Karnataka's Cabinet approved its Sustainable Data Centre Policy 2026–2031 at a meeting chaired by Chief Minister DK Shivakumar in Mangaluru, Deccan Chronicle reported. The policy carries a Rs 1,300.45 crore budget and targets 1 GW of cumulative data centre IT load by 2031, or 10–12% of India's AI-ready capacity. Its most consequential choice is not the gigawatt target. It makes operators publish plant-wise efficiency numbers and pays them for hitting thresholds.
The strongest case for the policy
The worry behind it is real. In March 2026, IT/BT Minister Priyank Kharge told the Assembly that a 1 MW data centre needs roughly 25 million litres of water a year, according to South First. The same report recalls that Bengaluru faced a 500 million-litre-a-day shortfall in 2024. The News Minute reports that the state's 32 existing data centres use about 4,000 kilolitres a day. A state that has just lived through an urban water crisis cannot be expected to wave through hundreds of megawatts of new load without asking where the power and water come from. Kharge called the policy "an important course correction", per Deccan Chronicle.
On that logic, the policy's design is largely right. Operators must report Power Usage Effectiveness (PUE) and Water Usage Effectiveness (WUE) plant by plant. The state also promises resource mapping to find areas with adequate power and water before approving expansion. A regulator that can see actual consumption can make proportionate decisions. One that sees only press-release pledges cannot.
Incentives for outcomes, not a ban on inputs
The policy rewards results. Per The News Minute, a facility achieving a PUE of 1.00–1.20 can receive up to Rs 20 crore in one-time support. Water-linked benefits include 100% reimbursement of treated-water charges, capped at Rs 8 lakh a year for five years. A further reimbursement of up to Rs 30 lakh a year for two years is available if WUE stays within 0.7–1.3 litres per kWh. Compare this with a moratorium or a blanket cap on water draw. A moratorium punishes the efficient and the wasteful alike. A payment tied to measured performance lets a hyperscaler with advanced cooling outcompete a legacy operator, which is how a pro-innovation sustainability policy should work.
The planned parks follow the same logic. The Hoskote park, about 500 MW, is to draw on Pavagada solar power and 60 million litres a day of secondary-treated water, with tenants doing their own tertiary treatment, The News Minute reports. Using treated wastewater for cooling rather than potable supply sidesteps the core objection.
Where the design weakens
First, the payout timing. The PUE support requires consistent performance for three consecutive years and is disbursed in year four, per The News Minute. For a developer financing a facility, that is a reward for surviving, not a reason to choose better cooling at the design stage. Operators will price the efficiency decision when they raise capital, and a payment arriving after year three barely enters that calculation. Faster, partially upfront disbursement against verified commissioning data would shift behaviour earlier.
Second, the measurement boundary. A PUE of 1.20 is demanding, and the figure depends heavily on what is counted: IT load, cooling, power conversion losses, on-site generation. WUE is similarly sensitive to whether treated water, recycled water and evaporative loss are counted. Plant-wise reporting only disciplines the market if the measurement method is fixed and published, and if auditors can check it. The coverage of the policy reviewed here does not make clear who audits these figures. Without that, the reward rewards creative accounting.
Third, the local-presence test. Eligibility requires registration with Karnataka Innovation and Technology Services (KITS) and a registered office in Karnataka, per The News Minute. Registration is a reasonable administrative step. Tying support to a state-registered office is a lighter version of a localisation rule. It is cheap for a large operator to satisfy and a nuisance for a smaller or foreign one. It does nothing for water or power efficiency, so it sits awkwardly in a policy whose logic is outcome-based.
The national context
Karnataka is competing in a market the Centre is actively courting. According to a Ministry of Electronics and IT statement on the Digital India site, national capacity grew from about 375 MW in 2020 to around 1,500 MW by 2025. Karnataka's 1 GW target is therefore large against the whole country's current base. The Union Budget 2026–27 added a tax holiday through 2047 for eligible foreign cloud providers using MeitY-notified India-based data centres, with services to Indian customers routed through an Indian reseller, according to a Press Information Bureau backgrounder of 14 February 2026. Capital is being invited in. States will win it by being predictable on power and water.
Note also that the three parks are planned at 1,000 MW combined, which equals the entire 2031 target. The target can only be met if the parks fill quickly, so the quality of the resource mapping matters more than the headline figure. Hoskote's 60 million litres a day of treated water is a planning assumption, and it has to be shown to hold under drought conditions before tenants commit.
What would make it better
Karnataka should keep the performance-based structure and fix the three weak points: publish the PUE and WUE measurement method and an audit protocol, move part of the incentive earlier against verified commissioning data, and drop the registered-office condition in favour of operating in the state. It should also publish the aggregate plant-wise figures it collects. Public data would let residents, researchers and investors test the state's claims, and would give other Indian states a template. Treated water and measured efficiency are a better answer to scarcity than restricting the industry, but only if the measurements can be trusted.