Powering Intelligence: Why India’s AI Ambitions Now Depend on Nuclear Reform
Introduction: India’s AI Moment and the Missing Link
India’s artificial intelligence inflection point has arrived faster than expected. In the span of just a few days, global technology leaders committed more than $67.5 billion to India’s AI and data-center ecosystem 1. Microsoft announced a $17.5 billion expansion, Amazon pledged $35 billion on top of earlier investments, and Google unveiled a $15 billion AI hub in Visakhapatnam.
These are not incremental bets. They signal a deeper shift in how global firms view India—not merely as a market or talent pool, but as a core pillar of future AI infrastructure.
Yet capital alone does not build AI capacity. Power does.
This is where the story takes a decisive turn. On December 12, 2025, India’s Union Cabinet approved the SHANTI Bill—short for Sustainable Harnessing and Advancement of Nuclear Energy for Transforming India. While positioned as a long-overdue reform of the atomic energy sector, the Bill may ultimately prove far more consequential. It directly addresses the single biggest constraint facing India’s AI ambitions: the availability of reliable, scalable, carbonfree power.
The Scale of India’s Data-Center Expansion
India’s data-center footprint has expanded at unprecedented speed. Capacity has grown from approximately 350 MW in 2019 to about 1,030 MW in 2024, and projections suggest it could reach nearly 9,000 MW by 2030 2—a nine fold increase from today.
What makes these figures striking is not just their size, but their speed. Hyperscale projects announced in the past six months in a single Tier‑2 Indian city already exceed the country’s total data-center capacity from just a few years ago. Investment flows mirror this trajectory, with nearly $60 billion committed between 2019 and 2024 and cumulative investments expected to cross $100 billion by 2027.
What is unfolding is not a gradual build-out but a compressed infrastructure cycle driven by AI demand.
Why Power Becomes the Binding Constraint
The expansion of data centers brings with it a fundamental constraint: electricity.
By 2030, data centers are expected to account for nearly 3 percent 3 of India’s total electricity consumption, up from a negligible share today. AI workloads amplify this challenge. Traditional server racks draw 3–5 kW per cabinet, while GPU-dense AI clusters routinely consume 30–50 kW or more and often run continuously for weeks.
For AI infrastructure, power must meet four conditions simultaneously:
A single hyperscale AI data center can require hundreds of megawatts of uninterrupted power—roughly equivalent to the electricity demand of a midsized city. The challenge is not merely generating enough electricity. It is delivering the right kind of electricity: power that is continuous, predictable, affordable, and increasingly carbonfree.
Energy reliability matters because AI systems do not tolerate interruption. Cost matters because power can account for 30 to 40 percent of a data center’s operating expenditure. Sustainability matters because every major hyperscaler has committed to carbonneutral or netzero operations. And availability matters because AI training cannot pause when the sun sets or the wind drops.
This is where India’s otherwise impressive renewable energy story encounters a structural limitation. Solar and wind have transformed the country’s power mix and will remain indispensable. But they are inherently intermittent. They excel at serving variable demand and daytime peaks. They are far less suited to powering highdensity compute infrastructure that must run, without interruption, twentyfour hours a day.
What the SHANTI Bill Changes
The SHANTI Bill represents the most significant reform of India’s nuclear sector since its inception. It ends decades of effective state monopoly and opens the door—carefully, but meaningfully—to private participation. In doing so, it unlocks what could be as much as ₹19.28 trillion, or roughly $214 billion, in cumulative nuclear investment as India works toward its stated target of 100 GW of nuclear capacity by 2047 4.
Today, India operates just 25 nuclear reactors with a combined capacity of about 8.9 GW, all run by the Nuclear Power Corporation of India Limited. The roadmap ahead is ambitious: 22 GW by 2032 and 100GW, a more than elevenfold expansion, by 20475. Achieving this scale simply is not possible without private capital, modern regulation, and a rethinking of how nuclear projects are developed and financed.
The SHANTI Bill enables several critical shifts. It allows private companies to participate across the nuclear value chain, including atomic mineral exploration, fuel fabrication, equipment manufacturing, and potentially certain operational roles. It supports the development of Small Modular Reactors (SMR), backed by a ₹20,000crore Nuclear Energy Mission aimed at deploying five indigenously developed SMRs by 2033. It reforms the civil liability framework—long viewed as the single biggest deterrent to investment—by clarifying operator and supplier responsibilities, introducing insurancebacked caps, and providing government guarantees. And it consolidates outdated legislation into a unified legal and regulatory structure capable of supporting expansion at scale.
For AI infrastructure, these changes are not abstract policy improvements. They are directly relevant to how data centers are powered.
Why Nuclear—Especially SMRs—Align with AI Infrastructure
Nuclear power, particularly in the form of SMRs, offers characteristics that align closely with AI datacenter requirements. Nuclear plants provide baseload power with capacity factors exceeding 95 percent, delivering consistent electricity regardless of weather or time of day. Their generation profile matches the steady, predictable load of data centers far more closely than intermittent sources.
SMRs also change the economics and timelines of nuclear deployment. Unlike traditional large reactors, which are built on site over five to ten years, SMR components are manufactured in controlled factory environments and assembled modularly. This can reduce construction timelines to 24–36 months, improve quality control, and lower financing risk. Their physical footprint is dramatically smaller, requiring far less land than wind or solar installations for equivalent output, and newer designs significantly reduce water consumption.
Perhaps most importantly, SMRs can be deployed close to—or even on—large datacenter campuses, enabling dedicated power supply. Modular configurations allow capacity to scale in line with demand, rather than forcing operators to overbuild years in advance.
Global Signals: Hyperscalers Are Already Moving
Global technology firms have already begun securing nuclear power to support AI expansion.
Amazon has invested in SMR developers and signed agreements linked to more than 5 GW of future nuclear capacity. Google has secured long-term nuclear supply from emerging reactor companies. Microsoft is exploring nuclear partnerships for Azure AI, while Meta has entered long-duration nuclear power purchase agreements.
These decisions reflect hard arithmetic. Global data-center electricity demand is growing at roughly 15 percent annually, and AI data centers alone are projected to consume nearly 945 TWh per year by 2030—roughly equivalent to Japan’s total electricity consumption6.
Renewables Remain Foundational—but Not Sufficient Alone
India’s renewable energy progress remains critical. As of March 2025, the country had installed approximately 220 GW of renewable capacity and is targeting 500 GW by 2030. More than 50 percent of installed capacity already comes from non-fossil sources7.
Renewables will continue to play a central role in serving variable demand and reducing overall system costs. However, for constant, high-density baseload power required by AI infrastructure, nuclear provides capabilities renewables cannot deliver on their own.
Why Timing Matters Now
Several factors make this moment uniquely consequential. SMR technologies are approaching commercial maturity. Liability reform is unlocking private capital. India possesses end-to-end nuclear manufacturing capability. Data-center demand curves align closely with projected SMR deployment timelines. Statelevel policies in Andhra Pradesh, Tamil Nadu, Maharashtra, and Karnataka already support hyperscale infrastructure through singlewindow clearances and fiscal incentives.
This convergence creates a narrow but powerful window for execution.
The Strategic Implication for India
India now sits at the intersection of three structural shifts: the global AI boom, the energy transition, and a renewed push for industrial leadership. The SHANTI Bill is not merely an energy reform. It is infrastructure policy for the AI era.
If executed well, the SHANTI Bill provides India with a durable competitive advantage—abundant, reliable, carbon-free power at scale. This advantage extends beyond data centers, anchoring a broader AI development ecosystem.
The risk is equally clear. Data centers are mobile. Capital can move. Power constraints cannot.
Conclusion: From Policy Intent to Execution
The SHANTI Bill reframes nuclear energy from a legacy sector into a future-critical enabler of India’s digital economy. Combined with India’s renewable leadership, it creates a credible pathway to power AI-driven growth at national scale.
Policy intent is now clear. The opportunity is historic. What remains is disciplined execution—at speed, and without hesitation.
Sources:
[1] https://erp.today/why-hyperscalers-are-investing-67-5-billion-in-india/
[2] https://www.nextmsc.com/report/india-data-center-power-solution-market-ic3679
[3] https://business.cornell.edu/article/2025/05/sustainability-challenge/
[4] https://economictimes.indiatimes.com/
[5] https://csiglobal.co/towards-100-gw-the-future-of-nuclear-energy-in-india/
[6] https://www.weforum.org/stories/2025/04/data-centers-hydrogen-technology-news-april-2025/
[7] https://www.pib.gov.in/PressReleasePage.aspx?PRID=2209478®=3&lang=2
