{"slug": "microsofts-zero-water-promise-can-indias-ai-boom-beat-the-heat", "title": "Microsoft’s ‘Zero-Water’ Promise: Can India’s AI Boom Beat the Heat?", "summary": "Microsoft president of India and South Asia Puneet Chandok said at the Microsoft AI Infrastructure Summit India on September 21 that the company's new India South Central cloud region in Hyderabad will use effectively zero water for cooling, describing it as a design choice rather than a measured result. Chandok said the region will be Microsoft's fourth in India alongside Pune, Chennai and Mumbai, with three availability zones supporting AI workloads, and that Microsoft has contracted 1 gigawatt of renewable energy capacity in India with 630 megawatts already flowing toward its 2030 carbon negative, water positive and zero waste goals. Microsoft's closed-loop liquid cooling design recirculates water between servers and chillers instead of evaporating it, though the company acknowledges the switch to mechanical cooling raises energy overhead and that kitchens and restrooms still use water.", "body_md": "# Microsoft’s ‘Zero-Water’ Promise: Can India’s AI Boom Beat the Heat?\n\nThe cloud has a plumbing problem.\n\nBehind every AI answer sits computing hardware that generates heat. Removing that heat can require water, electricity, or a carefully engineered combination of both. As India builds more data centres, the infrastructure behind a seemingly weightless digital economy is becoming harder to ignore.\n\nMicrosoft now says its new India South Central cloud region in Hyderabad will use effectively zero water for cooling.\n\n“We’re not needing water in the first place by design,” Puneet Chandok, president of Microsoft India and South Asia, reportedly said at the Microsoft AI Infrastructure Summit India on September 21.\n\nIt is a consequential promise. It also needs to be read with its last two words attached: for cooling.\n\n## Rise of the Pivotal Power\n\n18 Sep 2026 - Vol 05 | Issue 38\n\nThe BRICS declaration is a strategic victory for Modi\n\n[Read Now Rise of the Pivotal Power](https://openthemagazine.com/magazine/rise-of-the-pivotal-power)\n\nThe announcement concerns a specific operational demand. It does not establish that the entire cloud region, its electricity supply and the factories making its hardware have no water footprint.\n\n## What is Microsoft promising in Hyderabad?\n\nChandok described India South Central as Microsoft’s fourth Indian cloud region, alongside Pune, Chennai and Mumbai. It will have three availability zones and support AI workloads from the outset.\n\nHe also said Microsoft had contracted 1 gigawatt of renewable energy capacity in India, with 630 megawatts already “flowing”, and reiterated the company’s 2030 ambitions to become carbon negative, water positive and zero waste.\n\nThose are different commitments with different tests. A cooling design addresses water consumed during operation. Renewable procurement addresses electricity sourcing. Water replenishment concerns benefits delivered through water projects. None automatically proves the others.\n\nThe ANI account does not provide Hyderabad-specific operating measurements or the complete cooling specification. Its central claim is a statement of design intent, rather than a published record of performance through an Indian summer.\n\n## How can a cooling system contain water and consume almost none?\n\nThink of a car radiator: liquid carries heat away and circulates again. Microsoft’s published engineering work describes cold plates that transfer heat directly from chips into recirculating coolant. Cooling close to the heat source offers more precise temperature control, particularly for powerful AI hardware.\n\nIts newer design fills a closed water loop during construction and recirculates that water between servers and chillers, avoiding the evaporation used by conventional cooling systems. Microsoft explicitly says kitchens and restrooms still require water.\n\nThere are two jobs here: moving heat off a chip and releasing it from the building. Liquid cooling at the chip does not, by itself, tell you how much water the whole facility consumes. A system can circulate liquid around servers yet still use evaporation elsewhere to reject heat. Research by Pengfei Li and colleagues distinguishes these two cooling stages.\n\nThat is why Hyderabad’s complete design matters. Microsoft’s global engineering descriptions explain how its promise can work; they should not be treated as a disclosed blueprint for every component of this particular region.\n\n## Does saving water mean using more electricity?\n\nIt can. Microsoft acknowledges that replacing evaporative cooling with mechanical cooling increases energy overhead. It says warmer operating temperatures and efficient chillers mitigate the effect, describing the annual energy increase across the relevant designs as modest. That is a company assessment, not a published Hyderabad-specific result.\n\nGoogle makes the trade-off explicit too. In its explanation of cooling choices, it says water-cooled data centres can use about 10% less energy than many air-cooled facilities. That comparison is context-dependent; it is not a universal penalty applicable to Microsoft’s project.\n\nThe decision therefore depends on the location: water availability, local temperatures, electricity sources and the demands of the servers. A design that protects a scarce water supply may justify additional electricity consumption. The case becomes stronger when that electricity is reliably low-carbon.\n\n## Where is AI’s less visible water footprint?\n\nSome of it sits beyond the data-centre boundary. The study Making AI Less “Thirsty” separates on-site cooling water from water used in electricity generation and the manufacture of servers. It also distinguishes water withdrawn from a source from water consumed through evaporation or otherwise removed from the immediate water environment.\n\nThese distinctions change how a claim should be judged. Eliminating routine evaporative cooling can deliver a real local benefit. It does not erase water used upstream. Nor is there a single defensible water cost for every AI prompt. The model, hardware, location, weather and electricity supply affect the calculation. A dramatic per-query number stripped of those assumptions tells readers less than it appears to.\n\n## What do the global examples show?\n\nFinland: use the sea, recover the heat. Google’s Hamina data centre uses seawater from the Gulf of Finland for cooling and has an offsite heat-recovery project with local energy company Haminan Energia. The approach draws on the location’s resources and creates a use for waste heat. It remains water-based cooling.\n\nUnited States: give wastewater another job. In Douglas County, Georgia, Google uses treated municipal wastewater for cooling. Water that does not evaporate is treated before being returned to the Chattahoochee River. This reduces reliance on drinking-water supplies, while evaporation still consumes water.\n\nChile: communities can change the design. TIME reported in 2024 that opposition to a proposed Google facility in Cerrillos, amid prolonged drought, helped secure a switch to air cooling. An environmental tribunal also required further consideration of climate impacts. Google said its\n\nprojects account for local conditions. The episode shows how water concerns can become a question of whether a project wins public acceptance.\n\nMicrosoft’s US projects: distinguish plans from proof. In its December 2024 roadmap, Microsoft named Phoenix, Arizona, and Mount Pleasant, Wisconsin, for pilots of its zero-evaporation design in 2026. That timetable is evidence of an engineering programme; it does not establish the pilots’ current performance.\n\nThe examples solve different problems. Seawater substitutes for freshwater. Wastewater reduces demand for potable supplies. Closed-loop designs avoid evaporative losses. Heat recovery puts otherwise discarded energy to work. Comparing them requires more than putting “green” beside each company’s name.\n\n## Does 1 GW of renewable contracts settle the power question?\n\nChandok’s procurement figures indicate investment in renewable supply. They do not reveal what share of Hyderabad’s electricity demand will be matched with clean generation hour by hour.\n\nGigawatts measure capacity. Electricity generated and consumed over time is measured in kilowatt-hours and their multiples. Without the region’s demand profile, generation mix, storage arrangements and delivery details, those capacity figures cannot establish continuous renewable operation.\n\nGoogle’s stated ambition to operate on carbon-free energy every hour, on every grid where it operates, illustrates the more demanding test. Its own disclosures also distinguish contracted capacity from actual generation.\n\nThe useful question is straightforward: what powers the servers when contracted renewable sources are not producing enough?\n\n## What does “putting water back” mean?\n\nChandok also referred to “5 million liters of ground capacity being built”. The supplied account does not specify the project, the measurement period or the verified annual replenishment. That figure should not be rewritten as five million litres of groundwater already restored.\n\nMicrosoft separately describes water-restoration work with United Way of Hyderabad in Elikatta village, Telangana. That establishes a named initiative, but does not by itself connect it to the five-million-litre figure or establish the new region’s net water balance.\n\nFor residents, the practical questions are where the benefit occurs, when water becomes available and how the outcome is measured. Storage capacity and annual replenishment are different quantities. A project’s potential and its delivered benefit also need separate reporting.\n\n## Why does this matter for India’s AI buildout?\n\nChandok estimated India’s data-centre capacity at roughly 2 GW and projected a six- to sevenfold expansion by 2035. On his figures, that would mean about 12–14 GW. It is his projection, rather than an independently established outcome.\n\nThe international direction is clearer. The International Energy Agency’s 2025 Energy and AI report estimated global data-centre electricity consumption at 415 terawatt-hours in 2024 and projected about 945 TWh by 2030, with AI the leading growth driver. It also stressed that demand\n\nconcentrated in particular locations can create pressures far greater than the sector’s global share suggests.\n\nThat creates a harder test for efficiency claims. Using fewer resources per unit of computing is valuable. Rapid expansion can still push total demand higher.\n\nFor Hyderabad, the evidence to watch is annual and peak water use, measured electricity demand, cooling performance during hot weather, the timing of clean-power supply and verified local replenishment. Publishing those figures would make the promise testable.\n\nMicrosoft’s proposed design addresses a real problem at its source. If it performs as described, it could sharply reduce one recurring demand on local water supplies.\n\nThe next chapter should be written in meter readings.\n\n*(With inputs from ANI)*", "url": "https://wpnews.pro/news/microsofts-zero-water-promise-can-indias-ai-boom-beat-the-heat", "canonical_source": "https://openthemagazine.com/business/microsofts-zero-water-promise-can-indias-ai-boom-beat-the-heat", "published_at": "2026-09-21 10:44:38+00:00", "updated_at": "2026-09-21 11:23:28.881211+00:00", "lang": "en", "topics": ["ai-infrastructure", "artificial-intelligence"], "entities": ["Microsoft", "India South Central", "Hyderabad", "Puneet Chandok", "Microsoft India and South Asia", "Microsoft AI Infrastructure Summit India", "Google"], "alternates": {"html": "https://wpnews.pro/news/microsofts-zero-water-promise-can-indias-ai-boom-beat-the-heat", "markdown": "https://wpnews.pro/news/microsofts-zero-water-promise-can-indias-ai-boom-beat-the-heat.md", "text": "https://wpnews.pro/news/microsofts-zero-water-promise-can-indias-ai-boom-beat-the-heat.txt", "jsonld": "https://wpnews.pro/news/microsofts-zero-water-promise-can-indias-ai-boom-beat-the-heat.jsonld"}}