Subterranean Nuclear Power: The High-Stakes Bet to Fuel the AI Revolution

In the quest to power the insatiable hunger of artificial intelligence, a radical proposal has emerged from the American landscape: bypassing the traditional, sprawling nuclear power plant in favor of compact, modular reactors buried nearly two kilometers underground. As AI data centers proliferate, creating a massive energy bottleneck, the California-based startup Deep Fission is pitching a subterranean solution that promises to redefine the economics of baseload power. With 18.5 gigawatts (GW) of demand currently under inquiry—a figure that suggests a profound shift in how tech giants view their energy supply chains—the plan is as ambitious as it is controversial.

The Genesis of a Subterranean Vision

Founded in 2023 by the physicist and author Richard Muller and his daughter, Elizabeth Muller, Deep Fission seeks to upend the conventional nuclear paradigm. For decades, nuclear power has been synonymous with massive, capital-intensive infrastructure, lengthy construction timelines, and intense regulatory scrutiny.

Deep Fission’s design, by contrast, relies on a compact pressurized water reactor (PWR) that is lowered via reinforced cables into a borehole approximately 1.6 kilometers deep. This design does not require the massive, multi-meter-thick steel and concrete containment domes seen at traditional plants. Instead, the startup intends to leverage the Earth’s natural geology to assist in the cooling and pressure management of the reactor core.

Chronology of Development: From Concept to Groundbreaking

The journey of Deep Fission has been marked by rapid ambition followed by the harsh realities of engineering and regulatory physics.

• 2023: Richard and Elizabeth Muller incorporate Deep Fission, aiming to solve the "energy density" problem for the next generation of computing.
• December 2025: A milestone is reached in Parsons, Kansas, where the company conducts a ceremonial groundbreaking for an exploratory borehole. While it currently stands as little more than a construction site, it represents the first physical step toward proving the feasibility of their deep-well concept.
• June 2026: In the wake of intense international pressure from UN Secretary-General António Guterres, who called for AI giants to decarbonize their data centers by 2030, Deep Fission reveals a massive surge in interest. The company announces 18.5 GW in potential contracts, primarily from data center operators seeking reliable, "behind-the-meter" power.
• Present Day: The company is currently navigating the transition from conceptual design to the drilling of a 76-cm diameter shaft, significantly larger than the initial 20-cm exploratory hole, to house a prototype reactor.

Supporting Data: The AI-Energy Nexus

The urgency behind Deep Fission’s pitch is rooted in a global energy crisis driven by silicon. According to a new study from the United Nations University (UNU), data centers consumed more electricity in 2025 than all but ten countries on Earth. By 2030, this share is expected to double, accounting for nearly 3% of total global electricity consumption.

The International Energy Agency (IEA) warns that renewable energy sources—while critical—will likely only cover half of the projected increase in demand by 2030. The remaining gap must be filled by gas, coal, or nuclear power. For data center operators, the challenge is not just generation, but connection. In many regions, the wait time to connect to the public grid spans years.

Deep Fission’s value proposition is centered on "behind-the-meter" power. By building a 15 MW reactor directly adjacent to a data center, the operator bypasses grid congestion entirely. With a target of 5 to 7 US cents per kilowatt-hour, the company claims its electricity would be both cost-competitive and significantly more stable than the intermittent supply from wind or solar.

Engineering the Deep-Earth Reactor

The technical heart of the system is a standard pressurized water reactor, a technology with a proven safety record in naval and civilian applications. However, the operational environment is entirely new.

In a surface-level PWR, the high pressure required to prevent the coolant from boiling at 300°C is generated by an expensive, high-tensile steel pressure vessel. Deep Fission’s design aims to use the hydrostatic pressure of the borehole depth to aid in the thermal-hydraulic stability of the reactor. The steam generated by the core is then piped to the surface to drive a turbine, essentially using the Earth’s crust as a natural, massive, and free containment structure.

The scalability of the concept is a key selling point. While one reactor delivers a modest 15 MW, a cluster of 100 boreholes—occupying less than three hectares of land—could theoretically produce 1.5 GW of power, matching the output of a large-scale traditional nuclear plant.

Official Responses and Industry Skepticism

While the industry buzz is palpable, institutional caution remains high. Deep Fission’s initial target date of July 4, 2026, for the first three reactors to reach "criticality" has already proven overly optimistic. The company has since removed hard deadlines, reflecting the immense difficulty of obtaining regulatory approval from the Nuclear Regulatory Commission (NRC) in the United States.

Critics point to several hurdles:

1. Regulatory Hurdles: The NRC has never certified a reactor that is not permanently attached to a surface structure. The legal framework for a "subsurface" modular reactor is virtually non-existent.
2. Operational Maintenance: If a reactor malfunctions at a depth of nearly 2 kilometers, the maintenance protocol involves a delicate, high-stakes extraction process that has never been tested at scale.
3. Geological Uncertainty: The integrity of a borehole in different seismic or soil conditions remains a point of intense scrutiny by geological experts.

Implications for the Global Energy Market

If successful, Deep Fission could represent the "holy grail" for AI-heavy regions. It offers a path to provide 24/7, carbon-free, and decentralized baseload power. However, the company is competing against a crowded field of alternative energy strategies.

Conclusion: A High-Risk, High-Reward Frontier

The allure of Deep Fission is undeniably magnetic. In a world where the speed of AI innovation is hampered by the speed of grid infrastructure, a modular, "drill-and-deploy" energy solution is the dream of every Silicon Valley executive.

However, the chasm between a compelling PowerPoint presentation and a functioning, safe, and regulated nuclear reactor is vast. While the 18.5 GW of interest indicates that the market is desperate for new solutions, it is also a testament to the fact that the tech sector is willing to bet on high-risk, unproven technologies to secure their future. For now, the "atom in a hole" remains a brilliant concept awaiting the ultimate test: the cold reality of physics and the rigorous scrutiny of nuclear oversight. Whether Deep Fission becomes the savior of the AI boom or a footnote in the history of experimental energy remains an open, and high-stakes, question.