The rapid development of AI is inseparable from a massive power supply!
Latest news indicates that the Trump administration is pushing regulators to drastically reduce data center budgets. Approval process for grid connection to support artificial intelligence The construction boom of (AI) data centers .
Following the aforementioned news, shares of independent power producers collectively surged during Friday's trading session. By the close, Vistra rose over 5%, Constellation Energy rose over 6%, and NRG Energy rose 4%. Furthermore, nuclear power companies, boosted by the artificial intelligence wave, also saw significant gains, with Oklo rising over 9%, Lightbridge rising nearly 10%, and Centrus Energy rising over 11%.
In the AI field, two other news items have also attracted market attention. First, Google will supply up to 1 million dedicated AI chips to the emerging AI company Anthropic PBC; second, Apple... It was announced that the first batch of "Made in the USA" AI servers shipped from Texas.
Plans to accelerate the connection of AI facilities to electricity
According to Bloomberg, the Trump administration is actively pushing regulators to significantly shorten the approval process for data center companies to connect to the power grid in order to meet the industry's booming electricity demand.
According to documents obtained by the aforementioned media outlets, the U.S. Energy... Secretary Chris Wright recently urged the Federal Energy Regulatory Commission to implement a fast-track approval process for data center grid connection applications. According to Wright's proposed draft rule, such approvals would be completed within 60 days, a significant change from the current approval process, which can take years.
The report states that expedited approvals will help the Trump administration achieve its blueprint for artificial intelligence development. For hyperscale enterprises urgently needing to build high-energy-consuming data centers, the new regulations are a godsend—these companies are facing increasingly serious concerns about their electricity consumption, fearing that their massive energy consumption will drive up electricity bills for residents in surrounding communities.
In his letter to the Federal Energy Regulatory Commission (FERC), Wright emphasized: "To usher in a new era of prosperity for America, we must ensure that all citizens and domestic industries have access to affordable, reliable, and secure electricity. Achieving this goal requires the support of public utilities , including those responsible for artificial intelligence data centers." Large load users supplied by the company must be able to connect to the transmission network system in a timely, orderly, and non-discriminatory manner.
Since the U.S. Federal Energy Regulatory Commission rejected Teren Energy's direct shipment to Amazon Technology and energy executives have been eagerly anticipating rule changes since the application to supply electricity to Pennsylvania's nuclear power plants to a data center. However, new regulations could also face backlash from states, many of which are already struggling to cope with surging electricity demand from data centers, new factories, and electric vehicles, along with rising utility costs.
Under the proposed new regulations, data centers that are equipped with new power generation facilities or commit to cooperating with the regional power grid to reduce electricity consumption during periods of high demand, such as heat waves, can obtain expedited approval. However, data centers planned to be built adjacent to existing power plants (such as the Tyron- Amazon partnership) will require a special assessment to confirm that the power generation capacity is indeed necessary to maintain grid reliability.
Wright stated that the plan aligns with the Trump administration's strategic goals of "revitalizing domestic manufacturing and driving American AI innovation," both of which require unprecedented levels of electricity and necessitate significant investments in the power grid.
Data centers have huge power demands
The rapid development of AI is generating enormous electricity demand. Data from market research firm Collaborative Research Group shows that as of the end of the second quarter of this year, the United States had approximately 522 hyperscale data centers, accounting for about 55% of the global total computing power. It is projected that approximately 280 more such data centers will be operational in the US by the end of 2028.
According to data from the U.S. Department of Energy , before 2020, data centers consumed less than 2% of total electricity in the United States, but by 2028, data centers may account for 12% of total electricity consumption in the U.S. A report released by Deloitte in June of this year stated that by 2035, the electricity demand of U.S. artificial intelligence data centers may increase more than 30 times compared to 2024.
According to a report by the Global Times citing the Wall Street Journal, most US data center developers consider grid connection a top priority. Sridhar, founder of the US energy company Bloom Energy, stated that data centers have long taken for granted the readily available electricity: "Build the data center, then plug it in." Given the enormous power consumption of developing artificial intelligence—the power consumption of one data center is equivalent to that of 1,000 Walmart stores—this is no longer possible. For stores of similar size, the power consumption of a single AI search can be 10 times that of a single Google search.
Deloitte conducted a survey of 120 US power companies in April 2025. A survey of company and data center executives reveals that grid stress is a major challenge facing data center development. This highlights the conflict between the US government and corporate vision for artificial intelligence and the grid's power supply capacity: currently, some data centers face waiting times of up to seven years to connect to the grid. Market research firm Gartner analyst Johnson believes the US power industry is not yet capable of keeping pace with this trend.
Big moves by Apple and Google
In the field of AI, both Apple and Google have made significant moves.
Apple announced on Thursday that its Houston, Texas factory has begun shipping advanced servers for artificial intelligence applications. These servers are a core part of Apple's commitment to invest $600 billion in advanced manufacturing, suppliers, and other areas in the United States.
The rollout of server production capacity coincides with the US government's policy cycle promoting the return of manufacturing, and may echo US President Trump's long-standing policy of urging technology companies to expand domestic production.
It is understood that the servers at the Houston factory will be equipped with Apple's own custom-designed chips for Apple Intelligence and private cloud computing . The service provides computing power support. Apple COO Sabih Khan emphasized in a statement that the project team achieved early production by accelerating factory construction and plans to continue expanding next year to increase output. This production facility is expected to create thousands of jobs, marking a significant shift for Apple's server product line from overseas production to manufacturing in the United States. In February of this year, Apple first disclosed its plans for assembling servers in the United States, and in August, Apple CEO Tim Cook met with Trump and announced additional U.S. spending.
Additionally, on October 23rd local time, Google announced it will provide up to 1 million dedicated AI chips to the AI startup Anthropic. This collaboration will significantly enhance Anthropic's computing power and solidify Google's position in the AI field.
Google says its custom-designed Tensor Processing Unit (TPU, a chip that accelerates machine learning workloads) is planned for deployment in 2026 and will soon deliver more than 1 gigawatt of online computing power.
By using Google's TPU, Anthropic was able to access technologies other than Nvidia's. It is one of the most advanced chip infrastructures available, thereby reducing reliance on scarce and expensive graphics processing units (GPUs).
(Source: Securities Times) (Times.com)
