---
title: "American AI cannot cross the river while touching China"
type: "News"
locale: "en"
url: "https://longbridge.com/en/news/282034159.md"
description: "The giants of Silicon Valley in the United States face a dual challenge of computing power and energy infrastructure in the face of AI development. Musk's purchase of power plants, Meta's signing of nuclear energy contracts, and Google's acquisition of nuclear power stations demonstrate an urgent demand for stable, low-latency power supply. Electric resources are transforming into strategic assets, and competition among countries in computing power, green energy, and data sovereignty is intensifying. Infrastructure and capital investment have become key to unlocking the economic potential of AI, and in the future, the position of countries in the global AI economy will be influenced by their infrastructure construction capabilities"
datetime: "2026-04-08T11:55:37.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/282034159.md)
  - [en](https://longbridge.com/en/news/282034159.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/282034159.md)
---

# American AI cannot cross the river while touching China

In the past year, Silicon Valley giants have been scrambling, and what troubles figures like Zuckerberg and Musk is not only computing power but also the underlying energy infrastructure.

To address this issue, Musk purchased an entire power plant overseas and brought it back to the United States, frequently arranging teams to conduct research and procure photovoltaic equipment in China. Zuckerberg's Meta has signed at least three major nuclear energy contracts, and Google has spent $4.8 billion to acquire a nuclear power plant.

It can be said that in the United States, it takes 7 years to build a power grid, but the Silicon Valley giants cannot wait even a day.

The explosive demand for training and inference of large models has caused data centers to require **stable, low-latency, and sustainable power supply** far beyond traditional internet infrastructure, further forcing national power grids to undergo deep transformations in transmission and distribution capacity, energy storage technology, renewable energy absorption, and "power-computing" collaborative management.

At the same time, **electricity resources themselves are beginning to transform into new strategic assets**, with competition between nations and regions intensifying around "computing power availability," "proportion of green energy," and "data sovereignty," elevating data centers from mere technical facilities to key nodes that influence the global power structure.

## **The Political Economy of Infrastructure in the Age of Artificial Intelligence**

Just as railways once reshaped logistics speed and territorial spatial structure, and the internet changed information flow and business organization, the production mode centered on "computability" in artificial intelligence is also reconstructing the logic of value creation, giving rise to new industrial divisions, consumption patterns, and governance systems.

In this process, infrastructure and capital investment have become fundamental prerequisites for unleashing the economic potential of artificial intelligence.

**In other words, competition is not only reflected at the algorithmic level but also in who can build the corresponding infrastructure network faster, on a larger scale, and in a greener manner.** Therefore, capital flows are shifting from software to a new type of infrastructure encompassing "computing power - energy - network," which is an important signal of the current changes in the global economic landscape. The differences in infrastructure and investment capabilities will determine the future positions and influence of various countries in the global artificial intelligence economy.

Looking back over the past decade, the power demand of global data centers has indeed experienced significant growth, but this growth has not always been "explosive"; rather, it has gone through a **process of slow to accelerated growth**.

According to the International Energy Agency (IEA) analysis for 2024, from 2010 to 2018, **the energy usage of global data centers increased by about 6%**, with an average annual growth rate of about 0.7%. However, **since 2018, it has grown by about 50%–80%, with an average annual growth rate equivalent to 8%-13%**.

If this trend continues, it is expected that by 2030, global data center energy consumption will reach 600–800 terawatt-hours (TWh), and the IEA report for 2025 has been updated to 935 TWh (equivalent to a data center scale of 108GW capacity), accounting for 1.8%–2.4% of the projected global power demand for that year. If artificial intelligence drives a higher consumption rate (for example, the demand for large model training could lead to energy consumption growing at a rate of 20% per year), then by 2030, data center energy consumption could reach 1100–1400 TWh, accounting for about 3%–4% of the projected global power demand In China, it is expected that by 2030, the electricity demand of data centers will double compared to 2020, reaching 400 terawatt-hours.

The demand has transitioned from slow to accelerated growth, with different backgrounds in each phase.

Before 2018, although the volume of data services, network traffic, and storage demand significantly increased, the overall energy consumption of data centers during this decade did not follow the business volume with a "explosion" due to improvements in server hardware efficiency, advancements in cooling technology, and the trend of hyperscaler data centers replacing traditional inefficient small data centers.

However, after 2018, the energy usage of global data centers has significantly increased, with growth rates jumping into double digits. This shift is primarily driven by the demand for AI computing power, the expansion of hyperscale data centers, and the surge in traffic from video content platforms. Data centers have become one of the fastest-growing types of infrastructure in terms of global electricity consumption, bringing new pressures to energy systems, carbon emissions, and digital governance.

Especially after the emergence of large models, many regions have entered a rapid expansion phase for data center construction. There is no completely unified and widely accepted "official" number for the total number of data centers globally, as different countries have varying definitions, scale standards, and registration methods for data centers, making it possible to estimate their "total number" only as an approximation.

According to a summary from the market statistics agency Market.biz, **as of March 2024, there are approximately 11,800 data centers operating globally.**

In terms of regional distribution, Statista data shows that by November 2025, the United States will have the most data centers in the world, with 4,165 facilities, followed by the United Kingdom (499), Germany (487), China (381), France (321), Canada (293), Australia (274), India (271), Japan (242), and Italy (209).

It must be acknowledged that the current and future energy demand of data centers is unevenly distributed globally. For example, in the United States, data centers account for more than one-fifth of the total electricity consumption in Virginia. In Europe, the electricity demand of data centers in Ireland was 5.3 terawatt-hours in 2022, equivalent to 17% of the country's total electricity consumption. By 2026, as artificial intelligence applications rapidly penetrate the market, this electricity consumption will nearly double, reaching 32% of the country's total electricity demand.

The highly concentrated nature of data centers and their extremely high power density pose significant challenges at the local level, including issues related to grid access and capacity limitations, water resource consumption, and community opposition.

Another obvious trend is that **the electricity consumption of hyperscale data centers operated mainly by large technology companies has significantly increased in recent years**. From 2017 to 2021, the total electricity consumption of just four companies—Amazon, Microsoft, Google, and Meta—more than doubled, reaching approximately 72 terawatt-hours.

The explosive growth in the number of hyperscale data centers operated by technology companies has brought tremendous challenges to supply In **many countries, the power system is highly fragmented**—operated independently by multiple regional or local power companies, lacking unified scheduling and capacity planning—leading to issues such as voltage fluctuations, power shortages, or scheduling delays. Additionally, the significant differences in electricity prices, policies, and levels of power investment across different regions further complicate the construction and operation of data centers. Overall, the fragmentation of the power system not only restricts the scalability of data centers but also affects the reliability and energy efficiency of digital infrastructure to some extent.

A more fundamental challenge lies in the sources of energy supply itself.

**Many countries' data centers still rely on fossil fuels such as coal and natural gas**, which not only bring carbon emission pressures but are also susceptible to fluctuations in fuel supply and price changes; while renewable energy, despite its rapid growth, faces issues of uneven distribution and intermittency, lacking sufficient storage and intelligent scheduling means, making it difficult to continuously meet the "7×24 hours" power supply demands of data centers. In this context, nuclear energy is seen as a long-term viable solution. However, the long construction cycle, huge upfront investment, and the need for strict safety regulation and policy support pose challenges in its practical promotion, including technological maturity, social acceptance, and waste disposal capabilities.

In summary, the energy issues of data centers are not only technical problems of grid structure but also a long-term test of energy strategy and policy layout.

## **The American Model: Energy Constraints Driven by the Market**

The development of data centers in the United States has long relied on market mechanisms and private capital. This model was extremely efficient in the early days of the internet: companies could deploy large-scale data centers in places like Oregon, Virginia, and Texas based on electricity price differences and tax incentives.

According to a report by JLARC (The Joint Legislative Audit and Review Commission), Virginia's data center capacity accounts for about 25% of North America's total capacity and 13% of the global total. **Northern Virginia has more data centers than any other region, earning the title "World Data Center Capital."**

The JLARC report indicates that the data center capacity in Northern Virginia is more than twice that of the next largest competitor—Beijing, China—and three times that of Hillsboro, Oregon, the next largest data center hub in the U.S. The state's tax incentives have made Hillsboro a popular location for data centers, serving companies like Meta, LinkedIn, TikTok, and X. However, with the advent of the AI era, this market-driven expansion path is gradually encountering hard constraints at the infrastructure and institutional levels.

While the U.S. leads China in many aspects of artificial intelligence, especially in software and chip design, **the U.S. faces significant bottlenecks in the power supply and infrastructure approval for data centers**. The computing power of artificial intelligence is like an "electric tiger," wildly consuming the U.S. electricity resources, exacerbating an already fragile power grid Most of the facilities in the U.S. power grid were built in the 1960s and 1970s. Although the system has been upgraded through automation and some emerging technologies, the aging infrastructure is increasingly struggling to meet modern electricity demands.

According to an assessment by the American Society of Civil Engineers, the overall health of the U.S. power grid received a rating of only C+, with 70% of transformers exceeding their 25-year design lifespan, and the average age of transmission lines approaching 40 years.

When the "pulsed" electricity demand of artificial intelligence collides with the "aged body" of the power grid, this crisis not only severely limits the further development of the AI industry but also exposes the deep-seated contradiction between the long-term lag in U.S. infrastructure investment and the demand for emerging technologies. **If institutional barriers are not quickly broken and investment in the power grid is not increased, the computational advantage of the U.S. in the field of artificial intelligence is likely to evaporate due to power shortages.**

According to The Wall Street Journal, OpenAI's model named Orion consumed approximately 11 billion kilowatt-hours during two large training sessions lasting six months each. This figure is equivalent to the annual electricity consumption of 1 million American households and is close to the current annual electricity consumption of the U.S. steel industry. It is enough for a Tesla Model 3 to travel 44 billion miles, roughly equivalent to three round trips to Neptune.

The computational intensity and energy consumption during the operational phase are far lower than during the training phase, but as the number of users of such AI tools increases, the electricity demand during the operational phase will also continue to grow. Moreover, due to many companies and individuals fearing falling behind in the race for AI technology applications, the "latest and greatest" models often attract a large number of users, thereby exerting higher pressure on electricity demand.

On September 22, 2025, OpenAI announced a partnership with NVIDIA to build an AI data center with a power consumption of up to 10 gigawatts (GW). Andrew Chien, a computer science professor at the University of Chicago, stated, "A year and a half ago, they were discussing a 5 GW project, and now they have raised the target to 10 GW, 15 GW, or even 17 GW, showing a trend of continuous upgrades."

Each data center project by OpenAI is valued at approximately $50 billion, with a total planned investment of $850 billion. NVIDIA alone has committed $100 billion to support this expansion plan and will provide millions of new Vera Rubin graphics processing units.

While this example demonstrates significant electricity consumption, it is by no means an isolated case; other major players in the AI industry, such as Google, Meta, Microsoft, Amazon, and Anthropic, will also follow the same path when training the next generation of AI models.

Due to the urgent demand for energy, some data centers in the U.S. are choosing to build their own power generation facilities instead of relying on connections to state public grids. For example, in the western plains of Texas, a natural gas-powered generation project is under construction, which is not an investment project of traditional power companies but an important part of the $500 billion "Stargate" supercomputing center jointly built by OpenAI and Oracle At the same time, xAI is constructing two massive data centers named "Colossus" in Memphis, Tennessee, and has begun generating electricity using gas turbines. There are also more than ten data centers across the United States operated by Equinix, a global leader in digital infrastructure and data center services, that rely on fuel cells for power.

This trend is referred to as "Bring Your Own Power." Some call it an "energy wild west movement" that is reshaping the energy landscape in the United States.

However, there is significant social resistance at the local level. Although data centers involve substantial investment, their direct employment typically ranges from dozens to hundreds of jobs, far below traditional manufacturing projects. At the same time, their resource consumption is considerable: a large data center can use millions of gallons of water daily (equivalent to thousands of tons), primarily for cooling systems; its electricity consumption may reach 100 megawatts (MW) or more—equivalent to the electricity usage of a small city. Under this "high consumption—low employment" structure, local community dissatisfaction is gradually accumulating.

For example, in Loudoun County, Fairfax County, and Prince William County in Virginia, residents have protested the expansion of data centers multiple times, arguing that they drive up housing prices, occupy land, and exacerbate pressure on the power grid. Reports indicate that as of 2025, at least 25 proposed data center projects have been canceled due to local community opposition. In Oregon, some projects have been restricted by local governments due to water resource shortages. The visibility of these "infrastructure externalities" **has transformed data centers from mere commercial investment projects into local political issues in the United States.**

Overall, the development of data centers in the United States is subject to the combined effects of three constraints: first, the physical bottlenecks of grid infrastructure, which limit the speed of computing power expansion; second, structural instability during the energy transition, which raises electricity costs and risks; and third, conflicts between local communities and resources, which weaken the political feasibility of project implementation. Together, these factors create a new constraint mechanism that reveals the institutional boundaries of what was originally a highly flexible, market-oriented model of data center expansion in the AI era.

## **China's Unique Response**

**China's power grid has unique advantages in the global energy system, stemming from its scale, engineering capabilities, institutional coordination, and deep integration of technology and industrial chains.** It not only supports domestic industrialization, urbanization, and digitalization but also becomes an important strategic variable in the global energy transition and data center industry layout.

China's power system is the largest and most complex in the world—it has built the longest and largest UHV (Ultra High Voltage) transmission network, which, due to its long-distance and low-loss characteristics, can achieve "West-to-East Power Transmission" and "North-to-South Power Transmission," with no comparable examples globally. UHV enables the grid to connect to large renewable energy bases (wind, solar, hydro) and stably deliver power to load centers, providing a critical foundation for the absorption of new energy The interconnectivity and reliability of China's power grid are commendable, **with the power supply reliability in some cities reaching world-class levels**. The annual average power outage time in major cities of Beijing-Tianjin-Hebei, Yangtze River Delta, and Pearl River Delta is less than 1 hour per household, while the core areas of cities like Beijing, Shanghai, Guangzhou, and Shenzhen have annual outage times in the range of 1 minute, comparable to international first-tier cities like Tokyo and Singapore.

The structure of the large power grid brings economies of scale and redundant supply, enhancing system resilience.

At the same time, China has made breakthrough progress in UHV technology, covering the entire process from equipment manufacturing to engineering design, construction, and operation. **In the future, China's ultra-high voltage transmission projects will provide leading transmission solutions for more countries, and UHV will become China's "new business card."** In the field of UHV substation equipment, Chinese companies are world leaders, possessing a full range of ultra-high voltage products and dominating the formulation of international standards. In power equipment manufacturing and infrastructure construction, China has formed a complete industrial chain, giving it a global advantage in the cost, efficiency, and speed of large power grid projects.

China's power grid also has outstanding advantages in digital infrastructure and intelligent scheduling. Technologies such as AI-assisted scheduling, smart substations, and unmanned inspections have been deployed on a large scale, helping to manage the vast and complex multi-source power structure. At the same time, China is a global leader in the practice of "rapidly increasing the proportion of new energy while maintaining stable grid operation."

Such advantages of the power grid are gradually transforming into international influence.

Within the framework of the "Belt and Road" initiative, China has built or participated in large power projects in Southeast Asia, Africa, and the Middle East. Several Chinese power grid standards have entered the IEC and ISO systems, holding potential for standard-setting rights in future global power infrastructure upgrades (such as high-voltage direct current and smart grids).

China can also play a key role in the global energy transition: **To achieve an increase in the proportion of new energy globally, it is essential to rely on high-voltage transmission and Chinese-manufactured photovoltaic/wind turbine/storage equipment, which is why Musk's team is coming to China to procure equipment.** It can be said that China's large-scale experience in power grids and energy systems has a demonstrative effect globally.

Unlike the United States, which relies on global supply chains, China depends more on domestic industries for hardware and key materials, such as local servers, AI chips, optical fibers, and storage equipment. It also emphasizes domestic resource integration, green energy utilization, and coordination with national planning, presenting a uniquely Chinese infrastructure and digital strategic system, while also guiding domestic enterprises to participate in the global industrial chain, balancing self-control and international cooperation.

In terms of energy, China is vigorously promoting the integration of data centers with clean energy, laying out "green data centers" powered by photovoltaic, wind, and nuclear energy to reduce dependence on fossil fuels and enhance sustainable development capabilities.

Strategically, China emphasizes the combination of regional hubs and national planning—building ultra-large-scale data centers in core urban clusters such as the Guangdong-Hong Kong-Macao Greater Bay Area, Yangtze River Delta, and Beijing-Tianjin-Hebei, while connecting national regions through a "computing power network" to form computing power scheduling and inter-provincial collaboration capabilities.

However, it should be noted that China faces unique energy and structural risks in promoting the development of large-scale data centers First of all, the transformation of the energy structure is a long-term plan. **At this stage, China's data centers still have a significant reliance on coal power**, which brings considerable carbon emissions and environmental pressure. According to statistics, in 2024, the installed capacity of thermal power in China will still account for about 45% of the total installed capacity nationwide, with coal power being the mainstay. The demand for highly stable electricity from data centers makes it **realistically difficult to reduce the proportion of coal power in the short term**. High-energy-consuming industries are concentrated between the eastern coastal regions and the energy-producing areas in the central and western regions, which means that coordinating carbon reduction and energy supply is highly challenging.

Secondly, the development of data centers in China shows a clear east-west distribution pattern: computing power centers are mainly concentrated in eastern coastal cities such as Beijing, Shanghai, and Shenzhen, while electricity supply relies on the central and western regions. Long-distance electricity transmission inevitably generates line losses and increases dependence on the stability of the central and western power grids.

**The high concentration of data center layouts also brings potential systemic risks and resilience issues**. In the event of natural disasters, cyberattacks, or policy changes, it could cause a chain reaction impacting national AI services, cloud computing, and internet infrastructure.

Therefore, this year's government work report mentioned the concept of "computing power and electricity synergy," promoting the integration of computing power and electricity, optimizing the electricity supply structure, and eliminating risks related to stability. This is a more long-term plan and consideration that will take time to implement. However, it can be definitively stated that **in terms of energy supply, American artificial intelligence cannot cross the river by feeling the stones like China.**

Risk warning and disclaimer

The market has risks, and investment should be cautious. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this is at one's own risk

### Related Stocks

- [AI.US](https://longbridge.com/en/quote/AI.US.md)
- [GXAI.US](https://longbridge.com/en/quote/GXAI.US.md)
- [META.US](https://longbridge.com/en/quote/META.US.md)
- [ENFR.US](https://longbridge.com/en/quote/ENFR.US.md)
- [GRID.US](https://longbridge.com/en/quote/GRID.US.md)
- [AIYY.US](https://longbridge.com/en/quote/AIYY.US.md)
- [METU.US](https://longbridge.com/en/quote/METU.US.md)
- [GOOW.US](https://longbridge.com/en/quote/GOOW.US.md)
- [FBL.US](https://longbridge.com/en/quote/FBL.US.md)
- [GGLL.US](https://longbridge.com/en/quote/GGLL.US.md)
- [SRVR.US](https://longbridge.com/en/quote/SRVR.US.md)
- [POWR.US](https://longbridge.com/en/quote/POWR.US.md)
- [NUKZ.US](https://longbridge.com/en/quote/NUKZ.US.md)
- [ARTY.US](https://longbridge.com/en/quote/ARTY.US.md)
- [GOOG.US](https://longbridge.com/en/quote/GOOG.US.md)
- [TSLA.US](https://longbridge.com/en/quote/TSLA.US.md)
- [GOOGL.US](https://longbridge.com/en/quote/GOOGL.US.md)

## Related News & Research

- [Meta Offers WhatsApp Access to AI Rivals in Europe, but There's a Catch](https://longbridge.com/en/news/286959569.md)
- [China's Manus founders seek $1 billion to undo Meta takeover, Bloomberg News reports](https://longbridge.com/en/news/287194973.md)
- [Google gives first glimpse of new AI glasses ahead of fall launch](https://longbridge.com/en/news/286960810.md)
- [How High Can Meta's Stock Price Rise?](https://longbridge.com/en/news/287101660.md)
- [Meta's 8,000 Job Cuts To Foot The $145B AI Bill](https://longbridge.com/en/news/286819463.md)