India AI Data Centres Put Power Before the Job Boom

India AI data centres, built for artificial intelligence workloads, are becoming the country’s next industrial buildout. The opening is a new layer of skilled infrastructure work, from cooling engineers to fibre technicians, but the boom will be capped by electricity, water and grid planning unless those arrive with the server halls.

That makes this rush different from the software services wave that turned Bengaluru, Hyderabad, Pune and Gurugram into global back offices. Code still matters. The scarcer resource now is the physical stack beneath it: land, substations, graphics processing units, cooling loops, cables and people who can keep all of it running at high uptime.

A Capacity Race Measured in Megawatts

Data centres are counted in megawatts (MW, a measure of power demand) because the business begins with electricity. A 100 MW campus is not just a building with servers. It is a permanent industrial load that wants power every hour, backup systems behind it and enough cooling to keep dense computing racks alive through Indian summers.

Colliers, the global property advisory firm, said India’s data centre capacity across the top seven cities stood at 1,263 MW in April 2025 and could cross 4,500 MW by 2030 in the Digital Backbone data centre growth report. Deloitte, the professional services firm, puts the higher artificial intelligence (AI) path at roughly 8 to 10 gigawatts (GW, 1,000 MW) by 2030 if India resolves the power challenge.

The spread between those forecasts matters. The lower track still turns data centres into a major real estate and utility story. The higher track makes them a power system story with technology wrapped around it.

Policy Put the Floor Under Demand

India’s demand is not coming only from streaming, shopping and banking apps. Regulation has been pushing data closer to home for years. The Reserve Bank of India (RBI, the central bank) told payment system providers on April 6, 2018 that the entire data relating to payment systems operated by them must be stored in India, with compliance due within six months under the RBI payment data storage directive.

The Digital Personal Data Protection Act (DPDP Act, India’s privacy law for digital personal data) added a broader legal lever. Section 16 says the central government may restrict the transfer of personal data for processing to notified countries or territories outside India, according to the official DPDP Act text from MeitY, the Ministry of Electronics and Information Technology.

Budget policy has now moved from guardrail to lure. A Press Information Bureau note on the 2026-27 Budget says eligible foreign cloud service providers operating through India-based data centre infrastructure can get a tax holiday till 2047, a long runway for companies selling cloud services to global customers from Indian sites.

• Payments data created the first clear on-soil storage mandate for a high-volume digital sector.
• Personal data gave the government a wider transfer-control tool through the DPDP Act.
• Global cloud revenue now has a tax incentive if workloads run through Indian data centres.

That combination gives builders something they prize: predictable demand. Banks, insurers, public agencies and regulated platforms need local capacity. Foreign cloud firms can use India as a service base for overseas customers. Domestic AI builders want cheaper access to compute without shipping every model run abroad.

The Hyperscalers Want Latency, Sovereignty and Cables

The biggest commitments are no longer described as ordinary server farms. Google, Alphabet’s search and cloud company, broke ground on its Visakhapatnam AI hub in Andhra Pradesh and said the project is part of a $15 billion investment to provide foundational infrastructure for India’s digital economy. The company said Google’s India AI hub construction update includes gigawatt-scale compute with AdaniConneX and Nxtra by Airtel, plus a new subsea gateway in Vizag.

Microsoft, the U.S. software and cloud company, has put its own marker down. In December 2025, it announced $17.5 billion over four years, from calendar 2026 to 2029, for cloud and AI infrastructure, skilling and operations in India. The company said Microsoft’s India cloud and AI investment plan includes a Hyderabad-based India South Central cloud region set to go live in mid-2026.

Amazon Web Services (AWS, Amazon’s cloud unit) said in the AWS India cloud infrastructure announcement that its planned investment would support an estimated average of 1,31,700 full-time equivalent jobs annually at local businesses by 2030. Adani Group, the Indian infrastructure conglomerate, went even larger, announcing Adani’s sovereign AI infrastructure plan with a 5 GW target and a model linking renewable power, grid resilience and hyperscale compute.

The Job Boom Starts on the Construction Site

The first wave of employment will look less like an app campus and more like a high-spec industrial project. A hyperscale AI data centre needs civil engineers, electrical contractors, fibre crews, heating and cooling specialists, security staff, facilities managers and operations engineers before it needs a single consumer product manager.

That is why the skills conversation is changing. Rahul Takkallapally, co-founder of BharathCloud, told the source report that India already needs professionals in hyperscale infrastructure management, graphics processing units (GPUs, chips built to handle many calculations at once), GPU clustering, liquid cooling, cybersecurity, energy-efficient management and power management. Sudhir Kunder, chief business officer at DE-CIX India, pointed to another gap: interconnection architecture for multi-cloud connectivity, peering, private cloud access, routing resilience and secure data movement.

The U.S. is already showing what happens when construction outpaces trades. CBRE, a commercial real estate and critical infrastructure services firm, and Meta Platforms, Facebook’s parent company, announced LevelUp, a multi-year effort to recruit and train thousands of fibre technicians for Meta data centres. The CBRE and Meta fibre technician programme trains workers to install fibre-optic cable, racks, network gear and mission-critical equipment.

India will need its own version, and not only in engineering colleges. The useful models may come from short, employer-linked courses that train workers fast and then place them on live sites. Mind Cron has covered that shift abroad through Microsoft and Victoria University’s datacentre academy model, which shows how cloud companies are trying to turn local training into an infrastructure advantage.

The roles likely to grow fastest sit between blue-collar trades and software operations: fibre splicers who understand network latency, electricians who understand redundancy, cooling technicians who can work around liquid systems, and security analysts who know why local storage rules change facility design.

Power Is the Price of Sovereign Compute

The job story gets weaker if the power story fails. Deloitte says in the India data centre power challenge report that India contributes about 20 percent of global data consumption but hosts less than 5 percent of the world’s data centres. That gap is the opportunity. It is also the warning.

Government statements show the same tension. The Press Information Bureau said about 38,231 GPUs have been onboarded through 14 empanelled service providers and data centres under the AI compute capacity framework, with subsidised access at an average rate of ₹65 per hour. The same PIB data centre electricity and GPU update says data centre electricity demand is estimated to reach 13.56 GW by 2031-32.

• 13.56 GW is the government estimate for data centre electricity demand by 2031-32.
• 40-45 TWh is Deloitte’s estimate of additional AI-led power demand by 2030, measured in terawatt-hours (TWh, electricity used over time).
• 2.5% to 3% is Deloitte’s estimate for data centres as a share of national electricity demand by 2030.
• 38,231 GPUs have been onboarded under India’s AI compute capacity framework, according to the government.

Cooling adds another layer. The government says water needs depend on the cooling technology used and points to direct-to-chip liquid cooling, adiabatic cooling and immersion cooling as ways to reduce water use. Investors are already hunting for cleaner data centre power solutions, a theme Mind Cron covered in the clean fuel generator push for data centres. The broader risk is familiar: Big Tech can buy renewable power contracts while local grids still carry stress, as Mind Cron noted in the fossil fuel pressure behind AI data center growth.

Power discipline will decide whether this becomes national capability or a collection of expensive halls leased to the highest bidder.

Local Workloads Decide the Payoff

The strongest version of India’s data centre push is not measured only by foreign capital. It is measured by what runs inside the racks. Local-language models, banking workloads, public service AI, health data processing, manufacturing analytics and low-latency services for Indian users create more durable value than generic storage rented to global cloud traffic.

That is where the analogy with the old information technology services boom helps, but only up to a point. Two decades ago, India built an export engine around talent and process discipline. This time, the country also has to build the physical plant: power lines, fibre routes, land approvals, water management, chip access and safety standards.

If India pairs data localisation with reliable clean power and serious technician training, the AI data centre rush can produce a new class of skilled infrastructure jobs. If the buildout outruns the grid, the boom will still arrive, but the richest work will sit elsewhere.