Skip to content

AI Data Centres Are Rewriting the Power Grid: What the Utility Capex Boom Means for India

AI data centres are becoming a grid problem, not just a technology story. India’s power demand, transformer capacity, cooling and transmission build-out are next.

Market News Last reviewed: October 31, 2026 Official sources listed where provided
Cinematic AI data centre power-grid graphic showing U.S. electricity-price pressure and India transmission, transformer and cooling opportunity

AI is moving from the server rack into the electricity system

The global AI boom is no longer only about GPUs, HBM and software models. It is becoming a power-grid story.

In the United States, residential electricity prices have risen more than 10% nationally since 2024, while investor-owned utilities are planning almost $1.3 trillion of capital spending over the next five years. Data-centre demand is one of the forces pushing utilities toward new generation, substations, transmission lines, transformers and local grid upgrades.

That pressure is already creating political resistance. U.S. regulators and voters are asking how much of the cost of AI infrastructure should be carried by households and how much should be borne by the data-centre operators creating the load.

India is earlier in the same cycle, but the numbers are already large. Official government data puts installed Indian data-centre capacity at roughly 1.57 GW, with capacity projected near 8 GW by 2030 and electricity demand potentially reaching 17 GW by 2031–32.

That means the Indian AI investment thesis is expanding beyond chips and software. The next bottlenecks could be transmission, transformers, substations, cooling, storage and reliable round-the-clock power.

This connects directly with RegalTicker’s AI-memory/HBM shortage analysis: the physical AI stack is only as strong as its weakest infrastructure layer. HBM may be one bottleneck. Electricity can become another.

⚡ Quick answer

Which Indian sectors could benefit from AI data-centre power demand?

The clearest potential beneficiaries are businesses that supply the electricity infrastructure required by high-density data centres: large power transformers, switchgear, substations, grid automation, transmission networks, cooling systems, storage and reliable backup power. Hitachi Energy India explicitly says its new ₹2,000 crore transformer factory will support AI data-centre applications, while GE Vernova T&D India says it has already secured grid-automation packages from data-centre projects. POWERGRID sits at the national transmission layer. But not every power or transformer stock is automatically an AI beneficiary; actual orders, capacity, margins and customer exposure still need to be verified.

Key takeaways

U.S. residential electricity prices have risen more than 10% nationally since 2024.

U.S. investor-owned utilities plan nearly $1.3 trillion of capex over the next five years.

Alliant Energy says it already has 3.4 GW of contracted data-centre demand and is targeting roughly 1 GW more within six months.

India’s data-centre capacity has risen from 375 MW in 2020 to about 1.57 GW in 2026.

Official projections indicate Indian data-centre capacity could approach 8 GW by 2030.

Data-centre electricity demand in India could reach about 17 GW by 2031–32.

India plans to expand transmission network length from 5.09 lakh circuit-km to 6.48 lakh circuit-km by 2032.

Transformation capacity is planned to rise from 1,478 GVA to 2,345 GVA by 2032.

Hitachi Energy India is investing about ₹2,000 crore in a large-power-transformer plant that explicitly targets AI data centres among its applications.

GE Vernova T&D India disclosed grid-automation orders from data-centre projects.

POWERGRID operates 1,87,576 circuit-km of lines and 6,49,116 MVA of transformation capacity as of 30 September 2026.

Cooling, storage, renewable power and microgrids may become as important as computing hardware in the next AI infrastructure phase.

The opportunity is real, but valuation, order conversion, manufacturing capacity and return on capital still matter.

1. Why AI data centres are becoming a power-system problem

A high-density AI data centre needs continuous electricity, high-quality power, redundant feeds, substations, transformers, switchgear, cooling, backup generation or storage and enough local grid capacity to expand.

The physical difference from a traditional enterprise facility is density. As more accelerators are packed into a building, electricity delivery and heat removal become engineering constraints.

That means the bottleneck can move. First the market worried about GPUs. Then HBM. Now the question is whether the grid can deliver enough reliable power to energise all the hardware already being ordered.

This is why the hyperscaler AI-capex analysis matters: the AI investment cycle is no longer a pure technology capex story. It is becoming an infrastructure-financing problem as well.

2. The U.S. is showing what happens when data-centre demand reaches the grid

Reuters reported on 31 October that residential electricity prices have risen more than 10% nationally since 2024, with increases above 20% in some states. Investor-owned utilities are simultaneously planning almost $1.3 trillion of capex over five years.

Not all of that spending is caused by AI. Utilities also need to replace ageing infrastructure, connect renewables, harden grids and meet broader electrification demand.

But data centres are materially changing the growth assumptions.

Alliant Energy says it already has 3.4 GW of contracted data-centre demand from major customers and expects to sign around 1 GW more within six months. It sees peak system demand rising more than 60% by 2031.

That is industrial-scale incremental load arriving in regions whose grids were often planned before the current AI boom.

3. Why the politics of power prices matters to investors

Utilities earn returns by investing capital into regulated infrastructure. That model works when regulators allow companies to recover costs and earn an approved return through customer bills.

AI complicates this.

If utilities build substations, transmission and generation to serve data centres, regulators must decide who pays for upgrades, how costs are allocated, what return utilities can earn and what happens if a data-centre project is delayed.

That creates a new investment tension:

more data-centre demand can increase utility capex and earnings opportunity — but higher electricity bills can trigger regulation that compresses allowed returns.

So even in the U.S., “AI power demand up” is not automatically bullish for every utility stock.

4. India’s data-centre power demand is moving from small to material

India’s current data-centre base is still modest compared with total national electricity capacity, but the growth rate matters.

Official government data shows about 375 MW of installed data-centre capacity in 2020, roughly 1.57 GW in 2026, nearly 8 GW projected by 2030, and electricity demand potentially reaching 17 GW by 2031–32.

A 17 GW requirement is large enough to influence urban substations, state transmission planning, renewable procurement, grid balancing, transformer demand, power-quality equipment and storage.

India is already planning for that load rather than treating it as a distant possibility.

5. The transmission build-out is large even before AI reaches full scale

India’s National Electricity Plan includes major transmission expansion through 2032.

The government expects transmission network length to increase from 5.09 lakh circuit-km in June 2026 to 6.48 lakh circuit-km by 2032, transformation capacity to rise from 1,478 GVA to 2,345 GVA, and inter-regional capacity to expand from 120 GW to 168 GW.

Data centres are only one driver. Renewables, electric mobility, industrial expansion and general power demand are also increasing the load on the system.

That is actually positive for the investment thesis. A grid-equipment supplier does not need every order to come from AI. It can benefit from the broader requirement to expand and modernise the network that AI is helping accelerate.

RegalTicker’s India semiconductor boom article makes the same point from another direction: the strongest infrastructure themes often work through a broad ecosystem rather than one end product.

Hitachi Energy India is one of the easier listed connections to verify because the company itself has named AI data centres as a target application.

In June, Hitachi Energy announced an investment of approximately ₹2,000 crore to build a new large-power-transformer factory in Vadodara.

The company said the new plant will support high-voltage transmission, HVDC, power generation, AI data centres and large industrial applications. The factory is scheduled for completion in FY28.

That is stronger evidence than simply calling the company a “power equipment beneficiary.”

But it still does not guarantee a specific level of data-centre revenue. Investors need to watch factory completion, order intake, product mix, utilisation, export demand, margins and working capital.

Investor note

Direct exposure still needs financial proof

A company can have a credible AI-data-centre product and still deliver weak shareholder returns if orders, margins or utilisation disappoint. Verify the revenue contribution rather than buying only the theme.

GE Vernova T&D India provides another useful example because it has disclosed actual grid-automation activity connected to data-centre projects.

Its Q2 FY27 investor presentation said the company secured multiple Grid Automation packages from state utilities, EPCs and data-centre projects.

The company also manufactures power transformers, GIS and AIS switchgear, circuit breakers, substation automation, HVDC equipment, FACTS solutions and digital-grid systems.

Its order backlog at September 2026 was large, and Q2 order intake rose sharply year on year. The key point is not to label the whole backlog “AI.” The data-centre exposure exists, but it sits inside a much broader transmission and export business.

8. POWERGRID is the backbone exposure, not a pure AI trade

POWERGRID operates a huge part of India’s national transmission system.

As of 30 September 2026, the company reported 1,87,576 circuit-km of transmission lines, 293 substations, 6,49,116 MVA of transformation capacity and 99.81% system availability.

If India’s power demand grows because of data centres, manufacturing, EVs and renewables, national transmission investment should remain important.

But POWERGRID should not be described as a direct AI-data-centre beneficiary unless specific project disclosures establish that connection. Its exposure is broader.

That can reduce dependence on one technology cycle, but it also means investors should avoid attributing every transmission project to AI.

9. Cooling, storage and microgrids are becoming part of the AI stack

Power enters a data centre, but much of it eventually becomes heat.

The government says Indian data centres are increasingly adopting direct-to-chip liquid cooling, adiabatic cooling, immersion cooling, closed-loop liquid cooling and high-density rack systems.

Reuters has also reported that the AI scaling race is accelerating investment in microgrids and around-the-clock power solutions.

The physical AI chain increasingly looks like:

chips → servers → networking → cooling → electricity → transmission → storage → backup power

This creates opportunities across chillers, pumps, heat exchangers, controls, batteries, pumped storage, power electronics and backup generation.

But company mapping again requires proof. A general air-conditioning or battery company is not automatically a data-centre winner.

10. The biggest risk is building expensive infrastructure for demand that arrives late

AI data-centre forecasts are large, but not every announced project will arrive on schedule.

A data centre can be delayed by permits, chip availability, customer financing, water constraints, transmission delays, local opposition, power prices or changes in AI economics.

If utilities build too far ahead of actual demand, customers can be left paying for under-utilised infrastructure. If they wait too long, data-centre projects can move elsewhere because the power connection is unavailable.

For equipment suppliers, the equivalent risks are order cancellation, long lead times and customer concentration.

The Cash Flow Statement guide is especially useful because infrastructure booms often look strongest just before capex intensity becomes the key investor concern.

11. What Indian investors should watch next

The power side of the AI boom now has measurable checkpoints.

Watch whether India’s data-centre pipeline moves from announced investment to energised capacity. Watch transformer and switchgear lead times. Follow Hitachi Energy India’s Vadodara expansion and whether AI-data-centre orders become visible. Track GE Vernova T&D India’s grid-automation orders and POWERGRID’s broader network investment.

Also watch electricity prices and regulation. The U.S. shows that political pushback begins when households believe they are subsidising data-centre expansion.

Finally, watch AI economics themselves. If hyperscaler spending slows because returns disappoint, electricity forecasts can be revised just as quickly as chip forecasts.

That is why the Fundamental Analysis path and Risk Management path matter more than a simplistic “AI power stocks” basket.

Frequently asked questions

How much electricity could Indian data centres need by 2031–32?

Government data citing the Central Electricity Authority says data-centre electricity demand could reach about 17 GW by 2031–32.

How large is India’s data-centre industry today?

Installed data-centre power capacity was about 1.57 GW in 2026, up from roughly 375 MW in 2020.

How large could Indian data-centre capacity become by 2030?

Government material projects capacity near 8 GW by 2030.

Why are AI data centres putting pressure on power grids?

High-density AI servers require large amounts of continuous electricity, transformers, substations, cooling and backup power. New clusters can therefore create industrial-scale local demand.

Which Indian listed companies have verified data-centre power exposure?

Hitachi Energy India explicitly says its new transformer factory will support AI data centres, and GE Vernova T&D India says it has received grid-automation orders for data-centre projects. POWERGRID provides broad transmission exposure rather than a pure data-centre business.

Does every transformer or power stock benefit?

No. Investors should verify actual products, contracts, capacity, margins and customer exposure rather than relying on a thematic label.

Why is cooling important?

Dense AI computing creates large heat loads. Higher rack density is driving liquid cooling, immersion systems and other advanced thermal-management technologies.

Verify through official sources

Official references

Final takeaway

The next AI bottleneck may be measured in megawatts, not chips

The AI boom began as a software story, turned into a semiconductor story and is now becoming a power-infrastructure story.

The U.S. is already showing the consequences: electricity bills are rising, utilities are planning historic capex and politicians are arguing over who should pay for data-centre growth.

India has time to plan—but not much time to waste. Data-centre capacity could rise several-fold by 2030, while electricity demand could reach 17 GW by 2031–32.

That creates a credible long-term opportunity for grid equipment, transformers, automation, cooling, storage and transmission.

The investment discipline is the same as always: do not buy the label. Follow the orders, capacity, margins, cash flow and return on capital.

AI may drive the demand. The companies that create shareholder value will be the ones that can deliver the physical infrastructure profitably.

Educational disclaimer: RegalTicker provides educational and informational content only. This article is not investment advice or a recommendation to buy or sell Hitachi Energy India, GE Vernova T&D India, POWERGRID or any data-centre, power or infrastructure security. Data-centre forecasts, project timelines, grid-capex plans, electricity prices and order books can change. Verify current company, exchange, regulator and government disclosures before making financial decisions and consider valuation, liquidity, concentration and your own risk tolerance.

Dilip Kumar profile photo ✓
Written and reviewed by

Dilip Kumar

Founder & Author | Investor Education and Market Analysis Regal Ticker

Dilip Kumar is the creator behind Regal Ticker and focuses on investor education, technical analysis and stock-market learning. He simplifies complex concepts such as chart analysis, market trends, risk management and corporate actions through clear explanations and practical examples. His objective is to help investors build knowledge, verify information through official sources and develop a disciplined approach to market participation.

QualificationsB. Tech.
Experience10+ years studying Indian equity markets
Investor EducationTechnical AnalysisCorporate ActionsChart AnalysisMarket TrendsRisk ManagementStock-Market Basics
RegalTicker Briefing

Stay ahead with RegalTicker

Get research-led market updates, major IPO and corporate-action explainers, plus our daily, weekly and monthly market reviews in your inbox.

Daily context Weekly review Monthly perspective Major market news

No stock tips. No guaranteed-return claims. Unsubscribe anytime.