whoever controls the fuel will control the AI

In the deep mines of Kazakhstan and the data centers of Northern Virginia, two worlds that should never have touched are colliding. The digital speed of Artificial Intelligence faces the heavy inertia of nuclear physics. We have discovered, the hard way, that AI does not live in “the cloud” but on the ground. It has a ravenous hunger of a material that the world ignored for decades: uranium.

The end of the myth of efficiency. For years, the official Silicon Valley narrative was that chip efficiency would offset energy consumption. However, cHow an OilPrice analysis explainsthis idea has died because of the “Jevons paradox“Basically, the more efficient we make a chip, the more units we deploy and the more complex the models become. AI not only consumes data, but incinerates energy to create them.

This reality has forced a paradigm shift. According to a global survey to more than 600 investors, 63% already consider that AI electricity demand is a structural change in nuclear planning. It is not a temporary peak, it is the foundation on which the economy of the 21st century will be built.

The gap between the code and the steel. The fundamental problem is that software is moving at the speed of light, while the uranium supply remains “stuck in the mud” of 20th century industrial timelines. This temporary disconnection reveals an uncomfortable reality: the world has run out of room to maneuver. For two decades, humanity survived thanks to secondary supplies —reusing old Cold War warheads and surplus inventories—, but these strategic warehouses are practically exhausted today.

This shortage is a deep structural deficit. Uranium.io data reflect an alarming gap where the uranium coming out of the mines will cover less than 75% of what the reactors will need in the short term. This is what Sprott Asset Management define as a market that lives at “two speeds”: a superficial volatility that hides a deficit that widens like a canyon.

“AND“The silence of the electric companies”. On the Sprott Radio podcastexpert John Ciampaglia explains that, although 2025 seemed like a stagnant year for the price of physical uranium—anchored between $77 and $80—mining stocks rose 40%. This disconnection reveals that, while investors are already betting heavily on what is to come, electricity companies (utilities) are at a “stalemate”. They are delaying signing new contracts and burning down their last reserves in the hope that prices don’t skyrocket, but the pressure from AI is such that sooner or later someone will have to blink first.

Uranium as a strategic asset. If the semiconductors were the battlefield of the last decade, nuclear fuel is that of the next. Whoever controls the uranium will control the computing capacity. On the one hand, how the analyst describes for Oilpricewhen a tech giant signs a 20-year power agreement (PPA) with a nuclear plant, it is “locking up” the best clean electrons for private profit. The risk is the socialization of the cost, the companies take the clean energy, but the citizen pays to update the electrical network.

On the other hand, “Atoms for Algorithms”. The Director General of the IAEA describe this union as a “structural alliance”. AI doesn’t just need nuclear; The nuclear industry needs AI for the predictive maintenance of reactors, the design of new materials and the improvement of safety.

The strategy of the giants. The hyperscalers (Google, Microsoft, Amazon) have understood that to dominate AI they must secure the atoms before the competition.

• Vertical Integration: Google took a turn of the rudder by acquiring Intersect Power for $4.75 billion. The objective is to control the availability and cost of supply near your data centers, without depending on the public network.
• Modular Reactors (SMR): The International Atomic Energy Agency bet on SMRsmall reactors that allow a technology company to add nuclear power as it adds servers. It is literally bringing scalability from software to power.
• Sovereign AI: Companies like VivoPower they are redirecting capital towards markets such as Saudi Arabia or the United Arab Emirates. There, where the electrical grid is a bottleneck, the solution is to create computing infrastructures with its own energy generated “behind the meter.”

China: the provisional winner. While the West debates, China pours concrete. The Asian giant build reactors at a rate that no one else reaches, between ten and eleven per year. In fact, half of all the reactors being built in the world are in Chinese territory. According to the CNEAthe country will surpass France in nuclear capacity in 2026 and the United States in 2030.

Beijing not only seeks firm energy to sustain its renewables, but also total technological independence. It already produces 100% of its nuclear equipment and leads the fourth generation with high-temperature modular reactors. They are even “fishing” uranium from the sea with new absorption technologies to ensure centuries of autonomy. China has understood that nuclear energy is both a tool of decarbonization and energy diplomacy.

The wall of reality. In the software world, problems are solved by injecting capital or code; In the world of atoms, money cannot buy time. There are three physical obstacles that Silicon Valley capital will not be able to solve immediately:

1. The bottleneck of enrichment. There is no point in extracting the mineral if you cannot convert it into fuel, and that industrial capacity in the West is at its limit. As they warn in the podcastmuch of this vital process remains tied to Russian state interests, making AI power a national security issue.
2. The talent crisis. For an entire generation, the global message was that nuclear power was a dead technology. The result it’s a shortage criticism from engineers and specialists; There are simply no qualified “hands” to operate the new mines or manage the reactors. We have lost the know-how industrial while we were distracted with the digital world.
3. The “asking price.” Although uranium aims for the range of $100-120/lb by 2026, the figure of $135 is the one that it really marks desperation of the sector. That price is not a sign of a healthy market, but rather the minimum incentive to “beg” miners to reopen abandoned mines and face the ten-year bureaucratic hell required to obtain a single environmental permit.

From click to kilowatt. We are returning from a world of “clicks” to a world of “kilowatts.” The utopia of AI today depends on a metal that the world forgot. If the technology and nuclear industries do not synchronize their clocks, AI will hit an insurmountable physical wall. In the end, the nation that secures the uranium will be the one that leads the intelligence revolution. In the 21st century, computing power is, first and foremost, electrical power.

Image | IAEA Imagebank and freepik

Xataka | Something is going wrong with AI. The US is turning to energy solutions that it thought were buried to power data centers