Breaking Jen-Hsun Huang's heartfelt words: Why can AI only embrace nuclear energy? Four key points that 99% of people don't understand.

While the world is reveling over AI's computing power, NVIDIA founder Jensen Huang points to the next battleground: electricity. This is not just an energy choice, but the first needle to puncture the idealism of renewable energy, revealing the endgame of technology giants, national strategy and energy reality behind the development of AI. (Summary: Blue and white joint proposal to amend the upper limit of "nuclear power extension" for 20 years, fight to terminate non-nuclear homes) (Background supplement: Meta signed a 20-year nuclear power agreement to package an entire nuclear reactor to support AI computing power and cooperate with American Constellation Energy) This article is submitted by a person with bachelor's class experience in the Academy of Atomic Sciences of Qingda University, who has been engaged in media work in the field of science and technology and blockchain for nearly ten years You can disapprove of nuclear energy, but you can't think that the AI kingdom and the non-nuclear homeland can have both fish and bear's paw. When NVIDIA founder Jensen Huang suddenly lightning came to the stage this morning (22) and talked about the grand blueprint of how AI can change the world, he was like a magician, pulling out more and more powerful GPUs from his hat. But lately, the focus of his conversation has quietly shifted from chips to a more basic and controversial topic: electricity. He said that the future of AI is closely related to energy, why stigmatize energy...? Nuclear is an excellent energy option... I really hope that the local government will address our needs and solve all the necessary problems so that we can have our Asian headquarters here. This sentence, at first glance, sounds somewhat abrupt and even politically incorrect. In an era of shouting ESG and embracing renewable energy, why would the global AI overlord bet its future on an energy source full of historical baggage and public doubts? What Huang Jenxun sees is not the stock price in front of him or the next quarter's earnings report, but the "endgame" of this AI revolution. The subtext he didn't say is that our imagination of energy over the past two decades, especially the idealistic worship of renewable energy, is about to be ruthlessly punctured by the "electric tiger" of AI. This is not just a technology option, it is a paradigm shift that affects industrial structure, capital flows, and even geopolitics. AI energy hunger: a never-ending feast of computing power To understand why nuclear energy is an inevitable option for AI, we must first understand how terrifying AI's "appetite" is, and how picky its "eating habits" are. The traditional data center operates with peaks and peaks, just like an office building, where people are buzzing during the day and lights are turned off at night. But AI is different, especially the training of large language models, which is a 24/7, never-ending marathon. A piece of NVIDIA H100 GPU has a maximum power consumption of up to 700 watts, a cabinet of up to 8 GPUs, forming an AI server, and a server cabinet is composed of 4 such AI servers, the area and volume can be like a thick bookcase, but just a single hour power consumption is as high as 150 degrees, a day up to 3500 degrees, and such cabinets may have thousands in the server center, just a single day energy consumption of 18 million degrees, and this may be just a single enterprise of a computing cluster. Its operation will produce waste heat, you also have to spend other electricity to take the waste heat away, and cool, and according to the pan-science channel, in fact, the most power-consuming part of the AI server is the cooling part, which costs more energy than the AI server (graphics card cabinet), if you add heat dissipation power consumption, then the AI center operating 5,000 cabinets of GPU2 in a single daylight will consume 40 million degrees. When thousands of such GPUs form a cluster of computing power and operate day and night, its power demand is a stable and huge "flat curve", it has no day and night, and will not reduce power at night, early in the morning, and on weekends. It's like a giant beast that never stops once it starts devouring electricity. According to forecasts, by 2030, the power demand of global data centers will reach nearly 945 TWh per year, which is equivalent to the electricity consumption of the entire country of Japan, and the power consumption of AI will quadruple over the same period. This "constant, high-power" energy demand directly excludes the current mainstream renewable energy sources – solar and wind – from the main option. This is not a disparagement of renewable energy, but the unforgiving reality that the sun does not shine at night and the wind does not remain stable forever. Their essence is "intermittent". For them to be able to supply stable power 24/7, they must be paired with sky-high energy storage facilities, such as large-scale battery arrays. This not only greatly increases the cost, but also causes a huge waste in the energy conversion efficiency, such as energy storage during the day, in the night to the morning can only give AI 30% conversion rate, if you are an AI server, you also have to calculate the conversion rate and feasible deployment machine, which is a huge uncertainty and deployment cost. What AI needs is a "baseload power" (Baseload Power), an energy source that can output steadily throughout the year. Of all the low- or zero-carbon energy options, the only one that is perfectly suited for this role is nuclear energy. A nuclear power plant's Capacity Factor (the ratio of actual power generation to rated maximum power generation) is more than 92 percent in the United States, meaning it operates at peak efficiency almost all year round. This stability is precisely the most demanding requirement of AI, a picky gourmet. Investment in nuclear power plants: mostly assets rather than expenses Huang's call also heralds the next battleground for tech giants: vertical integration from competing for "computing power" to controlling "electricity". In the past, the core competencies of tech giants were algorithms, chips and data. In the future, whoever masters stable, cheap, large-scale low-carbon electricity will master the lifeblood of the AI era. There is a very shrewd business calculation behind this. For a company, electricity is an "expense" (Expense) that directly erodes profits. But a power plant is an "asset" (Asset) that can be included in the balance sheet and even deducted through depreciation. When giants such as Meta, Amazon, and Microsoft start investing directly or signing long-term nuclear power purchase agreements (PPAs), they are not only to reduce operating costs, but also to make a profound strategic layout. The most classic example is the cooperation between Amazon AWS and Talen Energy. AWS signed a ten-year contract to purchase 960 MW of carbon-free electricity from Pennsylvania's Susquehanna nuclear power plant to power its data center. The beauty of the deal is that the data center is built right next to the nuclear power plant, enabling "direct delivery" to maximize energy efficiency and minimize the risk of grid instability. This is not only buying electricity, it is "internalizing" energy as its own core infrastructure, getting rid of the dependence on traditional power companies and the restraint of price fluctuations. This has given rise to a "technology-energy" complex (Tech-Energy Complex) that has never been seen before. In the future, tech giants will no longer just be energy consumers, they will become energy producers and dispatchers. Nuclear energy, especially the small modular reactor (SMR) under development, will be the best piece of the puzzle to achieve this vertical integration due to its flexible site selection, short construction period and potential for higher safety. Imagine that in the future, every large AI data center campus will be equipped with several SMRs next to it, forming a self-sufficient "computing power-power" island. This is the endgame that tech giants are planning. "Nuclear waste...