India's Nuclear Energy Journey — A Complete Base Article for UPSC Mains

Think about what this meant. India could not import uranium. India could not import reactor components. India could not buy nuclear technology from anyone. For over three decades — from 1974 to 2008 — India's nuclear programme was essentially under siege.

But here is the twist in the story. These sanctions, painful as they were, forced India to do something remarkable: build everything at home. Every component going into India's nuclear plants — the reactor vessels, the steam generators, the control systems, the fuel assemblies — was designed, tested, and manufactured in India. The Atomic Energy Commission worked in patient partnerships with Indian firms. India's reactor unit size went up from 200 MW to 500 MW, and now to 700 MW. The article we are reading today by former Ambassador Meera Shankar puts a stunning number on this: India now makes the cheapest nuclear power plants in the world, at approximately $1,700 per kilowatt. Compare that with South Korea at $2,200, France at over $5,500, and the United States at a jaw-dropping $15,000 per kilowatt.

This is the core of the indigenisation story. The world shut the door on India, and India built a window. Today, India has the potential to become a major exporter of nuclear power plants — precisely because it was forced to make them on its own.

Why did the US do this? Two reasons worked together. First, the US saw India as a rising democratic counterweight to China and wanted to deepen the strategic relationship. Second, India's massive and growing energy needs meant a huge market for American nuclear companies like Westinghouse and General Electric.

But making this deal happen was like threading a needle through a hurricane. It took three years and required clearing multiple hurdles, one after the other:

The India Nuclear Separation Plan (March 2006) — India agreed to separate its nuclear facilities into civilian and military categories. Only civilian facilities would be placed under international inspection. Military facilities (needed for India's nuclear weapons programme) would remain outside inspections. This was a difficult political compromise within India. Critics said India was giving up too much sovereignty.

The Hyde Act (December 2006) — The US Congress passed a special law authorising nuclear trade with India, overriding existing American legislation that banned such trade with non-NPT countries.

The 123 Agreement (August 2007) — Named after Section 123 of the US Atomic Energy Act, this was the actual bilateral agreement between India and the US, specifying the terms of nuclear cooperation.

The India-IAEA Safeguards Agreement (August 2008) — India signed an agreement with the International Atomic Energy Agency (IAEA) to place its civilian nuclear facilities under IAEA safeguards (inspections). India also signed an Additional Protocol allowing more intrusive inspections.

The NSG Waiver (September 2008) — This was the hardest part. The 45-member NSG (now 48) had to reach consensus to grant India a special exemption from its rules. After intense lobbying by the US, and resistance from countries like China, the NSG gave India a "clean and unconditional waiver" on 6 September 2008. This made India the only country with nuclear weapons that is NOT part of the NPT but is still allowed to engage in nuclear trade with the world.

The result? The three-decade international embargo on nuclear trade with India was over. India could now import uranium, buy reactor technology, and sign nuclear cooperation agreements with multiple countries. Since 2008, India has signed civil nuclear agreements with the US, France, Russia, Canada, South Korea, Mongolia, Kazakhstan, Argentina, Namibia, Australia, and others.

In 2010, India passed the Civil Liability for Nuclear Damage Act (CLNDA). This law had a provision (Section 17b) that allowed the operator of a nuclear plant (say, NPCIL) to sue the equipment supplier if a nuclear accident was caused by defective equipment. This was unusual. In most countries, all liability sits with the operator, not the supplier. This is the global norm under the Convention on Supplementary Compensation (CSC). American and French companies like Westinghouse, GE, and Areva refused to set up plants in India because they feared being dragged into lawsuits if something went wrong.

This supplier liability problem stalled Western reactor projects in India for over a decade. The Kudankulam plant in Tamil Nadu, built with Russian help, moved ahead because Russia's Rosatom accepted the liability terms. But Western companies stayed away. This is one of the key problems the SHANTI Act of 2025 was designed to fix.

Going from 8.78 GW today to 100 GW in 21 years means roughly a twelve-fold increase. This is an extraordinarily ambitious target — more than any other country is attempting proportionally. Meeting it will require not just building reactors fast, but building the entire ecosystem: fuel supply chains, trained manpower, regulatory capacity, safety culture, and massive capital investment.

In the West, SMR designs are mostly still under development; no commercial SMR is operating yet. India's approach is interesting. The DAE has offered its existing 200 MW reactor technology to new private sector entrants. India is also developing its own SMR designs — the Bharat Small Modular Reactor (BSMR-200) and the SMR-55. The Nuclear Energy Mission targets at least five operational indigenous SMRs by 2033.

The article by Meera Shankar makes a sharp point here: there would be little justification for deploying an untested foreign SMR design in India as a first-of-a-kind experiment. If an SMR has not operated satisfactorily for a few years elsewhere, India should not be the testing ground. This is the self-reliance argument applied to safety: use proven domestic technology first, let foreign designs prove themselves abroad before importing them.

Think of it this way. When you ask ChatGPT or a similar AI model a question, that query does not float in the air. It runs on thousands of specialised processors (GPUs) packed into massive server racks, housed inside data centres the size of football fields. These data centres need electricity — not sometimes, not during the day, but every second of every day, without a single interruption. A single AI server rack consumes five to six times more power than a conventional server rack. Training a large AI model can consume as much electricity as a small town uses in a year.

Now multiply that by the scale of what is happening. Globally, data centres consumed roughly 300–400 terawatt-hours (TWh) of electricity in 2022 — about 1–2% of all electricity used on the planet. The International Energy Agency (IEA) projects this could reach 945 TWh by 2030 and 1,200 TWh by 2035. To put that in perspective, 1,000 TWh is more than the total electricity consumption of Japan.

And India is right at the centre of this explosion. Despite hosting less than 5% of global data centres, India accounts for about 20% of global data consumption. The gap is closing fast. The Indian AI market is projected to grow to over $17 billion by 2027. Data centre capacity in India is expected to rise from under 2 GW in 2025 to 8–15 GW by 2030. Reliance Industries has announced a ₹1.6 lakh crore investment in a 1.5 GW AI cluster in Visakhapatnam. Google has committed $15 billion for a 1 GW hyperscale hub. Microsoft's CEO Satya Nadella announced $17.5 billion in India investments over four years starting 2026. India is being called the "next hyperscale battleground."

Here is the problem. Solar and wind energy are cheap, but they are intermittent — the sun does not shine at night, the wind does not blow on demand. AI data centres cannot tolerate downtime. They need what energy engineers call "base-load power" — electricity that is available 24 hours a day, 365 days a year, rain or shine. Coal can provide this, but it defeats India's climate goals. Gas is expensive and import-dependent.

This is where nuclear energy enters — not as a nice-to-have, but as perhaps the only realistic clean energy option that can provide guaranteed, round-the-clock, carbon-free power at the scale data centres need. Union Minister Ashwini Vaishnaw said explicitly after the SHANTI Act was passed that small and modular nuclear reactors offer "feasible and practical solutions" to meet the huge energy demand of data centres. He noted that new reactor designs can be set up on as little as 14 acres of land — compact enough to sit alongside a data centre campus.

Globally, this marriage between nuclear and AI is already happening. Microsoft signed a deal to restart the Three Mile Island nuclear plant in the US specifically to power its data centres. Amazon, Google, and Meta are actively investing in nuclear energy projects. France, where 70% of electricity comes from nuclear, is marketing its nuclear-powered grid as an AI advantage.

For India, the connection is direct: the SHANTI Act opens the door for private companies (many of whom are also building data centres) to invest in nuclear reactors. SMRs, with their smaller footprint and dedicated output, are particularly suited for powering data centre clusters in cities like Hyderabad, Mumbai, and Visakhapatnam. The Nuclear Energy Mission's target of five SMRs by 2033 is not an abstract clean-energy goal — it is a practical necessity if India wants to remain competitive in the global AI race.

There is a deeper strategic point here. Whoever controls the energy supply for AI infrastructure controls the pace of AI development. If India has to depend on imported fossil fuels to power its data centres, it is building its AI future on someone else's energy. Nuclear energy — especially indigenous nuclear energy — makes AI infrastructure truly self-reliant. This is the indigenisation argument applied to the digital economy: just as India built its own rockets when others refused to share technology, India now needs to build its own energy backbone for AI when the stakes are equally high.

But there are real concerns too. Data centres are enormous water consumers — AI data centres in India could drive water usage to 1,068 billion litres annually by 2028, according to some estimates. Nuclear reactors also need cooling water. Placing both near the same water sources creates competition. Regulatory approvals for nuclear plants near data centre hubs (often in or near cities) will face scrutiny. And the public perception challenge is real — communities may resist having a nuclear reactor next to their neighbourhood, even a small modular one.