Is India's Data Centre Boom Built On Promises And Not Numbers?
What if the biggest winner of India's AI boom isn't the company building the most powerful technology, but the one making investors believe in the biggest future? Over the past year, “AI” and “Data-Centre” have turned into the stock market's favourite magic words.
Companies have announced billion-dollar investments, massive data centres and thousands of GPUs, while some stocks have jumped several hundred percent. From small technology firms to India's biggest business houses, everyone seems to want a piece of the AI gold rush. But there's one thing that doesn't always move as fast as share prices: actual earnings.
Global Tech leaders and PM Narendra Modi at the AI Impact Summit Source: Bloomberg
And that's where the story gets interesting. Nearly a year after the AI boom took off, some companies are already making serious money, others are still building what they promised, and a few are asking investors to bet on a future that hasn't arrived yet.
So, we went back to the financials, order books, earnings calls and project updates to find out who's actually delivering and who's still selling the dream. Because in a market where everyone is talking about the next AI goldmine, the real question is who has found the gold and who is still digging.
The Original AI Darlings That Started The Frenzy
Before the data centre story turned into a race to announce gigawatt-scale projects, investors had already picked their AI favourites. Between August 1 and early October 2025, Netweb Technologies jumped 115.86 percent, E2E Networks gained 90.51 percent, and MTAR Technologies rose roughly 40 percent.
As Nvidia and its GPUs dominated AI headlines, investors looked for Indian beneficiaries. Netweb made powerful computing systems, E2E rented out GPUs, and MTAR supplied components for fuel cells that could power energy hungry data centres. But nearly a year later, what do their financial numbers actually show?
Starting with Netweb Technologies . In Q1FY26, the company earned Rs. 301.2 crore in revenue. By Q1FY27, that had reached Rs. 819.7 crore, up 172 percent, while net profit climbed from Rs. 30.5 crore to Rs. 85.3 crore.
More importantly, AI systems alone brought in Rs. 510.6 crore, accounting for 62 percent of quarterly sales and growing 484 percent year-on-year. Its confirmed order book stood at around Rs. 2,507 crore in June 2026, with another Rs. 848 crore where it was the lowest bidder. These are real equipment sales, although pending orders still need to be executed.
Source: Netweb Technologies Q1FY27 Investor Presentation
E2E Networks' story was more complicated. It spent heavily buying GPUs before demand fully caught up. In Q1FY26, revenue was just Rs. 36.1 crore and the company posted a Rs. 2.8 crore loss. It eventually closed FY26 with a Rs. 15.6 crore loss and Rs. 245.6 crore in revenue.
But by Q1FY27, revenue had surged to Rs. 156.8 crore, up 334 percent, and net profit reached Rs. 43.9 crore. Its live GPU capacity crossed 5,100, including newer Nvidia B200 machines. Management had targeted a March end monthly revenue run rate of Rs. 35-40 crore in Q2FY26 earnings call and achieved Rs. 37.4 crore.
By June, the figure had reached Rs. 71.8 crore. So the investment is finally showing results, although GPUs remain expensive assets that bring depreciation and financing costs. However, its quarterly profits have been highly volatile, swinging between losses and profits over the past year, raising questions about whether the recent sharp improvement in profitability can be sustained.
Source: E2E Networks Q1FY27 Investor Presentation
MTAR Technologies offers a different kind of exposure. It supplies parts used in fuel cells, including those made for Bloom Energy, whose customers include data centre operators like Oracle. MTAR's FY26 revenue grew to Rs. 876.2 crore, while Q1FY27 revenue jumped 130 percent to Rs. 360.7 crore and profit exploded 364.5 percent to Rs. 50.2 crore.
Its clean energy business contributed Rs. 220.8 crore in Q1FY27, but this includes businesses beyond data centres. MTAR also secured a direct data centre infrastructure order of about Rs. 45 crore. However, initial product qualification was still underway in July, with execution expected by March 2027. Its Rs. 5,143 crore total order book also spans several industries, not just AI. Management's FY27 guidance of 80 percent revenue growth remains a target.
Source: MTAR Technologies Q1FY27 Investor Presentation
Then came Fractal Analytics , which listed in February 2026 amid growing interest in AI-related listings and names such as Sify and Yotta. Unlike the others, Fractal sells AI and analytics services rather than physical infrastructure. FY26 revenue stood at around Rs. 3,300 crore with a 19 percent growth, with net income growing 30 percent to Rs. 287 crore. In Q1FY27, revenue grew 20 percent to Rs. 912.5 crore and net income rose 92 percent to Rs. 72.3 crore.
But revenue growth was only 9 percent at constant exchange rates. Its smaller Fractal Alpha business also posted a Rs. 14 crore segment loss, showing that newer products still require investment. Its BMC hospital AI project was also only a pilot, not a citywide rollout.
Source: Reuters
So these companies were not merely selling dreams, but their valuations tell an interesting story. Netweb Technologies' P/E ratio has cooled from a peak of around 216x to 102x, even as its shares continued to rise, suggesting that earnings growth is gradually catching up with its valuation. Similarly, MTAR Technologies' P/E has fallen sharply from nearly 500x in June 2026 to around 175x, despite its stock moving higher, reflecting stronger earnings.
However, E2E Networks tells a different story. Its P/E has jumped from around 100x at the beginning of the year to nearly 476x, even as its share price continues to climb, suggesting investors are pricing in substantial future growth despite volatile profits. Meanwhile, Fractal Analytics' P/E has fallen from around 60x at listing to 32.8x, with its share price also declining from highs of around Rs. 1,100 to Rs. 700.
The Billion-Dollar Race To Build India's AI Future
By late 2025, it was no longer enough from smaller companies to talk about fancy words like “AI” and “Data Centres”. India's biggest companies started announcing data centres measured in gigawatts, with investments running into billions of dollars. Reliance, Adani, Airtel, TCS, L&T and Lodha all wanted a place in the race. But almost a year later, how much has moved beyond the headlines?
Source : Trade Brains
Reliance Industries announced gigawatt-scale AI data centres in Jamnagar, alongside a Google Cloud region powered by its renewable energy. In October 2025, management said the Google project could reach a little over 100 MW within two years. By April 2026, Reliance confirmed work had begun on data centres for its own use and partners.
In July, it disclosed a more specific 168 MW Jamnagar project with Meta, covering infrastructure, power, connectivity and managed services. However, management kept the commissioning date confidential and did not disclose project-specific revenue or spending.
And that's what makes Reliance's progress interesting. As India's biggest listed company and a business that has shaken up industries like telecom and retail, investors might have expected Mukesh Ambani to move much faster in the AI race.
After all, this is the man who changed India's internet market with Jio in a decade. But nearly a year into the AI excitement, Reliance's biggest data centre plans are still taking shape.
Source: Trade Brains
Adani Enterprises has been more specific about operating capacity. Its AdaniConneX business secured a new 400 MW hyperscale order linked to the older Google contract in Visakhapatnam, taking total tied-up capacity above 960 MW. But as of June 2026, only 65.4 MW was operational, including 9.6 MW newly handed over to a customer in Pune in May end.
Facilities in other locations remain at different stages of construction. Management expects capacity to reach roughly 500 MW over the next three years. That is visible progress, but still a considerable gap between the gigawatt level contracts, USD 250 Billion worth of ecosystem promised and the current operating facilities.
And this could turn into an interesting fight between two of India's richest businessmen. Mukesh Ambani and Gautam Adani have largely avoided competing directly with each other. But India's AI boom could change that.
Source: Adani Enterprises Q1FY27 Investor Presentation
Bharti Airtel , through Nxtra, is targeting 1 GW over the coming years. In Q1FY27, management described its existing base as around 120-130 MW. The company announced a USD 1 billion fundraise for Nxtra and has an agreement to build data centres for Google in Visakhapatnam (the same DC that Adani has also announced), alongside connectivity and a cable landing station.
Management also said more contracts were in place, land acquisition was progressing and several hundred megawatts were in the development pipeline. The full gigawatt, however, is still a future target.
Google Breaks Ground on India AI Hub, Launching a National Industrial Ecosystem Alongside India's Digital Infrastructure Milestone
Source: Bharti Airtel
TCS entered the race through HyperVault, targeting 1 GW over five to seven years. It announced a USD 1 billion equity partnership with TPG and partnered with OpenAI for an initial 100 MW, expandable to 1 GW. It also tied up with AMD and brought together Tata Power, Tata Projects, Tata Communications, GE, Honeywell, ABB and Siemens for infrastructure and equipment.
By Q4FY26, TCS had secured customer commitments and was finalising land parcels. In Q1FY27, its HyperVault partnerships helped win two large IT services deals. Separately, TCS won an AI-led transformation contract worth over USD 800 million from SKF, while its annualised AI services revenue reached USD 2.6 billion. So the partnerships are creating business opportunities, although TCS has not disclosed any commissioned HyperVault capacity.
Among both the big conglomerates and smaller AI players, TCS has arguably made the most noise with its partnerships and announcements. And with its huge cash reserves and financial strength, it certainly has the resources to back those ambitions. But can all this noise actually turn into operational data centres and meaningful revenue from HyperVault?
Source: Tata Consultancy Services
L&T had much bigger ambitions, proposing gigawatt-scale AI data centre infrastructure through its collaboration with NVIDIA under the IndiaAI Mission. Its plans included expanding GPU capacity at its 300 acre Chennai campus to around 30 MW and developing a 40 MW facility in Mumbai.
But the actual numbers tell a smaller story. During FY26, its Vyoma business (DC Business) commissioned 12 MW at Kanchipuram, with another 6 MW nearing commissioning taking the total commissioned capacity to 30 MW. Management later discussed 30 MW under construction at Mahape and a target of around 200 MW over time, backed by roughly Rs. 10,000 crore of planned investment over five years.
Interestingly, L&T also holds a reported 18.45 percent stake in E2E Networks , with both companies working together on cloud and data centre infrastructure. So L&T has moved beyond announcements, but its operational capacity is still far smaller than its gigawatt-scale ambitions.
An image of L&T’s Data Centre In Mumbai
Source: Larsen & Toubro
Lodha is playing a different game. Its Rs. 1.3 lakh crore data centre announcement includes customer investments, not just its own. At its 660-acre Palava park, backed by 3 GW of power, it had sold 132 acres to operators including AWS, STT and Digital Edge by June 2026. Land sales, including non-data-centre transactions, contributed around Rs. 600 crore to Q1FY27 profits. Lodha expects another Rs. 9,000 crore from land sales over the next three to four years.
But the bigger bet is its proposed 1 GW powered-shell business, where Lodha supplies buildings and power rather than GPUs. It targets over Rs. 2,000 crore in annual rent by FY32, while Morgan Stanley estimates around Rs. 2,100 crore and over 20 percent potential valuation upside from data centres.
Management expects its first lease in FY27 and construction to begin in calendar 2026. Yet by June, just Rs. 20 crore had been invested against Rs. 10,000-11,000 crore still planned, with no completed capacity. So land monetisation is real, but the much larger rental opportunity remains a future promise.
Source: Lodha Developers Q1FY27 Investor Presentation
Clearly, the announcements have not simply vanished. Land has been sold, deals signed and some facilities commissioned. But most of the headline gigawatts remain capacity to deliver, not capacity already earning money.
The Optical Fibre Gold Rush
After the first AI stocks and gigawatt data centre announcements grabbed headlines, investors found another way to bet on the boom: optical fibres. These cables connect data centres, GPUs and communication networks.
The Union Budget also added excitement by proposing a tax holiday until 2047 for eligible foreign cloud companies using Indian data centres. But the benefit was conditional, and it was not a direct tax break for fibre manufacturers.
In fact, Sterlite Technologies had already started rallying in late January, gaining over 1,000 percent by October 2026, while HFCL surged as much as 320 percent during the same period. The rallies showed how investors were looking beyond companies building data centres and increasingly betting on those supplying the infrastructure needed to power India's AI boom.
Source: Tradingview
But unlike some of the biggest data centre announcements, there were real orders behind this rally. In January, Sterlite Technologies completed multi-core fibre trials with Colt in London. In March, it launched its hollow-core fibre cable designed for faster data transmission.
More importantly, in May, its subsidiary received a hyperscaler product award with a potential value of USD 1.11 billion for supplying optical connectivity products through FY29. In October, another filing disclosed a long-term supply agreement potentially worth USD 1.2 billion through 2030. However, the customers were unnamed, and purchase orders would be released in stages. These figures cannot simply be treated as revenue already earned or automatically added together.
Source: Sterlite Technologies
And the financials have started reflecting the demand. In Q1FY27, STL's revenue jumped 87 percent year-on-year to Rs. 1,910 crore, EBITDA rose 184 percent to Rs. 397 crore, and net profit reached Rs. 197 crore.
The most telling number was the data centre business itself: it contributed 21 percent of quarterly revenue, compared with just 1 percent in FY26. North America accounted for 54 percent of sales, showing the importance of overseas hyperscaler demand.
However, management's biggest targets are still ahead. STL wants revenue to jump from Rs. 4,750 crore in FY26 to Rs. 20,000 crore by FY29, with EBITDA margins rising from 13.2 percent to more than 27 percent. That is a massive step up, even after the strong first quarter.
Source: Sterlite Technologies’ Investor & Analyst Meeting Call Transcript
HFCL has also moved beyond announcements. During FY26, it secured a global optical fibre cable supply contract valued at about $1.1 billion. In Q1FY27, revenue more than doubled from Rs. 871 crore to Rs. 1,915 crore, while net profit swung from a Rs. 29 crore loss to a Rs. 246 crore profit.
Exports rose from Rs. 210 crore to Rs. 1,063 crore over the same period. Its total order book reached Rs. 26,665 crore, although that includes telecom networks and other businesses, not just AI.
For data centre connectivity specifically, HFCL is expanding manufacturing capacity fivefold and expects more than Rs. 700 crore of revenue in FY27. In September, it approved additional fibre and cable expansion, taking planned spending on those manufacturing projects to roughly Rs. 1,800 crore. That is planned investment, not completed expansion. Its hollow-core fibre work with an IIT Delhi-led consortium is also still a technology-development effort rather than established sales.
Source: HFCL
So the fibre boom has numbers behind it: stronger exports, bigger orders and much better quarterly profits. Yet a 1,000 percent stock rally and billion-dollar agreements raise another question. Are investors paying for the revenue these companies earn today, or the far bigger businesses management hopes to build tomorrow?
The AI IPO Frenzy
The AI excitement did not stop with already listed companies. By 2026, investors were looking for the next data centre winner in the IPO market, and ESDS Software Solution became an example of how quickly expectations could run ahead of results.
Its Rs. 720 crore IPO was subscribed 135.88 times. On September 4, its shares listed at Rs. 757 on the NSE, a 76.46 percent premium over the Rs. 429 issue price. The stock then rose as much as 146 percent from its listing price in just 13 trading sessions, before falling around 26 percent following its September results, as of October 5.
Source: Bloomberg
But what were investors actually buying? ESDS was no newcomer. Its business includes cloud infrastructure, data centres, software and managed services. According to its RHP, FY26 revenue rose from Rs. 361.34 crore to Rs. 472.21 crore, while net profit more than doubled from Rs. 55.61 crore to Rs. 120.82 crore.
It planned to use Rs. 576 crore of IPO proceeds to purchase and install cloud computing equipment and other infrastructure. However, the prospectus said the relevant equipment orders had not yet been placed when it was filed.
The first quarterly results published after listing painted a less dramatic picture. In Q1FY27, revenue grew only 7.28 percent to Rs. 133.66 crore, while net profit increased 14 percent to Rs. 29.28 crore. Infrastructure-as-a-Service contributed around 51 percent of sales, but this includes existing cloud and colocation work, not just AI GPUs.
Management also explained that FY26 had benefited from roughly Rs. 85 crore of one-time technical design revenue related to the Sharon AI project. This was not regular GPU rental income.
That distinction matters because the biggest AI business was still waiting to start. Under its arrangement with Sharon AI, ESDS was preparing to offer 8,200 Nvidia GPUs from a Sydney facility, but deployment had slipped by around one to 1.5 months.
Management expected the project to begin generating recurring revenue in Q3FY27, around November. The company had received roughly Rs. 1,100-1,200 crore in customer advances against the project, but advances are not the same as revenue already earned. These advances also helped lift FY26 operating cash flow to Rs. 1,368 crore.
Management also spoke of a potential pipeline exceeding 50,000 GPUs and a domestic order book of around Rs. 3,000 crore spread over three years. Yet it admitted that only a small number of GPUs had gone live by Q1, while delivery shortages could cause months of delay. It planned to deploy around 1,500 of its own GPUs by Q4FY27 and spend another Rs. 1,500 crore on capital expenditure during FY27. Those remain execution targets.
Meanwhile, other companies were also lining up. Sify Infinit Spaces had reportedly postponed its proposed Rs. 3,700 crore IPO despite operating 14 data centres with 188 MW of installed IT capacity as of June 2025.
Yotta Data Services was looking to raise up to USD 1.5 billion through a possible January-March 2027 IPO, having announced a USD 150 million private capital raise at a valuation of about USD 3.9 billion.
Source: Kotak Neo
So the IPO frenzy was not built on imaginary businesses. ESDS had real revenue, Sify had operating data centres, and Yotta was looking to raise fresh capital. But there was a catch.
Investors were already putting big price tags on GPUs that had yet to switch on, projects that had yet to be built and profits that had yet to arrive. In India's AI IPO rush, it seemed the valuations were running at full speed while the GPUs were still waiting to boot up.
One Year Later, Where Is The Actual Money?
A year ago, India's AI boom was largely a race of announcements, ambitious targets and soaring share prices. Today, the picture looks quite different. The promises haven't disappeared, but the numbers have started separating companies that are actually benefiting from AI from those still preparing for the opportunity.
On one side, there are companies such as Netweb Technologies, E2E Networks and Sterlite Technologies. Netweb's AI systems now contribute 62 percent of its quarterly revenue, E2E has moved from losses to profits as its GPU business expanded (Earnings are still volatile).
Sterlite's data centre revenue contribution has climbed from just 1 percent in FY26 to 21 percent in Q1FY27. These are actual sales, not just management projections. HFCL has also reported a sharp improvement in earnings, although its business extends well beyond AI.
But the picture changes when we look at the big companies building the data centres themselves. AdaniConneX has more than 960 MW of tied-up capacity, but only 65.4 MW was operational as of June.
Airtel has an existing base of around 120-130 MW against its 1 GW ambition, while Lodha has completed land sales but has yet to generate rental income from its proposed powered-shell business. Reliance and TCS have major partnerships, but their biggest projects are still awaiting commissioning.
That does not make these announcements empty promises. Building such infrastructure takes years, and several projects remain within their original timelines. However, there is a big difference between signing an agreement, spending money on construction and actually earning revenue from a completed facility.
The stock market, meanwhile, has moved much faster. Sterlite's extraordinary rally and ESDS's spectacular listing showed just how much investors were willing to pay for future growth. Yet ESDS reported only 7.28 percent revenue growth in Q1FY27, with its biggest GPU project still awaiting commercial operations.
So, is India's AI boom built on promises rather than numbers? Not entirely. The numbers are beginning to arrive, but they are arriving at very different speeds. Some companies have already built profitable businesses around AI, while others are still turning billion-dollar ambitions into physical infrastructure.
India's AI boom is real. But a significant part of the excitement still rests on what companies are expected to earn tomorrow, rather than what they are earning today.