The Economics of Data Centres
Tracing the shift from manufacturing-led development to computation-led economic growth.
From Factory Floors to Server Floors
For most of modern economic history, industrialisation had a remarkably simple formula.
If a state wanted to become richer, it needed factories.
Factories converted land into industrial estates, labour into wages, and wages into consumption. Every new manufacturing plant created a small economic ecosystem around itself. Workers rented homes, children attended schools, restaurants flourished, transport companies expanded, and local businesses multiplied. Economists call these multiplier effects, but most people simply know them as jobs creating more jobs.
This is why governments fought fiercely to attract automobile manufacturers, textile companies, steel producers and electronics firms. Tax breaks, subsidised land, cheaper electricity and faster approvals became standard tools in the competition between states.
The prize wasn’t merely the factory itself.
It was the entire local economy that grew around it.
Today, however, something unusual has happened.
The world’s fastest-growing industry often doesn’t manufacture anything you can touch.
Instead, it manufactures computation.
As our lives increasingly move online—from shopping and banking to entertainment and artificial intelligence—the modern economy runs less on assembly lines and more on server racks. The new industrial asset isn’t a steel furnace.
It’s a data centre.
The Factory Hasn’t Disappeared. It Has Changed What It Produces.
At first glance, calling a data centre a factory sounds strange.
After all, nothing comes out of it wrapped in cardboard boxes.
But that’s because we’re still thinking like an industrial economy.
Traditional factories took physical inputs—iron ore, cotton or crude oil—and transformed them into physical products with higher economic value.
Data centres perform exactly the same economic function.
Only their raw material is different.
Instead of iron ore, they receive billions of pieces of digital information every second. Searches, payments, videos, GPS signals, business records, medical scans, emails, photographs and AI prompts constantly flow into these facilities.
Inside, this data is stored, organised, analysed, secured and transmitted almost instantly across the world.
In other words, the product isn’t steel.
The product is computation.
And in an economy increasingly driven by software and artificial intelligence, computation has become as economically important as manufacturing once was.
Not All Data Centres Are Built for the Same Job
The phrase data centre sounds deceptively simple, almost like calling every building a house.
In reality, today’s digital infrastructure resembles an entire city, with different buildings performing completely different functions.
At the foundation are cloud data centres, the everyday workhorses that keep businesses running. They host office software, websites, enterprise applications and cloud storage—the invisible backbone that powers much of the modern internet.
Above them sit hyperscale data centres. These are industrial-scale facilities operated by global cloud giants, containing tens of thousands of servers designed to support millions of users simultaneously. If cloud centres are apartment buildings, hyperscale facilities are entire metropolitan cities.
Then comes the newest and fastest-growing category: AI compute centres.
These facilities are purpose-built for graphics processing units (GPUs), the specialised chips that train and run artificial intelligence models. Unlike conventional servers, GPU clusters consume extraordinary amounts of electricity while producing immense heat, making power availability and advanced cooling systems just as important as computing hardware itself.
Finally, there are edge data centres. Rather than serving entire countries from one location, these smaller facilities are positioned close to users. Whether it’s autonomous vehicles making split-second decisions, financial traders executing transactions in milliseconds or factories relying on real-time automation, physical distance suddenly becomes an economic cost. Edge centres reduce that delay, making digital services faster and more reliable.
Together, these layers form the digital infrastructure on which the modern economy increasingly depends.
And unlike factories of the past, their most critical raw materials are no longer coal, steel or labour.
They are electricity, fibre-optic connectivity, land, cooling systems and uninterrupted power.
The factories may have changed.
But the competition between states has only intensified because in the age of artificial intelligence, the ability to host computation may matter just as much as the ability to manufacture goods.
The New Competition Isn’t for Factories. It’s for Digital Gravity.
If factories were the magnets of the twentieth century, data centres are becoming the magnets of the twenty-first.
Think about how a shopping mall works. People don’t visit because every store is equally attractive. They visit because one or two big anchor stores draw the crowd, and hundreds of smaller businesses benefit simply by being nearby. Remove the anchor, and the footfall disappears.
The digital economy behaves in much the same way.
A large hyperscale data centre rarely arrives alone. It pulls in cloud service providers, AI startups, cybersecurity firms, software companies, telecom operators, semiconductor designers, Global Capability Centres (GCCs), and thousands of highly skilled professionals. Once enough of these businesses cluster together, the ecosystem begins feeding itself. More companies attract more talent. More talent attracts more investors. More investors fund more startups. Economists call these agglomeration effects. Everyone else simply calls them “the place where all the opportunities are.”
This is precisely why states have become so eager to host data centres. They are not chasing buildings filled with servers. They are chasing the economic gravity those servers create.
Why Are States Giving Away So Many Incentives?
At first glance, the incentives seem almost excessive.
Governments are offering subsidised land, waiving stamp duty, reducing electricity taxes, providing faster approvals, and even helping finance these projects through interest subsidies. For a building that may employ only a few hundred permanent workers, that appears to be a surprisingly generous bargain.
Until you realise what governments are actually buying.
A century ago, a state wanted an automobile factory because it created manufacturing jobs.
Today, a state wants a data centre because it creates digital infrastructure.
The objective has quietly shifted from creating employment directly to creating an environment where future industries choose to locate.
It’s the same logic that once made ports, highways and airports transformational investments.
A port doesn’t manufacture anything.
Yet cities with good ports historically became centres of trade because every merchant preferred to be close to them.
Similarly, a data centre doesn’t write software, build AI models or create fintech companies.
But software firms, AI startups, banks and cloud businesses increasingly prefer locating near high-quality digital infrastructure because it reduces costs, improves speed and increases reliability.
In economics, infrastructure rarely creates value by itself.
It creates value by making everyone else more productive.
The New Industrial Race Is Already Underway
This shift isn’t theoretical anymore. It is visible in the industrial policies of almost every major Indian state.
Instead of competing over automobile assembly lines, they are now competing over megawatts, fibre connectivity and GPU clusters.
Uttar Pradesh has perhaps made the boldest statement of intent. The state aims to attract more than ₹2 lakh crore of private investment while building nearly 2 GW of data centre capacity. Beyond conventional incentives such as subsidised land, stamp duty exemptions and interest subsidies, Uttar Pradesh has introduced dedicated AI Compute Booster incentives aimed specifically at GPU-intensive facilities. Greater Noida has consequently emerged as one of India’s fastest-growing digital infrastructure hubs, attracting large campuses from Yotta Data Services and AdaniConneX.
Maharashtra, already India’s financial capital, wants to remain its digital capital as well. Under its Developed Maharashtra 2047 vision, the state is targeting 30–40 GW of green-energy-backed data centre capacity over the coming decades. Large green data centre parks investing ₹60,000 crore or more are eligible for industrial incentive grants covering up to 75% of capital investment, alongside interest subsidies. The Mumbai–Navi Mumbai region already hosts more than 790 MW of operational capacity, with further expansion led by players such as CtrlS, Yotta and Adani Infra.
Meanwhile, Telangana is betting that Hyderabad can become India’s AI capital. The state expects to add 1,000–1,200 MW of capacity by 2030, supported by single-window clearances, subsidised power costs and industrial land allotments. Hyderabad’s ambitions received a major boost with Neysa Networks announcing a $1.18 billion, 400 MW AI-ready facility, alongside continued investments from global operators such as NTT.
Further west, Gujarat is taking a different approach. Rather than focusing only on conventional cloud infrastructure, it is leveraging its long coastline and renewable energy ecosystem. The state targets 2 GW of capacity while offering incentives for desalination infrastructure and renewable-powered facilities. Jamnagar, already known for one of the world’s largest refinery complexes, could soon house Reliance Industries’ planned multi-gigawatt AI data centre powered entirely by renewable energy.
Perhaps the most ambitious vision comes from Andhra Pradesh. The state has set an astonishing target of 6 GW of hosting capacity by 2030, supported by full reimbursement of state GST during construction, discounted land prices and electricity tariff concessions lasting fifteen years. Visakhapatnam is expected to anchor this strategy, with Google’s proposed $15 billion, 1 GW campus and Reliance’s planned ₹1.6 lakh crore, 1.5 GW cluster potentially transforming the city into one of Asia’s largest digital infrastructure hubs.
Even Karnataka, is planning to establish three sustainable data center parks with a combined capacity of one gigawatt (GW), distributed across Bengaluru, Mysuru, and Mangaluru. This strategic move aims to solidify the state's position as a key hub for digital infrastructure in India.Bengaluru, a prominent technology hub, is slated to host a 500 megawatt (MW) data center park near Hoskote. The state government has committed to supplying solar power generated at Pavagada directly to this park.
Looking across these policies, one pattern becomes obvious.
Every state is competing using slightly different tools cheaper electricity, faster approvals, subsidised land, renewable energy or AI-specific incentives.
But they are all chasing the same prize.
The next generation of economic growth.
Why a Data Centre Creates More Value Than Its Own Revenue
The obvious criticism is also the most reasonable one.
“If these buildings employ so few people, why are governments investing so heavily in them?”
Because their biggest contribution isn’t what happens inside the building.
It’s what becomes possible outside it.
Economists call these spillover effects.
Imagine trying to build India’s next AI startup while hosting your computing infrastructure thousands of kilometres away. Every request travels farther. Every millisecond adds delay. Every expansion becomes more expensive. Now imagine the same infrastructure located within your own state, connected through high-speed fibre, powered by reliable electricity and supported by local cloud providers.
Suddenly, the cost of innovation falls.
This is exactly why data centres matter.
They reduce latency for businesses, lower cloud computing costs, strengthen cybersecurity, improve service reliability and make it easier for startups to scale. They also make a region significantly more attractive for multinational companies establishing Global Capability Centres, research facilities and engineering teams.
The impact is already becoming visible across India’s digital economy.
Digital services currently contribute roughly 12% of India’s Gross Value Added, but several estimates suggest this could rise to nearly 20% by FY30. That growth cannot happen on software talent alone. Every AI model trained, every fintech transaction processed, every streaming video delivered and every enterprise application hosted ultimately depends on the invisible computing capacity sitting inside these data centres.
Consider India’s own digital infrastructure.
The Unified Payments Interface (UPI) now processes over 630 million transactions every day. Every payment triggers authentication, fraud detection, account verification and settlement within seconds. Add streaming platforms, e-commerce, cloud gaming, digital banking, telemedicine and artificial intelligence, and the scale of computing required becomes almost unimaginable.
The data centre, therefore, is not just another real estate project.
It has quietly become one of the most important pieces of economic infrastructure in the digital age.
Just as reliable electricity powered the Industrial Revolution, reliable computation is increasingly powering the Intelligence Revolution.
And that explains why states are no longer asking, “Which factory can we attract?”
They are asking a far more future-oriented question:
“Can we become the place where the digital economy chooses to think?”
But What Happens When the Digital Factory Moves Next Door?
So far, the story sounds almost irresistible.
Data centres bring investment, strengthen digital infrastructure, attract technology companies and position cities for the AI economy.
If that’s true, why would anyone oppose them?
Interestingly, some of the strongest resistance isn’t coming from developing countries worried about growth.
It’s coming from some of the world’s richest economies.
Across the United States and Europe, communities are increasingly asking a question that economists love but politicians often avoid:
Who gets the benefits, and who bears the costs?
Because while the internet feels weightless, the infrastructure behind it is anything but.
Every new data centre occupies land, draws enormous amounts of electricity, consumes water, requires transmission lines, installs massive diesel backup generators and changes the character of neighbourhoods around it.
The cloud, it turns out, casts a surprisingly large shadow.
When Digital Growth Meets Local Reality
One of the growing concerns around large data centres is the gap between national benefits and local costs. While governments and companies highlight billions of dollars of investment and improved digital infrastructure, nearby residents often worry about more immediate issues:
Whether electricity supply will become less reliable, whether water resources will come under pressure, whether industrial cooling systems will create constant noise, and whether open land will be replaced by large concrete facilities.
Economists describe these concerns as externalities costs generated by an economic activity but borne by people who may not directly benefit from it. Data centres can create several such externalities at the same time.
Large facilities may strain local electricity grids, increase demand for scarce water resources, require regular testing of diesel backup generators, generate mechanical noise from cooling equipment, and alter local ecosystems through land conversion and thermal discharge.
As a result, the broader economic gains may be national, but the inconveniences are often experienced most intensely by local communities.
Virginia: The World’s Largest Data Centre Hub Is Beginning to Push Back
No place illustrates this tension better than Northern Virginia, often called the world’s data centre capital.
The region hosts more than 5,000 MW of operational capacity, earning it the nickname “Data Center Alley.” Almost every major cloud provider has a presence there.
For years, the growth seemed unstoppable.
Then local communities began asking difficult questions.
One of the biggest controversies surrounds the proposed Prince William Digital Gateway, a project covering nearly 2,000 acres. Residents have opposed the development not because they oppose technology itself, but because of concerns over its location near historic Civil War battlefields, its impact on local watersheds and the sheer visual transformation of the landscape.
Others complain about something far less glamorous than artificial intelligence.
Noise.
Large cooling systems operate continuously, while backup diesel generators—some as large as railway carriages—must be tested regularly to ensure reliability during power failures. Near residential areas, sound levels can reportedly reach 85 to 100 decibels, comparable to standing beside heavy traffic or industrial machinery.
The irony is striking.
The infrastructure powering our quiet online lives can be remarkably noisy in the physical world.
Ireland: When Servers Consume More Power Than Homes
Ireland offers another fascinating lesson.
Over the past decade, the country became one of Europe’s preferred locations for data centres, thanks to favourable tax policies, cool weather and strong international connectivity.
Success, however, created a new problem.
By 2023, data centres accounted for roughly 21% of Ireland’s metered electricity consumption more electricity than all urban households combined.
That statistic fundamentally changed the public debate.
The question was no longer whether data centres created investment.
The question became whether the electricity system could continue supporting unlimited expansion without affecting everyone else.
Eventually, Ireland’s grid operator, EirGrid, imposed restrictions on new data centre connections around Dublin until at least 2028, arguing that the local electricity network simply needed time to catch up.
Environmental groups added another layer to the debate, challenging policies that allowed large facilities to rely on natural gas backup generation and questioning whether rapid digital expansion was compatible with the country’s climate goals.
The lesson wasn’t that data centres are undesirable.
It was that infrastructure planning cannot be an afterthought.
The Biggest Input Isn’t Data. It’s Electricity.
When most people imagine a data centre, they think about servers.
Rows and rows of blinking machines storing photos, streaming movies and answering AI questions. But if you asked the engineers running these facilities what keeps them awake at night, many wouldn’t say servers.
They’d say electricity.
Because a data centre has a surprisingly simple business model.
It buys electricity and converts it into computation. Everything else is a consequence of that one transaction.
The processors consume power to perform billions of calculations every second. The memory modules need power to keep information instantly accessible. Then comes an invisible but equally important consumer of electricity—the cooling systems. Every watt consumed by a processor eventually turns into heat, and unless that heat is removed continuously, the world’s smartest AI chip quickly becomes an expensive paperweight.
You can think of a modern data centre as two factories operating inside the same building.
One factory performs computation.
The other works tirelessly just to stop the first one from melting.
Why Cooling Has Become as Important as Computing
Imagine running your laptop on a hot summer afternoon. Within minutes, the fan starts spinning louder. The harder you push the processor, the hotter it gets. Now multiply that laptop by hundreds of thousands.
That’s essentially what happens inside a hyperscale data centre.
Electricity enters the building from the grid and immediately splits into two enormous streams. One powers the computing equipment—processors, GPUs, memory and storage devices that perform the actual work. The other powers the infrastructure supporting them: industrial chillers, cooling towers, fans, pumps, transformers, UPS systems and backup equipment that ensure the servers never overheat or lose power.
This explains why electricity dominates the economics of a data centre.
Over its lifetime, power typically accounts for nearly 50% to 60% of the facility’s operating costs, making it by far the single largest recurring expense. Unlike many businesses that can temporarily slow production when demand weakens, a data centre cannot simply switch itself off. The internet doesn’t close at night, cloud services don’t take weekends off, and AI models don’t stop training because electricity prices have risen.
A standard 100 MW hyperscale data centre consumes electricity continuously throughout the year, requiring nearly 876,000 megawatt-hours (MWh) annually. To put that into perspective, that’s roughly enough electricity to power around 100,000 urban Indian households for an entire year.
In the digital economy, computation doesn’t just require power.
It requires uninterrupted power.
How Power-Hungry Are Data Centres, Really?
This naturally raises an important question.
Are data centres simply the new steel plants when it comes to electricity consumption?
The answer is both yes and no.
An integrated steel plant still consumes far more electricity overall because steelmaking itself is an extraordinarily energy-intensive industrial process. A large 3 million-tonne-per-year steel plant can consume nearly 17.5 million MWh annually, requiring a continuous power load approaching 2,000 MW.
But here’s where data centres stand apart.
Their electricity demand is concentrated inside a single building.
A 100 MW hyperscale data centre consumes nearly 876,000 MWh every year, substantially exceeding the electricity needs of many medium-sized manufacturing facilities. For comparison, a typical automobile assembly plant producing around 200,000 vehicles annually consumes roughly 121,000 MWh, while a modern 1 million-tonne cement plant may require around 89,000 MWh of electricity.
In other words, one large data centre can demand several times the electrical energy of factories that employ thousands of workers.
India’s Next Infrastructure Challenge May Be the Power Grid
This shift is already visible in India’s electricity projections.
In 2024, data centres accounted for only about 0.5% of India’s total electricity consumption, using roughly 13 terawatt-hours (TWh) of power.
That number sounds comfortably small.
It won’t remain that way for long.
According to projections from the Ministry of Power, electricity demand from India’s data centres could rise to 13.56 GW of connected load by 2031–32, accounting for nearly 8% of all additional peak electricity demand expected during this period.
Notice the phrase peak demand.
Power systems are not built around average consumption. They are designed for the moments when everyone wants electricity simultaneously.
As clusters of AI data centres emerge around cities like Mumbai, Hyderabad, Chennai, Noida and Visakhapatnam, local electricity networks face an entirely new engineering challenge: supplying massive amounts of reliable power twenty-four hours a day without interruption.
This is one reason why governments are no longer thinking about electricity and digital policy separately.
Increasingly, they are becoming the same policy.
The Biggest Criticism: Where Are the Jobs?
The biggest criticism of data centres is also the simplest: where are the jobs? For decades, governments judged industrial projects by the number of people they employed. Textile mills hired thousands, automobile plants created hundreds to thousands of factory jobs, and steel plants often supported entire industrial townships. Data centres, despite attracting massive investments, break this traditional relationship between capital and employment.
A ₹10,000 Crore Investment, But Few Permanent Jobs
A hyperscale data centre may require investments of ₹10,000 crore or more, demanding land, electricity, and generous government incentives just like any other large industrial project. During construction, these facilities generate substantial employment. A $10 billion (around ₹83,000 crore) data centre campus can support more than 4,000 temporary construction jobs, employing civil engineers, electricians, HVAC specialists, fibre installers, and thousands of construction workers. However, once operations begin, automation takes over. Most facilities require only 200–300 permanent employees, largely comprising network engineers, cloud architects, cybersecurity specialists, electrical engineers, and facility managers. In effect, every ₹275 crore invested creates roughly one permanent operational job, making data centres among the least labour-intensive investments in the economy.
Looking Beyond Direct Employment
The contrast becomes even sharper when compared with other industries. An investment of ₹10,000 crore in a hyperscale data centre typically creates only 36–50 direct operational jobs and 270–370 indirect jobs. The same investment in an automobile plant can generate 850–1,000 direct jobs and another 4,000–5,000 jobs across suppliers, logistics, and dealerships. An integrated steel plant supports around 8,400 direct jobs, while the textile and apparel sector can employ hundreds of thousands of workers. Given that land, electricity, and water are scarce resources, the obvious question is why governments continue to compete so aggressively for one of the least labour-intensive industries.
Governments Are Investing in an Ecosystem, Not Just Jobs
The answer lies in shifting the focus from direct employment to economic spillovers. Governments are no longer asking how many people work inside a data centre; they are asking how many industries become possible because it exists.
A data centre may employ only a few hundred people directly, but it enables a much larger digital ecosystem. It lowers cloud computing costs, reduces internet latency, attracts Global Capability Centres (GCCs), supports AI startups, improves banking infrastructure, strengthens cybersecurity, helps SaaS companies scale, encourages university-industry collaboration, and attracts both venture capital and skilled talent.
Conclusion:
The contest for data centres is, at one level, a contest over the shape of future growth. States are no longer merely trying to attract factories that employ thousands of workers on assembly lines; they are trying to attract the infrastructure that will power artificial intelligence, cloud services, digital finance, and the next generation of high-value industries.
And yet, this transition brings a profound trade-off into view. Data centres promise digital competitiveness, foreign investment, and technological ecosystems, but they also demand vast quantities of electricity, water, land, and public incentives while generating relatively few direct jobs. The question is not whether computation matters; it clearly does. The harder question is how much of a state’s scarce resources should be devoted to enabling that computation, and what other opportunities are forgone in the process.
Perhaps the real shift is philosophical. In the industrial age, governments competed to host places where people made things. In the intelligence age, they are competing to host places where machines process information. One model converts labour into wages; the other converts electricity into insight. Both can create prosperity, but they distribute costs and benefits in very different ways.
The future may belong to regions that can balance these two imperatives: building digital infrastructure without neglecting employment, managing resource consumption without slowing innovation, and ensuring that the gains of the AI economy are shared more broadly than the ownership of the servers that power it.
So the questions to ponder are difficult ones. Should public policy reward industries for the number of jobs they create today, or for the productivity they may unlock tomorrow? Can a state become digitally rich while remaining socially inclusive? And when the factories of the future run on electricity and data rather than labour and raw materials, what will economic development mean for the ordinary citizen?
SourceS- InvestIndia, KPMG, DCNTGLOBAL, IEA, PIB, FinancialExpress, USEIA, CanaraBank, TheGaurdian, QZ, ET, TimesIndia, CareRating, HT, assocham, Datacentre.com.
The ₹275 crore per permanent job number deserves more weight than the piece gives it because it exposes a deeper structural problem. The multiplier effects that justify the incentives, the ecosystem, the startups, the GCCs, the talent, only materialise in cities that already have them. Mumbai, Hyderabad, Bengaluru, Noida. The same agglomeration logic that makes data centres attractive ensures they cluster where digital infrastructure already exists.
So the states competing hardest for data centres are the ones that need them least. The states that need economic development most can't win the competition because they lack the grid capacity, the fibre connectivity, and the talent pool that data centres require to operate. The development gap between India's digital hubs and its manufacturing states doesn't narrow with this model. It widens. Data centres don't distribute growth. They concentrate it. That's the trade-off the incentive frameworks haven't priced in.