The year 2026 has made the statement, "The ultimate limit of computing power is electricity, and the ultimate limit of electricity is energy strategy," more tangible than ever. As the world's largest electricity producer, China has also established the biggest and fastest-growing renewable energy system globally, with nearly 4 out of every 10 kilowatt-hours consumed coming from green power. Simultaneously, China ranks second worldwide in total computing power capacity, giving it a distinct advantage in promoting the integration of computing and electricity. The national "15th Five-Year Plan" outlines the goal of coordinating the layout of green power and computing resources. This year, the State Council’s government work report included "compute-power synergy" for the first time as part of new infrastructure projects, accelerating the deep integration of these two critical infrastructures.

Recent data highlights significant shifts. On the international model aggregation platform OpenRouter, Chinese large language models now account for 85.7% of global weekly API calls. Among the top five most frequently used models globally, four are developed by Chinese companies. Why are these changes occurring now? Primarily, the commercial logic of artificial intelligence (AI) has evolved, becoming more reliant on "data + algorithms + computing power." AI itself is a key variable for China in seizing developmental initiative and shaping competitive advantages. In a recent speech, NVIDIA CEO Jensen Huang noted that AI has moved beyond the pure training phase into the realm of inference. This shift has directly increased the required volume of tokens and computational load by approximately ten thousand times compared to before.

Tokens are the smallest units of data used in text processing. In language models, long texts are broken down into token-based data structures, segmented for model computation. For example, in Chinese, the phrase "我很开心" might be split into three tokens: "我," "很," and "开心." Much like "traffic" in the internet era, tokens have become the hard currency of the intelligent age. The recent popularity of OpenClaw (nicknamed "Lobster") has accelerated the value transition of tokens. Unlike previous AI interactions that followed a simple question-and-answer format, OpenClaw enables AI to run over ten subtasks simultaneously in the background, continuously calling contextual information and performing iterative cycles, leading to an exponential rather than linear increase in token consumption. Analyzing the cost of tokens reveals that computing power and electrical energy are its core components. The "compute-power synergy" policy thus carries strategic significance.

For Sichuan, this represents a historic opportunity. To seize it, the province must first assess its inherent advantages. Sichuan holds two strong cards. The first is undoubtedly energy. Jensen Huang recently proposed a "five-layer cake" theory for the AI industry, placing energy at the foundation, upon which all upper-layer applications depend. This means any successful application relies on the stable operation of underlying infrastructure, even down to power plants. In this regard, Sichuan is ahead in large-scale green energy supply—last year, it achieved a historic breakthrough with installed hydropower capacity exceeding 100 gigawatts, maintaining its position as the national leader. Moreover, Sichuan’s hydropower potential remains 30% to 40% untapped. Additionally, installed new energy capacity has surpassed 32 gigawatts and is projected to reach 82 gigawatts by 2030, accounting for over 45% of the province’s total power capacity. These resources provide ample, green energy to drive computing power.

The second card is an established computing power base. As a national hub node in the "East Data, West Computing" project, Sichuan has built facilities such as the National Supercomputing Center in Chengdu and the Chengdu Intelligent Computing Center, forming an integrated supply system that includes general, intelligent, and supercomputing capabilities. Intelligent computing alone totals 16.8 EFlops, representing nearly 60% of the total. A province-wide integrated computing power monitoring and scheduling platform is already operational, supporting flexible allocation of computing resources. These two advantages position Sichuan with a first-mover edge in implementing "compute-power synergy."

How can Sichuan translate this early advantage into industrial and developmental success? The key lies in understanding why electricity and computing power are not yet fully synergistic. Expert analysis points to mismatches in both space and time. Spatial mismatch refers to the current "east computing, west electricity" pattern, where computing demand is concentrated in the economically developed eastern regions, while green power resources are largely in the west. Furthermore, computing facilities have short construction cycles and rapid iteration, whereas power grids and new energy projects require large investments and long timelines, leading to situations where "computing waits for power" or "power waits for computing." Temporal mismatch stems from the intermittent nature of green energy—solar and wind power depend on weather conditions, generating power during daylight or windy periods but lacking output at night or during calm weather; hydropower is affected by wet and dry seasons. The emergence of intelligent agents like "Lobster" further demands stable, continuous computing power.

Previously, electricity and computing operated independently, often resulting in waste. "Compute-power synergy" aims to integrate the two, enabling more efficient use of stable energy and computing resources. Experts suggest that as synergy mechanisms mature, western inland regions can leverage low green power costs to transition from mere energy exporters to strategic hubs offering both "green power + computing power." This brings the discussion back to token costs. As a new form of "commodity," tokens could be traded nationally and even globally. If Sichuan achieves compute-power synergy, it could lower overall operational costs based on its electricity cost advantage, significantly boosting profit margins for AI-related businesses.

During the National People's Congress sessions, representatives actively proposed suggestions on compute-power synergy. Industry professionals recommended supporting provinces rich in hydropower, like Sichuan, in conducting direct green power connections and promoting local conversion of green energy resources. Others proposed enabling direct links between computing centers and green, hydro, or nuclear power plants, with policy support in transmission fees and project approvals to reduce energy costs. Sichuan is already accelerating its deployment of compute-power synergy. In April 2025, the province issued an implementation opinion supporting accelerated integration, proposing the construction of data centers directly supplied by green power and prioritizing 10,000-card computing clusters in regions like Panzhihua, Aba, Garze, and Liangshan, which meet criteria such as abundant clean energy and constrained power transmission sections. This year’s provincial government work report further emphasized speeding up and optimizing the national hub node for "East Data, West Computing," advancing the construction of compute-power integration parks, scientifically planning data and computing infrastructure, and supporting public cloud development.

Concurrently, Sichuan is formulating a work plan for compute-power integration, improving support measures for siting "green power + computing" facilities in clean energy-rich areas, establishing data industrial parks, implementing direct green power supply and electricity price support policies, and optimizing clean energy power supply and backup plans. Reflecting on Jensen Huang’s "five-layer cake" theory, Sichuan’s strength lies at the foundational layer. While a solid foundation is crucial, merely serving as a "base" would prevent sharing in upper-layer value. Transitioning from selling electricity to selling computing power and then to selling tokens—maximizing the value of each kilowatt-hour—is an essential question for energy-rich provinces and a strategic move for Sichuan to grasp the opportunities of the new technological revolution.