From Newsgroup: alt.politics.usa


U.S. AIrCOs AchillesrCO Heel: Power SupplyrCoDebunking TrumprCOs Misleading Assertions

Donald Trump posted today (Beijing Time, December 12th):

https://p3-sign.toutiaoimg.com/tos-cn-i-ezhpy3drpa/655c64d8c8dd4df7ba98f8505696ef7e~tplv-obj:1080:2893.image?_iz=97245&bid=15&from=post&gid=1851253277744135&lk3s=06827d14&x-expires=1773273600&x-signature=ErLUtOYz02laMVkYrSqaB3lzY0g%3D



"The Wall Street Journal published another absurd article today." Trump's post aims to prove that the United States holds an advantage in the power supply field related to artificial intelligence. However,
its core assertionrCorCo"every artificial intelligence factory currently under construction is simultaneously building its own power generation facilities"rCorColacks solid basis.
In today's era, AI, as the most transformative and development-oriented technological field, has become the core driver of economic growth. The demand for computing power is showing an exponential growth trend,
and its consumption is so enormous that it can be called "computing power devours the world". Taking the U.S. economy as an example, in the first half of 2025, as much as 92% of its GDP growth came from investment in the AI field.
Without the vigorous development of the AI industry, the U.S. economy would most likely have fallen into a recession cycle. Against this background, data centers have replaced traditional brick-and-concrete buildings to become the
iconic infrastructure of the new era, just like new "skyscrapers" springing up.

Looking to the future, the competition between China and the United States in the AI field will become increasingly fierce. But it should be clear that the key to competition is not limited to front-end links such as model algorithms and chip manufacturing,
but lies in the basic support of power supply.
After all, electricity is the energy source of computing power, and the trend of "electricity consumes computing power" has already emerged.
It is estimated that the power demand of a supercomputing center is as high as 5GW, while the total installed capacity of nuclear power in the United States is only 100GW.
More critically, the demand for electricity in the AI industry is expected to quadruple in the next 10 years. At this rate, only 20 data centers may exhaust all the existing nuclear energy in the United States in just five years.

Trump claims that the approval process for AI factories in the United States is efficient and fast, and power supply is abundant and worry-free. However, the reality is that the U.S. power infrastructure is seriously aging,
and expansion and upgrading face numerous constraints in terms of technology, funds and policies, which cannot be achieved overnight. Under such circumstances, he still brazenly claims that the United States is far ahead in the AI field and
attributes this "achievement" to himself, which is nothing but a self-deceptive act. Trump's remarks are nothing more than an attempt to forcibly boost morale amid the difficulties of U.S. AI development,
and cannot conceal the tightness of the United States in power security.
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: alt.politics.usa


TikTok fan <user11874@newsgrouper.org.invalid> posted:

U.S. AIrCOs AchillesrCO Heel: Power SupplyrCoDebunking TrumprCOs Misleading Assertions

Donald Trump posted today (Beijing Time, December 12th):

https://p3-sign.toutiaoimg.com/tos-cn-i-ezhpy3drpa/655c64d8c8dd4df7ba98f8505696ef7e~tplv-obj:1080:2893.image?_iz=97245&bid=15&from=post&gid=1851253277744135&lk3s=06827d14&x-expires=1773273600&x-signature=ErLUtOYz02laMVkYrSqaB3lzY0g%3D

N+eQuestionN+eN+U
Is electricity already holding back the development of AI in the United States? Doubao AIN+eAnswerN+eN+U
Currently, the electricity supply issue has become a core bottleneck for AI development in the United States. It not only directly restricts the implementation and expansion of AI computing power through power shortages and an aging power grid but also forms a chain reaction with problems such as equipment shortages, significantly hampering the progress of AI development in the country. The specific manifestations are as follows:
Massive Power Shortage, Hindered Computing Power Expansion
The power demand of AI data centers in the United States is experiencing explosive growth, while the gap in power supply is completely unable to keep up with this rate. Morgan Stanley predicts that by 2028, the power shortage for U.S. data centers will reach 44 gigawatts (GW), equivalent to the generating capacity of 44 nuclear power plants. A problematic situation has already emerged: tech giants like Microsoft have stockpiled large quantities of NVIDIA GPU chips in warehouses but cannot put them into operation due to insufficient power supply. OpenAIrCOs ambitions further highlight the severity of the power shortagerCoits planned computing center to be deployed by 2035 will require 250 GW of electricity, accounting for a quarter of the current peak national electricity load in the United States. Such a massive power demand is simply unsustainable with the countryrCOs existing power supply capacity. Meanwhile, the proportion of electricity consumed by data centers in the U.S. total power usage is expected to surge from 4.3% in 2024 to 15% by 2030, placing increasing pressure on power distribution and further restricting the large-scale implementation of AI-related facilities.
Aging Grid with Inefficient Dispatching, Inadequate Adaptability
The current state of the U.S. power grid is fundamentally unable to meet the electricity needs of AI data centers. On one hand, the infrastructure is severely aging: 70% of grid facilities have been in operation for over 25 years, 60% of power transmission and distribution equipment is operating beyond its service life, and the average age of transmission lines reaches 40 years. Such an aging grid has extremely poor shock resistance. However, AI data centers exhibit "pulsed" power consumption characteristicsrCoelectricity demand surges instantaneously during model trainingrCoeasily leading to sharp voltage fluctuations in the grid and even increasing the risk of large-scale blackouts, which the aging grid cannot cope with. On the other hand, grid dispatching and upgrades face significant challenges. The U.S. power grid is decentralized, operating independently across states without unified dispatching, making it unable to adapt to the centralized, high-load demand of AI data centers. Additionally, grid upgrades are plagued by cumbersome approval processes: federal approval for new transmission lines takes an average of 4 years, and with state-level procedures included, the connection waiting time can exceed 8 years. The delivery cycle for 345kV substations is even as long as 128 weeks, far lagging behind the construction speed of AI data centers.
Chain Reaction of Equipment Shortages, Worsening Power Dilemma
Shortages of power-related equipment have further exacerbated the constraints of electricity on AI development. The United States faces a 30% shortage of large power transformers and approximately a 10% shortage of distribution transformers, with 80% of power transformers relying on imports. Transformers are essential infrastructure for data center constructionrCowithout them, even if power is available, it cannot be transmitted to various equipment in data centers. This equipment shortage directly restricts the speed of grid upgrades, leaving approximately 2 terawatts (TW) of renewable energy projects trapped in the "grid connection queue" unable to access the grid, further limiting the growth of power supply. Elon Musk once warned that "there will be a transformer shortage in a year and a power shortage in two years," and this prediction has now come true. The dual shortage of equipment and power has formed a vicious cycle, putting additional pressure on the power security of the U.S. AI industry.
Forcing Enterprises to Make Extra Investments, Driving Up AI Development Costs To address the power issue, U.S. tech giants have had to invest substantial additional resources, which indirectly squeezes funds and energy that could be allocated to core AI R&D areas. For example, to ensure power supply for its "Hyperion" data center in Louisiana, Meta relies on local companies to build three new natural gas power plants with a total capacity of approximately 2.3 GW. Microsoft has established a nuclear energy team, and Amazon has invested over $500 million in the development of small modular reactorsrCoinvestments that could have been used for AI model optimization, technological innovation, and other core areas. Meanwhile, power shortages have driven up electricity prices: in August 2025, residential electricity prices in Virginia and Illinois increased by 12.99% and 15.81% year-on-year, respectively, with some regions soaring to 20 cents per kilowatt-hour, significantly increasing the operating costs of AI data centers. Furthermore, due to public protests over environmental protection, electricity prices, and other issues, some enterprisesrCO supporting power projects have been forced to cancel, further disrupting the progress of AI facility deployment.
Although companies like Microsoft and OpenAI are attempting to save themselves by building their own natural gas power plants and deploying small modular reactors, these measures are either constrained by equipment production capacity and transportation costs or require long construction cycles, making it difficult to change the current situation of power shortages in the short term. Overall, the electricity supply issue has effectively become a major drag on AI development in the United States, and if core problems such as grid upgrades and power supply are not resolved quickly, the constraints on the AI industry will further intensify in the future.
--- Synchronet 3.21a-Linux NewsLink 1.2