The Atomic Revival: Why Tech Giants and Governments Are Betting Big on Nuclear Energy

For most of the last four decades, nuclear power in the developed world was a story of caution rather than growth. Construction slowed after Chernobyl and Fukushima, financing dried up, and utilities generally preferred gas and renewables, which were cheaper to build and faster to bring online. That pattern has reversed with unusual speed over the past two years, and the reason has little to do with climate policy and everything to do with artificial intelligence.

AI training and inference workloads run continuously, around the clock, and at a scale that is straining electricity grids faster than new generation capacity can be added. Solar and wind remain intermittent by nature, and battery storage has not yet closed that gap at data centre scale. Nuclear power, by contrast, offers the one thing hyperscalers need most: firm, always-on, carbon-free baseload electricity. That single requirement has turned some of the world's largest technology companies into direct financiers of nuclear infrastructure, and it has pushed governments, including India's, to rewrite decades-old nuclear legislation.

This piece looks at what is actually happening in this buildout, who is funding it, where small modular reactors fit into the picture, and what the economics and risks look like for anyone following the theme from an investment perspective.

Why AI data centres are turning to nuclear power

The scale of the power problem is the starting point for understanding why nuclear has re-entered the conversation. Global data centre electricity consumption stood at roughly 415 terawatt-hours in 2024 and is projected to more than double to around 945 terawatt-hours by 2030, according to International Energy Agency estimates cited in recent industry analysis. That growth is concentrated in a short window, which matters because grid interconnection has become the binding constraint on AI expansion, arguably more than chip supply. The queue of projects waiting for grid connection in the United States has swelled past 2,600 gigawatts, with average wait times of roughly five years and high withdrawal rates, a backlog driven by insufficient transmission capacity and regulatory bottlenecks.

Renewables alone cannot solve this problem for a data centre that needs to run at close to full utilization every hour of every day. Nuclear plants, once operating, deliver capacity factors well above 90%, making them structurally suited to a workload that does not pause for weather. That is the core logic driving hyperscalers toward nuclear, not a sudden ideological shift toward atomic energy, but a practical response to a baseload gap that renewables and gas alone have not been able to close fast enough.

The scale of Big Tech's nuclear commitments

The commitments made by technology companies over the past two years are large enough to be described, without exaggeration, as a reindustrialization of the nuclear supply chain. Every major AI company, Microsoft, Google, Amazon, and Meta, has signed at least one nuclear power agreement, with combined commitments now exceeding roughly 10 gigawatts of capacity across more than a dozen deals, according to industry tracking as of mid-2026.

Microsoft has taken the most direct route by funding the restart of an existing plant. The company committed roughly 16 billion dollars over a 20-year agreement to bring back online the sister reactor at Three Mile Island, now renamed the Crane Clean Energy Centre, with power expected to begin flowing around 2027 or 2028. Because it involves restarting existing infrastructure rather than building new reactors, this is expected to be the fastest of the major deals to deliver actual electricity.

Amazon has pursued a dual strategy of near-term offtake and long-term reactor development. It expanded an agreement with Talen Energy to purchase 1.9 gigawatts through 2042 from the Susquehanna nuclear plant in Pennsylvania, while separately investing roughly 700 million dollars in X-energy to support development of up to 12 Xe-100 high-temperature gas-cooled reactors, alongside funding for new small modular reactors being developed with Energy Northwest in Washington state.

Google has taken a more concentrated bet on next-generation SMR technology, committing to roughly 500 megawatts of capacity from Kairos Power's molten salt-cooled reactor design, with commercial deployment targeted around 2030.

Meta has emerged as the most aggressive buyer by committed capacity. Its agreements with Vistra, TerraPower, Oklo, and Constellation Energy could deliver up to 6.6 gigawatts of electricity by 2035, including a 20-year power purchase agreement for 1.1 gigawatts from the Clinton Clean Energy Centre in Illinois, announced in mid-2026. Meta's deals span both restarted conventional reactors and newer reactor designs still working through development and licensing.

Two patterns are worth noting across these deals. First, restarts of existing reactors deliver power on a much faster timeline than new construction, typically by 2027 or 2028 versus the early-to-mid 2030s for greenfield SMR projects. Second, technology companies are no longer simply purchasing electricity through standard offtake agreements. In several cases they are directly financing construction, effectively functioning as project developers rather than passive buyers, a shift that has helped make nuclear projects financeable again after decades of stalled economics.

What small modular reactors actually are

Small modular reactors are factory-fabricated nuclear reactors, generally in the 50 to 300 megawatt range per unit, designed to be built largely in a controlled manufacturing environment and then transported to site for assembly, rather than constructed entirely on location the way traditional gigawatt-scale reactors are. The appeal is straightforward: shorter construction timelines of roughly three to five years versus a decade or more for conventional plants, lower upfront capital requirements per unit, and the ability to scale capacity incrementally by adding additional modules as demand grows.

Three reactor technologies dominate current SMR development. High-temperature gas-cooled reactors, the design behind X-energy's Xe-100, use helium as a coolant and can reach higher operating temperatures suited to industrial applications. Molten salt reactors, including Kairos Power's design uses liquid salt as both coolant and, in some variants, fuel carrier, offering passive safety characteristics. Sodium-cooled fast reactors, the approach taken by TerraPower's Natrium design and Oklo's Aurora reactor, use liquid sodium coolant and can potentially use a wider range of fuel types, including some reprocessed material.

Market forecasts for this technology have expanded considerably. Industry research firm IDTechEx projects the global SMR market to reach approximately 53.8 billion dollars by 2036, growing to nearly 300 billion dollars by 2046, driven substantially by data centre demand. Roughly 1.3 billion dollars in equity funding flowed into SMR developers in 2025 alone, alongside expanding support from the U.S. Department of Energy.

The economics remain a genuine open question, however. The current levelized cost of energy for restarted conventional nuclear plants under long-term power purchase agreements runs in the range of 70 to 95 dollars per megawatt-hour. SMR cost projections, inclusive of available tax credits, range from roughly 48 to 85 dollars per megawatt-hour, a range that would make them competitive if development and construction costs are actually controlled at scale. Given that no fleet of commercial SMRs has yet been built and operated at volume anywhere in the world, that cost assumption remains unproven rather than demonstrated.

Governments are rewriting nuclear policy to keep pace

The United States and Canada announced plans in mid-2026 to jointly build ten new nuclear reactors, described as the largest coordinated nuclear push North America has undertaken in decades. This government-level commitment runs alongside, and is partly reinforced by, the private capital hyperscalers are putting into the sector.

India's policy shift has been just as significant, and arguably more structural, since it involves opening a sector that had been legally closed to private ownership for more than six decades. The Atomic Energy Act of 1962 restricted nuclear power generation exclusively to two state-owned entities, the Nuclear Power Corporation of India Limited and Bharatiya Nabhikiya Vidyut Nigam Limited. That changed with the passage of the SHANTI Act, formally the Sustainable Harnessing and Advancement of Nuclear Energy in India bill, which opens the sector to domestic private investment and, to a limited extent, foreign participation, though it stops short of permitting foreign direct investment.

The Union Budget for 2025-26 established a Nuclear Energy Mission with an outlay of roughly 20,000 crore rupees, targeting at least five indigenously designed small modular reactors, referred to as Bharat Small Reactors, to be operational by 2033. India's broader ambition is to expand nuclear capacity from approximately 8.8 gigawatts today to around 22 gigawatts by 2031-32, and ultimately to 100 gigawatts by 2047, a roughly elevenfold increase over roughly two decades. The government has also removed customs duties on nuclear equipment imports through 2035 and increased research and development funding for the sector by 88% year-on-year year in the 2026-27 budget, signalling a shift toward attracting private capital and global technology partnerships rather than relying solely on state-funded construction.

The Indian angle for investors

NPCIL itself is not publicly listed, so market exposure to India's nuclear buildout runs through the engineering and infrastructure sector rather than through a reactor operator directly. Larsen & Toubro and Bharat Heavy Electricals have historically been the two largest suppliers to India's nuclear program, with BHEL having supplied turbine generator sets to 14 of the country's 24 operating reactors and recently winning the turbine island package for a new fleet of ten 700 MW pressurized heavy water reactors. L&T's role spans structural engineering and critical components, including steam generators and pressure vessel forgings, produced partly through its joint venture heavy forge facility in Gujarat. Beyond these two established names, a longer tail of smaller suppliers, including Walchandnagar Industries, MTAR Technologies, and several others, participate in specific parts of the reactor component supply chain, from calandria vessels to cooling system pumps.

It is worth being clear about what this opportunity is and is not. It is not, at this stage, a story about buying shares in a nuclear utility with a clear, near-term earnings ramp. NPCIL's own budget allocation has actually been trimmed even as SMR-focused R&D funding rises, reflecting a government preference for private capital to fund new construction going forward. It is, instead, a longer-duration industrial and engineering supply chain theme tied to multi-decade capacity targets, private-sector participation rules that are still being finalized, and reactor designs, in the case of Bharat Small Reactors, that are still moving through regulatory approval with deadlines that have already been extended multiple times.

Because nuclear and utility projects involve heavy capital expenditure and complex regulatory timelines, conducting deep stock analysis on debt coverage ratios and long-term power purchase agreements is vital before allocating capital. A project's economics depend heavily on the length and pricing of its offtake agreements and on whether the entity building it can service the debt taken on for construction without the multi-year delays and cost overruns that have historically plagued large nuclear projects in India and elsewhere.

Risks and open questions

The nuclear-for-AI thesis carries real execution risk that is worth stating plainly rather than glossing over.

Timelines remain long relative to the pace of AI infrastructure buildout. Reactor restarts are expected around 2027 to 2028 at the earliest, while new SMR projects are generally targeted for the early 2030s, meaning that even the fastest nuclear projects will not close the near-term power gap that is constraining data centre growth today. Goldman Sachs has identified energy availability, rather than chip supply, as the largest infrastructure constraint on AI expansion, and nuclear's long lead times mean it addresses that constraint only partially and only over a multi-year horizon.

Cost overruns and schedule delays have a long history in nuclear construction globally, including in India. The Kakrapar Atomic Power Project units 3 and 4 were budgeted at roughly 11,459 crore rupees but ran eight years behind schedule, with final costs rising to approximately 19,220 crore rupees by the time commercial operation began. There is no strong evidence yet that SMRs, despite their factory-fabrication design intent, will avoid similar cost and schedule risk once built at a commercial scale rather than as demonstration units.

Regulatory and liability frameworks are still being finalized in several markets. India's SHANTI Act was a major structural reform, but implementing rules, the treatment of civil nuclear liability for private operators, and the specific terms under which private companies can build and operate reactors under NPCIL supervision are still being worked out. NPCIL's own request for proposals for Bharat Small Reactors has had its submission deadline extended multiple times, most recently to March 2026, an indication that the private sector response has been more cautious than the policy ambition.

None of this means the theme lacks substance. Government targets, corporate offtake commitments, and capital flowing into reactor developers are all real and verifiable. It does mean that the gap between announced capacity targets and actual operating reactors remains wide and that investors should treat this as a long-duration structural theme rather than a near-term earnings catalyst.

Conclusion

The nuclear revival underway is unusual in that it is being driven less by traditional utility planning and more by a small group of technology companies whose AI ambitions require power at a scale and reliability that current grids cannot fully provide. That has made nuclear projects bankable again after decades of stalled economics, and it has pushed governments, India included, to open sectors that had been closed to private capital for generations.

For investors, the opportunity sits less in predicting which specific reactor technology wins and more in understanding the industrial supply chain, the financing structures, and the multi-decade timelines involved. The capital commitments are real. So are the execution risks. Both deserve equal weight before drawing conclusions about where value will ultimately accrue.

Disclaimer- The content of this document is provided solely for informational and educational purposes and does not constitute financial, investment, legal, or tax advice.