost of an AI data centre, sold to clouds and server makers.6
Why chokepoint: Nvidia booked US$ 193.7 billion of data-centre revenue in its year to January 2026, far ahead of any rival, and its software stack locks in users; nearly all leading AI chips are made at one foundry with scarce advanced packaging. Technologies GPUs with high-bandwidth memory Custom AI ASICs (TPU-type) Server CPUs (x86 and Arm) Scale-up interconnects (NVLink-type) AI software stacks (CUDA-type) Depends on Commodities Si SiliconCu Copper Chemicals ESG Electronic specialty gases Energy DC Data centres Industries SemiconductorsExamples: Nvidia, AMD, Broadcom, IntelBuild and integrate servers, storage arrays, switches and full AI racks, selling to clouds, colocation tenants and enterprises.
Why important: Several branded and Taiwanese contract makers compete, but rack integration capacity for the newest AI systems is tight. Technologies Rack-scale AI systems Liquid-cooled server trays Ethernet and InfiniBand switches NVMe flash storage Optical transceivers Depends on Commodities Cu CopperAu GoldAl AluminiumIn Indium Chemicals EPX Epoxy resinsFR Flame retardants Energy PE Transformers and power electronics Industries Computers and electronicsSemiconductorsExamples: Dell Technologies, Supermicro, Quanta, Arista Networks, Netweb TechnologiesLink campuses to each other, to internet exchanges and to subsea cable landings, selling dark fibre, wavelengths and cross-connects.
Why important: Major hubs have several carriers, but a campus without diverse fibre routes cannot sell to clouds or banks. Technologies Dark fibre and DWDM Internet exchanges and peering Subsea cable landing stations Data-centre interconnect optics Depends on Commodities Ge Germanium Energy Dist Distribution grids, mini-grids and interconnectors Industries TelecommunicationsExamples: Tata Communications, Bharti Airtel (Nxtra), Lumen Technologies, Sify TechnologiesRun secure, powered and cooled halls, leasing space and power by the rack or the megawatt to clouds, banks, telecom firms and enterprises.4
Why important: Many global and Indian operators compete in major hubs, but moving a deployed tenant is slow and costly. Technologies Tier III and Tier IV redundant design Data-centre infrastructure management software Physical security and compliance Water-efficient cooling Heat reuse Depends on Commodities H₂O Water Chemicals WT Water treatment chemicals Energy DC Data centresDist Distribution grids, mini-grids and interconnectorsHP Heat pumps, air conditioning and refrigerationICE Internal combustion engines and gensets Industries Water and sanitationReal estateElectrical equipmentExamples: Equinix, Digital Realty, NTT Global Data Centers, STT GDC India, CtrlSBuild and operate global fleets of data centres and rent compute, storage, databases and AI services by the hour to businesses, governments and AI developers.2,3
Why critical: Amazon (28%), Microsoft (21%) and Google (14%) took about 63% of cloud infrastructure revenue in Q4 2025 and migrating large workloads takes months to years, though Oracle, regional clouds and GPU clouds offer alternatives. Technologies Virtualisation and containers Custom server and AI chips Global software-defined networks Managed AI model services Sovereign and local cloud regions Depends on Energy DC Data centresPV Solar photovoltaicsNFR Nuclear fission reactors Industries Computers and electronicsSoftware and IT servicesTelecommunicationsConstruction and infrastructureExamples: Amazon Web Services, Microsoft Azure, Google Cloud, Oracle Cloud, Yotta Data Services · ≈ US$ 419 billion of cloud infrastructure services revenue in 20252