India AI Data Centre Stocks Face a Power Reality Check

India’s operational data-centre capacity has already climbed past 1.8 GW of IT load in the first half of 2026 and is on course to multiply several times by 2030, according to successive industry tallies. Five listed companies now sit in the middle of that build: they make the servers, rent the GPUs, raise the halls, pull the fibre and supply the high-voltage gear. The boom is real. The second-order question is whether the grid can keep up with the order books.

Wood Mackenzie projects operational capacity rising from 2.2 GW in 2025 to 12 GW by 2030, a roughly 40 percent compound annual growth rate, with AI-dedicated capacity expanding nearly 24-fold from 275 MW to 6,546 MW. Savills India sees a path to more than 7 GW IT by the same date after a 59 percent jump in additions in the first half of this year. Government figures put the base near 1,500 MW by 2025. The numbers differ by source, yet every series points the same direction.

Five Layers of the Same Machine

Strip an AI cluster to its bones and five physical layers appear. Each of the five stocks owns one.

These are the listed, liquid ways an Indian investor can own a piece of the physical body of the AI build without owning Nvidia silicon or a hyperscaler equity stake.

Netweb Builds the Boxes the Chips Sit In

Netweb Technologies assembles high-performance AI servers in India and holds official partnerships with both Nvidia and AMD. In the quarter ended June 2026 its revenue jumped to ₹8.2 billion from ₹3 billion a year earlier, a 172 percent rise. Net profit climbed 180 percent to ₹850 million. AI systems revenue grew nearly six times and now accounts for close to two-thirds of the total.

The confirmed order book stood at roughly ₹25 billion at the end of June, with additional L1 positions and a much larger pipeline of opportunities under pursuit. The company has already secured work under the government’s IndiaAI Mission, including clusters for foundational models and sovereign compute facilities. That visibility is why the shares trade at more than 100 times earnings. One soft quarter would test the multiple hard.

• AI systems share of revenue: ~62 percent in Q1 FY27
• Order book: ₹25 bn confirmed plus further pipeline
• Valuation: triple-digit earnings multiple leaves little room for execution slips

Recent market chatter has also noted a possible QIP fundraise in the range of ₹1,200 crore, a reminder that even high-growth server makers still need capital to keep pace with GPU generations.

E2E Networks Rents the Compute by the Hour

Not every buyer wants to own the metal. E2E Networks is India’s first listed pure-play GPU cloud provider. For years it was small and loss-making. Then the AI wave arrived. In the June 2026 quarter revenue rose more than four times to ₹1.57 billion and the company swung to a profit of ₹439 million. Monthly rental collections nearly doubled in three months.

The engine is a fresh cluster of 1,024 Nvidia B200 chips running near full utilisation. Total GPUs under management have reached about 5,100. The company split its stock ten-for-one in June to widen the shareholder base. Chips are expensive and age quickly, so E2E must keep spending just to stay current. After the run-up the shares also trade at elevated multiples that assume the utilisation rates hold.

Anant Raj Turns Property Into Hardened Halls

Servers and clouds need buildings with backup power, cooling and security. Anant Raj, long known as a Delhi-region real-estate developer, now runs 28 MW of data-centre capacity across Manesar and Panchkula. Management is targeting 63 MW by December 2026 and 117 MW by 2028. A larger long-term plan points toward several hundred megawatts by the early 2030s, backed by a memorandum with Haryana that contemplates investment up to ₹250 billion over time and a Singapore arm aimed at cloud and AI work abroad.

In the first half of the prior financial year the data-centre business already contributed about ₹584 million and most of the group’s operating profit. The company plans to hive the digital business into a separately listed entity so investors can own the pure play. Data centres devour cash. Funding the ramp while keeping debt in check remains the central execution risk.

Sterlite Pulls the Fibre That Ties the Racks Together

Modern AI halls pack GPU racks so tightly that they need far more optical fibre than older computer rooms, by some estimates dozens of times more. Sterlite Technologies makes the fibre, cable and optical connectivity products. In the June 2026 quarter revenue rose 87 percent to ₹19.1 billion and profit hit a record ₹1.97 billion. After years of heavy debt the balance sheet has swung to net cash.

The order book stood at ₹186 billion in June, more than double the figure three months earlier. One contract explains most of the jump: a multi-year optical supply deal above $1 billion (roughly ₹100 billion-plus) to equip next-generation AI data centres for a hyperscaler between now and 2029. Managing director Ankit Agarwal said the company would support building AI data-centre infrastructure in the United States for that customer and enable the connectivity backbone.

Under this agreement, STL, through its optical solutions, will support building AI data center infrastructure in the US for this hyperscaler. We are enabling connectivity backbone for the AI data centers.

Ankit Agarwal, managing director of Sterlite Technologies, said that in the May 2026 announcement. Fibre remains a cyclical, price-sensitive business and Sterlite has disappointed investors before. For the first time in years the order book and the balance sheet point in the same direction.

Hitachi Energy Feeds the Power That Runs Everything

An AI data centre is first of all a giant electricity customer. One large campus can draw as much power as a small town. Hitachi Energy India supplies the transformers, switchgear and grid links that carry that load. It is the sturdiest of the five businesses, selling to utilities, railways and industry as well as data centres.

In the June 2026 quarter its order backlog reached a record ₹322 billion. Management said data centres ranked among the largest sources of fresh orders, including a single power project for a data centre in Hyderabad. The company is building its twentieth factory in India to meet demand. Quality businesses attract quality multiples; the stock trades near 150 times earnings. That is a steep toll even for a multi-year power-equipment up-cycle.

Why Power Is the Real Sorting Mechanism

Wood Mackenzie’s research makes the second-order point explicit: reliable, cost-competitive power has overtaken land and capital as the primary constraint on data-centre development. Captive generation and long-term renewable power-purchase agreements are becoming the preferred routes. States with liberalised open-access rules and competitive transmission charges will win projects; those without will lose them.

Electricity demand from data centres is forecast to grow twenty-fold by 2040 from a 2025 base of about 10 TWh and could account for 7 percent of total Indian electricity demand. That is the backdrop against which Hitachi’s backlog and every other order book must be judged. AI power demand rewriting converter makers playbooks is already visible far beyond India. The same constraint shows up in global surveys where 83 percent of firms need agentic AI infrastructure yet few have solved the underlying electricity and cooling maths.

Policy is trying to pull demand forward. The Union Budget 2026-27 proposed a tax holiday till 2047 for foreign cloud providers that route global cloud services through notified Indian data centres, with domestic transactions remaining taxable and a 15 percent safe-harbour margin for related parties. The measure is meant to give multi-decade certainty to capital-intensive builds. It also raises the volume of power and fibre that will eventually be required.

1. 2020-2025: Capacity rises from roughly 375 MW to about 1,500 MW.
2. H1 2026: Additions surge 59 percent; operational IT load reaches ~1.8 GW.
3. End-2026 target range: Industry and government expect a crossing of 2 GW and further acceleration.
4. 2030 forecasts: Wood Mackenzie 12 GW operational; Savills above 7 GW IT; AI-dedicated capacity multiplies dramatically.

Across Asia-Pacific the same capital wave is visible: APAC executives flag a $160 billion infrastructure bet on AI appreciation. India’s slice is large enough that local listed names can capture meaningful revenue even if global hyperscalers own the ultimate economics.

Debt, Multiples and How Long the Ramp Lasts

Three practical questions sit under every bullish presentation. First is leverage. Most of the five have avoided heavy debt. Sterlite has moved to net cash. Netweb and Hitachi Energy carry little. Anant Raj and E2E have leaned on equity raises rather than loans. That is comfort, not immunity. Capex intensity remains high.

Second is valuation. Four of the five already price in years of rapid growth. A rich multiple holds only while the growth arrives on schedule. Third is duration. Capex waves of this steepness rarely move in a straight line. When the slope flattens, the question becomes how far the stocks fall from elevated bases. Order books and capacity targets are promises. Profits arrive only when the concrete is poured, the fibre is lit and the transformers are energised.

Crowd conversation on X tracks the same tension: Sterlite’s multi-bagger move after the hyperscaler award is cited as proof that one contract can reprice a cyclical name, while Netweb’s growth and possible capital raise are watched for signs of either strength or dilution. Power-sector names appear on the same watchlists because the electricity constraint is no longer theoretical.

What Converts and What Does Not

India is wiring itself for AI at speed. Unlike the chips themselves, the server, cloud, building, fibre and power layers trade in plain sight on Indian exchanges. The direction of travel is hard to argue with once capacity has already more than quadrupled since 2020 and Wood Mackenzie sees capacity rising from 2.2 GW to 12 GW by 2030.

The second-order filter is simpler. Projects that secure reliable power and competitive tariffs will convert order books into multi-year earnings. Projects that do not will slip. The five stocks will not move in lockstep when that sorting begins. Investors who treat them as a single AI proxy will discover the differences the hard way. The concrete is going down now. The electricity bill arrives next.

Disclaimer: This article is for information purposes only. It is not stock advice or a recommendation to buy or sell any security.