Computing power surge ignites global electricity demand

If you’re trading stocks, just look at reports from analyst Jinqilin—authoritative, professional, timely, and comprehensive, helping you uncover high-potential themes and opportunities!

Reporter | Cao Enhui

Editor | Zhang Xing

A global wave of electricity demand ignited by computing power has already been set in motion.

According to media reports, on April 1, 2026, Microsoft, Chevron, and the investment fund Engine No. 1 reached an exclusive agreement on a power generation and power supply matter. The agreement will support a large energy complex located in West Texas, aiming to provide electricity to a large-scale data center (AIDC) park.

This is another “power scramble” staged by global tech giants as they seek a stable source of electricity supply. AI computing power, taking on the posture of a “power-eating black hole,” is pushing the world into an unprecedented super electricity cycle.

China has already responded in a well-thought-out way. In 2026, “computing-and-power coordination” was included in the government work report for the first time, rising to the strategic level of a new category of national infrastructure buildout.

In this global super electricity cycle, how should Chinese companies seize the lead?

A “power-eating black hole” is taking shape

In the Intelligent World 2035 series report released by Huawei last September, it predicted that by 2035, the total amount of computing power for all of society globally will grow by 100k times. With AI-driven computing power, the growth pace can no longer be explained by the traditional Moore’s Law.

A supporting data point: this year in March, China’s daily Token call volume broke through 140 trillion, growing over 1,000 times in two years.

The exponential expansion of AI computing power is reshaping the growth curve of global electricity consumption.

A report released this year by the International Energy Agency (IEA) points out that electricity demand outpacing economic growth will become a common trend in the coming years. From 2026 to 2030, global electricity demand is expected to grow at an average annual rate exceeding 3.5%, with major drivers coming from the growth in electricity usage for industry, electric vehicles, air conditioners, and data centers.

A set of figures further reflects the “quality” of data centers as a “power-eating black hole.” Also from the IEA’s forecast, in 2024 global data centers’ total electricity consumption reached 415TWh (terawatt-hours), roughly equal to the UK’s electricity use for the entire year; by 2030, this number will be close to Japan’s total electricity consumption nationwide. A reporter from 21st Century Business Herald noted that in recent years Japan’s total electricity consumption across society has been about 1,000TWh. Over the past five years, the compound annual growth rate of global data center electricity consumption has been 12%, which is more than 4 times the global average electricity consumption growth rate.

In recent years, global tech giants have been deeply laying out data centers, whose electricity consumption far exceeds traditional electricity demand. “In the next five years, computing power will see explosive growth in relation to electricity demand.” Lin Boqiang, president of the China Energy Policy Research Institute at Xiamen University, said in an interview with a reporter from 21st Century Business Herald.

Why does AI computing power become a “power-eating black hole”? An industry insider told a reporter from 21st Century Business Herald: “Four factors make its energy consumption far exceed that of traditional computing: high-density, high-power-consumption hardware; demand for compute at the exponent/number scale; the dual losses from data movement and thermal dissipation; and continuous expansion of inference throughput.”

In addition, the popular science newspaper Ke Pu Shi Bao, citing the viewpoint of an engineer from the Institute of Computing Technology of the Chinese Academy of Sciences, stated that AI’s electricity consumption is mainly concentrated in two areas: model training and inference. “Taking the GPT-3 large model as an example, its total training electricity consumption is about 1,280 megawatt-hours, i.e., 1.28 million kWh, equivalent to the electricity usage of 6,400 typical Chinese households in one month.”

Since GPT-3 was released, and especially since 2025, demand for computing power has shown rapid growth. In November 2025, Amin Vahdat, head of Google’s AI infrastructure, said that Google must double its AI computing power every 6 months, and achieve an additional 1,000-fold growth within the next 4 to 5 years, to meet continuously rising demand for AI services.

Tech giants are accelerating their investments related to AI. According to a credit rating firm’s (Pengyuan Credit) statistics, in 2026, the combined planned capital expenditures of the four major U.S. tech giants—Amazon, Alphabet (Google’s parent company), META, and Microsoft—are expected to reach $650 billion, a year-on-year increase of 60%. Meanwhile, most of this large spending is directed to building new data centers and related supporting equipment.

Once large-scale data centers are built, they will form long-term, stable electricity consumption. Therefore, there is a view that electricity is no longer a “supporting factor” for computing power, but rather the “ceiling” that determines how computing power grows.

“Grid-building anxiety” in Europe and the U.S., and the opportunity for China’s power sector under “Token going global”

As the global super electricity cycle arrives, however, the electricity markets in Europe and the U.S. have fallen into “anxiety”—AI’s electricity-hungry compute demand is surging, colliding with the “aging period” of the power grids in Europe and the U.S., and supply-demand contradictions are instantaneously magnified, exposing the fragility of power infrastructure.

The power grids in Europe and the U.S., which were generally built in the 1960s and 1970s, are entering a phase of “overextended service.”

In 2020, the U.S. Department of Commerce released a report stating that the actual average service life of U.S. transformers is 30 to 40 years, far exceeding their expected lifetime of 25 years. Moreover, 80% of transformer supply depends on imports. When the European Commission released the “Grid Action Plan” in 2023, it also said that about 40% of the EU’s distribution grids have already been in use for more than 40 years, and it is expected that it will require €100k (1 hundred billion euros) for maintenance, improvements, and upgrades to Europe’s power grid and related facilities.

Analysts have been quick to see the windfall era for power equipment. In a research report recently jointly released by its teams across multiple regions, Jens Spiess, an analyst at Morgan Stanley’s Mexico division, said: “The U.S. power grid is experiencing a situation of supply-demand mismatch, and this situation will last at least until 2030. The market size for large power transformers (LPT) will expand with an annual compound growth rate of about 14%.”

It further pointed to factors on the demand side—America’s old and outdated grid infrastructure urgently needs to be updated; large-scale grid connections of wind power and solar PV create new transmission demand; and the explosive expansion of data centers.

Analysts focus on transformers because they are core equipment for connecting data centers to power grids and for integrating new energy. However, affected by issues such as shrinking domestic production capacity, a shortage of skilled technicians, and material shortages like upstream-oriented silicon steel, transformers in Europe and the U.S. are caught in a supply dilemma. In essence, this is a concentrated surge driven by long infrastructure investment cycles, an imperfect industrial chain, and misaligned timetables for the energy transition.

To address this, both Europe and the U.S. have rolled out large-scale investment plans to upgrade and transform their power grids. CICC (China Aviation Securities) stated that in December 2025, Europe proposed a package plan for the power grid, expected to unlock €1.2 trillion in grid investment; and the Global Times, citing foreign media, said that from 2025 to 2029, U.S. power companies are expected to spend $1.1 trillion on grid upgrades.

China’s power opportunities under “Token going global”

On March 23, 2026, the head of the State Data Bureau, Liu Lihong, officially announced at the China Development High-Level Forum: the Chinese translation of the core terminology “Token” in the field of artificial intelligence is “Token.”

After the Spring Festival, “Token going global” became a new business narrative. Chinese AI large-model companies turn advantages in electricity, computing power, and the like into high value-for-money Tokens, and provide inference services to the world via API interfaces, charging based on processing volume, thereby achieving a “digital export” of computing power and electricity.

In practice, although the business narrative logic above is still under discussion, behind it lies the initiative gained from China’s power industry having laid out plans in advance—namely, through a way of organizing and coordinating at the national-strategy level, with forward-looking planning and integrated collaboration across the whole industrial chain, China has built a new development model of “computing-and-power coordination,” establishing a first-mover advantage in the super electricity cycle.

The outline of the “15th Five-Year Plan for 2026–2030” proposes accelerating the construction of new-type energy infrastructure. Focus on building a new-type power system, comprehensively improve the power system’s complementary support and security and resilience, optimize the layout of electricity flows nationwide and cross-regional channels, accelerate smart grid construction, improve urban and rural distribution grids, scientifically plan pumped storage, and vigorously develop new-type energy storage.

A reporter from 21st Century Business Herald noted that during the “15th Five-Year Plan for 2026–2030” period, it is precisely the critical five years when China’s power grids are being upgraded and to support the explosive growth of computing power. On the one hand, the grid continues to increase investment—State Grid and Southern Grid each plan to invest ¥4 trillion and ¥2 trillion, respectively, with key funding directed to supporting facilities such as ultra-high-voltage, smart distribution networks, new-type energy storage, and computing-and-power coordination, thereby fully building an integrated “generation-transmission-distribution-storage” new-type power system; on the other hand, policy sets the tone for the “computing-and-power coordination” strategy—explicitly proposing “implementing new infrastructure projects such as ultra-large-scale intelligent computing clusters and computing-and-power coordination,” which marks China’s top-level design shifting from “separation of electricity and computation” to “integration of electricity and computation.”

“Currently, the world is entering a new energy era driven by decentralization, digital intelligence, and decarbonization.” Gao Feike, Executive Vice President of Schneider Electric and head of the energy management business, said. In this profound transformation, China shows global leading advantages by leveraging its strengths in both “speed and quality,” becoming the first important power country in human history. “Its leading role not only comes from its huge market scale, but also from its outstanding vision and systematic layout and planning—through comprehensive, continuous, and coordinated investment on the supply side, the power grid, and the demand side, thereby forming competitiveness.”

In Gao Feike’s view, today industries such as industry, transportation, and buildings are accelerating their adoption of electrification; electricity supply-demand balance management has been upgraded into a highly complex systems engineering project. “And that is the core battleground where digital-intelligence technologies such as artificial intelligence can deliver disruptive value.”

“The core of ‘computing-and-power coordination’ is to break down the barriers where computing power and electricity each operate independently, enabling two-way matching of ‘compute follows electricity and electricity follows compute.’ And this power super cycle is also a direct manifestation of the deep integration of energy infrastructure and digital infrastructure.

From big-grid investment commitments, to policy leadership for computing-and-power coordination, to the supporting capabilities across the entire industrial chain, China’s power industry has already seized the opportunity ahead of others.

		Sina Statement: This news is a repost from Sina’s partner media. Sina.com has published this article for the purpose of transmitting more information, and does not mean that it agrees with or confirms the views or descriptions in the report. The article content is for reference only and does not constitute investment advice. Investors act at their own risk.

A wealth of information, precise analysis—available in the Sina Finance APP

Editor-in-charge: Song Yafang