Daybreak - India's $60 billion data-center bet, and the jobs it won't actually create

Episode Date: August 20, 2026

Every Indian state wants to be the country's data-centre capital. Gujarat, Andhra Pradesh, Maharashtra. They're all handing out land, power, decades of tax breaks, all chasing the same promise: jobs.The Ken reporters Inderpal Singh and Mrunmayee Kulkarni ran the numbers on what the jobs actually cost. And the math does not hold up.On this episode of Daybreak, we go inside what's causing the data-centre boom. Why these buildings are designed to need almost no one, what states are actually betting on, and if the data-centre industry will see the same ecosystem build out as other sectors in India.Daybreak is produced from the newsroom of The Ken, India’s first subscriber-only business news platform. Subscribe for more exclusive, deeply-reported, and analytical business stories.

Imagine the last part of Terminator Judgment Day where he's in an empty building with a lot of computers, that's a data center. It's a very cold building. It's very silent. No desks or no chairs here, just racks of servers. You don't really have a lot of windows, so it's dark all the time. It's under a really harsh glare of LED lights. At one corner, I could just hear water just sort of moving through different pipes. and everywhere you look there is this little hum.

The kind of hum you have in like hotel rooms basically. When it's all closed up, the windows don't open and the AC hums at a very weird angle. You just heard my colleague, the Ken reporter Munnikoal Kani. And here, she is talking about her experience from inside a data center. The infrastructure that powers the artificial intelligence capacity of LLMs that you and I both use, like Chad GPD, Claude and Gemini. And as you heard, a data center isn't exactly the most welcoming of places.

My colleague Indarpal Singh, who worked on the story with Nunmay, in fact describes it as a space that is resentful of human presence. But it's still the kind of facility that almost every Indian state is rushing to build. So much so that some are nearly emptying out their purses to build out these structures. India itself is targeting 10 gigawatts of this capacity by 2030, up from about 1.57 gigawatts today. That's the race that these states are currently locked in. Gujarat announced its data center policy last month, quoting about 6,000 crore rupees in investment. Meanwhile, Andhra Pradesh has managed to bring in reliance and Google by offering more than 40,000 crore rupees in incentives. And that's at the stage where the state is already in a revenue deficit.

But what exactly are they even buying? Mostly the promise of jobs. And to be fair, the promise does look shiny. Government officials have talked about nearly 100,000 new engineering jobs by 2030, created by data centers. But that kind of sounds like a pretty loaded promise, right? Especially for what is mostly an empty building? That's the question that prompted my colleagues to take a closer look at the numbers and see what exactly each of these promised jobs would actually cost the state.

In this episode, you will hear from both of them. Because turns out, the math isn't really in the state's favor. Welcome to Daybreak, a business podcast from the Ken. I'm your host, Richard Verguise, and every day of the week, my co-host, Nikaa and Sharmana will bring you one new story that is worth understanding and worth your time. Today is Friday, the 21st of August. The data centers built for AI as we know it today look like the set of Terminator by design.

The safety and security of all of the sensitive pieces of data traveling from server to server pretty much demand this kind of a deserted condition. Here, Sandhapal, explaining why in more detail. A modern data center is, I think, a very lonely place, perhaps the loneliest of all. because there's a lot of risk involved here because this is somebody's data and you can't risk breaching it and every extra person in a data center facility means extra breach risk that you have. And a fun fact I came across during my reporting was that a lot of data centers in India at least are hiring retired army officers for security.

So that should tell us about how secure these places need to be. that's why one of the people I interviewed for the story, Wimal Kaur, he's the CEO of the largest data center company in India. So he told me that he also doesn't have access to a lot of his own data centers in India. Imagine, Kaur runs 18 data centers in India. And even then, he and the other people who are in charge of security who are already on the data center's payroll are rarely allowed inside. that seems to drastically reduce the number of jobs a data center opens up, right? So I asked Indyarpal, what exactly is the pitch that companies are making to the states?

So a lot of these pitches which Murdoin Bha talked about are packaged with a promise of jobs. Recently, one of the union ministers said that the data center sector in India would create 1,000 lack engineering jobs. Data center jobs are different from traditional jobs. I mean, it's a little apples to oranges comparison here, but I mean, it takes a lot of investment per job in data centers compared to something like, say, textiles or something like, you know, mobile manufacturing or software. Indarpal mentioned textiles and mobile manufacturing exactly because they are industries which have historically. turned investment into employment.

And in comparison to these industries, data centers do look like a more extravagant way to buy jobs. Let me explain. You see, the textile sector takes about a few thousand dollars in CAPEX to create a new job. Mobile manufacturing costs a bit more at about a hundred thousand dollars. A data center in India, on the other hand, the Ken's calculations show, takes as much as 4.5 million dollars. To put that into perspective, let's take a look at how much exactly India's total investment is likely to be. Kaur and Narendra Sane, the founder of Neve Cloud, which is a smaller data center operator, helped the Ken with this breakdown. You see,

one megawatt of capacity costs about $5 to $6 million to install. If you scale that up to the targeted 10 gigawatts in the next four years, the total bill comes up to about $60 to $70 billion. And here's where exactly all the money goes. So about $53 billion of this goes into the core infrastructure equipment, IT, data center equipment, and then about $300 million to $1 billion are reporting short. This goes into chip design and engineering. So where exactly do the data center specific jobs come in? Mostly at the beginning.

It takes about two to four years to set up a center. And according to a study from University of South California, late last year, a data center in the U.S. needed less than one to two workers per megawatt. For India's target, that same number works out to a peak construction workforce of 26,000. Which, as the study notes, means that by the time the building is up, the jobs are gone. The picture is slightly different in India, though, as Indarpal explains. we try to do a little back of the envelope calculation for this. So for 100 megawatt of data center creates about 100 to 150 jobs.

So you scale that to a targeted 10 gigawatt capacity across the country. That translates to about 15,000 full-time sustainable jobs. But then that's a little misleading number because we also have to look at it from an ecosystem angle. So, when you put the entire picture together, you get an inverted pyramid. So when a data center is being erected, I mean, our calculation showed that you need about 60 to 70,000 people, but then this is not permanent labor. This lasts only as long as the data center is being constructed.

And then slowly when you move to the testing phase and the commissioning phase, this starts dwindling and then once the data center is live and up and running, you need a very small bunch of people to run them. 15,000 full-time sustainable jobs. That's it. That's all $60 billion is buying. Essentially, India is being sold the top of the inverted pyramid that Indirpal mentioned. But once the money is invested and the data centers are actually live, India will be left with

just the bottom of this upside-down pyramid. which raises the obvious question. If the math for the jobs isn't adding up, why are states so eager to invest? More on this in the next segment. Every Indian state that wants a data center makes some version of the same pitch. She planned a discounted power

and a tax exemption for 15 to 20 years. And now, they are even competing with each other quite publicly. Karnataka's IT minister, Priyang Kareg, He called out the Andhra Pradesh government for providing so much money to Google in tax exemptions when the state itself is in debt for a lot of years. Minister Kharki's public callout is actually a reference to a real fact. You see, despite its massive investment, which like I mentioned earlier, is about 40,000 croix and incentives to Google and reliance,

Andra does not have that amount of cash to burn. Because its revenue deficit crossed 110% of its budgeted estimate in just the first of, first quarter of FY25. Still, this has not dampened the state's need to one-up each other. Part of the problem is the fact that data centers are just, well, new as a concept. You see, when the state pitches exemptions and cheap plan to businesses, in return they get a business plan with employment generation projections. For any other industry, the state would check the plan against all the years of

precedent they have. But a data center doesn't come with that kind of precedent. The result, as an official at the Andhra Pradesh Industrial Infrastructure Corporation, explained, is that the state's incentives are then tied to milestones like capital invested and infrastructure completed. When I started figuring out which state had what incentive, it really felt like the beginning of a fairy tale to me. So if, I mean, it's almost like once upon a time, every Indian state wanted to build multiple smart cities. and it looks like this year everyone wants to be the data center capital of the country. So if you guys remember in July, like last month this year,

HCL Tech announced that it would build a new AI center in Gujarat's GIF city with Sarvam. And in the same week, Gujarat had unveiled its own data center policy. It was like 6 lakh crores in investment. That was their target. And they wanted to build about 7.5 gigawatt of capacity by 2029. Every week there is a new state and it signs a new MOU and it's almost always bigger than the state which was before it. Like Gujarat signed a big MOU. It's bigger than Andra, which was bigger than Marashtra,

which was bigger than Tamil Nadu, which was way bigger than Karnataka. Like an official at the Andra's IT department told me that if a state announces a major policy, we study it and we try to top it. Basically land, land, power, tax exemptions, interest subsidies. It's a situation where every state wants to be more generous to the industry. than the states before it. Right now the situation is that data center projects by various companies are getting subsidies which are a little unheard of. Like, for example, the Reliance Project in Andhra Pradesh is getting subsidies of like $19,000.

The Google project is getting subsidies of $22,000. But even within the state, officials don't necessarily think that these data centers are going to generate as many jobs as they have promised, which means the bet is something else entirely. When I spoke to someone in the system, they said that nobody inside the government expects a hyperscaler campus to become a major mass employer once the construction is over. What they're doing essentially over here is building stronger digital infrastructure. They're building better cloud availability. They're sort of banking on creating an AI readiness that might pull in other businesses behind it. So the idea is basically an ecosystem.

The kind that other sectors in India have seen, like the automobile industry in places like Pune or Chakan. And a sort of example for the kind of scale possible for data centers does exist. Nunei found a comparison with a province in China, which surprisingly wasn't even trying to solve for jobs. It had decided to build something else entirely. Employment by itself is not a metric for incentives when it comes to data centers. An example over here is a province called Guizu in China, which was the country's first national

big data pilot zone in way back in 2016. What they did was they built a decade-long cluster strategy around data center infrastructures. They brought in Huawei, they brought in a lot of cable networks, and they built up, they ended up building a base beyond just data centers. They ended up building a system which other tech companies wanted to plug into. So by, like I said, it started in 2016, and by 2023, digital services made up nearly half of the province's GDP.

So Huawei's data center alone was serving more than 380 companies inside that province. So what it wanted to become was the computing hub that companies would build around. So basically, Guizu knew that it didn't just want to host a hyperscalor. And its clear plan was the reason why companies like Apple had a presence there, which ended up pulling in a Huawei and eventually even Qualcomm and so on until the zone became investable all by itself and had nothing to do with what kind of incentives the government was handing out. But Indian states don't really seem to have a plan like that. Just a series of independent deals negotiated with each player like Reliance, Google, Amazon and each of them negotiated

on their own terms. And none of them really working together. Stay tuned. So is there a chance that this could actually turn into an ecosystem like Visus? To answer this, Runmay compared the data center industry to the one I had mentioned earlier, automobile manufacturing, which did Ciat ecosystem build up around it in a similar manner. The crucial difference here, I think Rachel, is that automobile manufacturing didn't happen just because one company came and set up a megafactory and things like that. Basically, it was a company followed by a bunch of component makers, then a lot of engineering firms, there were testing facilities, there were skill centers,

there was an ecosystem that developed over here. I spoke to someone in the MIDC, that's the Marastra Industrial Development Corporation, and even they said that a data center ecosystem won't quite look the same as, say, a manufacturing or a semiconductor. She said what is emerging here is smaller, more service-shaped companies. Like, there would be companies specializing in electric systems. There'd be companies specializing in cooling infrastructure and fiber, and maybe facility management, cybersecurity, cloud operations.

But whether that becomes a cluster comparable to automobiles is something we look only know over the coming decades. But even if this is only going to play out over the coming decades, the cost itself is still going to show up sooner for the country. And mostly, in the form of tax revenues that they are not collecting. Taxes are also a very high cost here. So just this year in the budget, the finance minister Nirmala Sita Raman announced that we now have a tax holiday

for data centers in India till 2047. So that's almost a 20-year tax holiday for a sector. So according to some estimates, India is going to lose about 1.4 to 1.7 lakh crores over this period. We are not the only country which is losing revenue as tax exemptions. For example, Ohio recently reported a loss of $1.6 billion in taxes thanks to its data center tax holiday. and they had an installed capacity of like 230 megawatts. Now, the point here is that every state is going after an industry that neither gives India ownership nor labor while taking away our resources. For example, by industry estimates roughly 90 to 95% of the total installed capacity will be leased by global hyperscalers, which is Amazon, Microsoft, Google, Oracle,

META, leaving nibbles for domestic enterprise customers. So the pitch which is repeated across every state's data center policy is one of digital sovereignty, India's data on Indian soil, powering Indian industry. But on ground, India is supplying the land, the concessional power, the water and the tax breaks, whereas the hypers supply the servers, keep the software and own the customer. Basically, India's data sovereignty pitch, which is India's data hosted on India's data. soil, powering Indian industry, isn't getting fulfilled either. So it's not surprising that some states are pulling back from their commitments. Tamil Nadu, for example, has already done it.

In its 2021 data center policy, the state had signed 59 MOUs worth more than 35,000 true rupees. The policy expired in March this year and the state still has not renewed it. It's now even started moving to incentives for renewables and other. subsidies. Still, most other states are very firmly on the data center bandwagon. And that raises the final question. Is there a more lucrative way for states to make these deals? Runei's research does show some ideas. The government could take equity instead of just giving away revenue. It's literally what's happening with the India AI mission. For example, the government did give subsidy to Sarvam. It's not a data center, totally different. But it was structured as a

compulsarily convertible debenture and it converted to an actual equity stake when the company's valuation went up. So in case of a data center, if the state is going to forgo about 20,000 crore, for example, in electricity duty, converting some portion of that into equity

or a revenue share means that the state would participate in the upside if the data center is making money instead of just, you know, bearing the risk and gaining none of the rewards in it. They could charge for water and power at something close to the true cost and then negotiate everything else, basically.

See, hypers care a lot about guaranteed uptime and fast approvals and grid reliability rather than the raw discount. Nobody is really very clear or very vocal about what do you expect to happen after a data center is built. A little bit of foresight, I think, would be better. Rana is right. A little bit of foresight could always be better. But for now, construction is in process anyway.

If the states do listen to the criticism, they might even. even plan for what happens next. But until then, it's still a waiting game.