Price Spread Reaches up to 2000x! Calls for Building a National Computing Power Exchange Are Growing

Log in to Sina Finance App and search for 【Information Disclosure】 to see more evaluation levels

When it comes to stock trading, look at Jin Qilin analyst reports—authoritative, professional, timely, comprehensive, helping you discover potential thematic opportunities!

【Introduction】Price differences reach up to 2000 times, calls for establishing a nationwide computing power trading platform are growing

China Fund News Reporter Lu Ling

Computing power is becoming the “water, electricity, and coal” of the AI era. The “14th Five-Year Plan” proposes to appropriately and proactively build a nationwide integrated computing power network.

However, the current pricing system for computing power services is severely distorted, with price differences sometimes reaching hundreds or even thousands of times.

Against this backdrop, the call to establish a nationwide computing power trading platform is gaining momentum.

Price differences in computing power services reach hundreds to 2000 times

According to Zhang Yunquan, a member of the National Committee of the Chinese People’s Political Consultative Conference and researcher at the Institute of Computing Technology, Chinese Academy of Sciences, currently, the prices for similar computing power services vary by hundreds of times—Lite version of Doubao costs only 0.3 yuan per million tokens, while Wenxin 4.0 version costs up to 120 yuan, reflecting a serious distortion in the pricing system.

At the infrastructure level, Nvidia H100 GPU rental prices soared from a peak of $8 per hour in 2023 to about $2 per hour by the end of 2025, a decline of over 70%, approaching a deterministic loss price of $1.65 per hour. At the application level, in February 2025, DeepSeek launched off-peak discounts of 75% to $0.035 per million tokens, while at the same time, GPT-4.5’s pricing was as high as $75 per million tokens, with a price gap reaching 2000 times. Some local governments subsidize 10-15% of the investment in computing centers, further distorting market prices.

Moreover, current computing power trading still mainly relies on bilateral negotiations, with different trading rules across regional platforms, lacking a unified computing power price index or risk management tools like futures. High-end GPU resources are concentrated in large tech companies, while small and medium enterprises and research institutions have limited access, leading to severe information asymmetry. Compared to mature commodity markets like oil and electricity, the price discovery and trading mechanisms for computing power are significantly lagging behind infrastructure development.

Calls for establishing a nationwide computing power trading platform are growing

According to Zhang Yunquan, referencing the “kilowatt-hour” standard in the electricity market, the computing power market urgently needs to establish a standardized measurement unit to lay the foundation for fair pricing.

In the proposal “On Regulating the Formation Mechanism of Computing Power Prices and Developing a Healthy and Sustainable Computing Power Economy,” Zhang Yunquan suggests formulating and improving the “Computing Power Service Measurement and Pricing Standards,” establishing a unified computing power measurement system, and creating an intelligent computing power pricing standard based on “Yuan per million tokens.” He proposes setting up an independent third-party national computing power trading platform, separate from the demand and supply sides, to build a unified national computing power trading market; establishing a computing power matching center to provide centralized listing, open bidding, and matching services, covering spot trading and forward contracts; implementing a computing power information disclosure system, compiling and releasing a “National Computing Power Price Index,” and exploring the launch of computing power futures and options when conditions are ripe, as well as future asset securitization of computing power (computing power REITs).

National People’s Congress deputy and Dongfang Caifu chairman also believes that China can explore “computing power futures” and even establish a “computing power exchange.” Computing power futures could reduce companies’ price fluctuation risks and stabilize expectations; when computing power service prices fluctuate sharply, the trading market for computing power futures could lock in costs and stabilize expectations.

“Within 2-3 years, related computing power trading products can be launched,” Zhang Yunquan told reporters.

At the “360 Security Lobster” AI agent launch event on March 14, Zhou Hongyi, member of the National Committee of the Chinese People’s Political Consultative Conference and founder of 360, told reporters that once AI computing power is used to drive intelligent agents like lobsters, making them personal assistants for everyone, the future demand for computing power could increase by 100 to 1000 times; under such massive demand, computing costs will further decrease, potentially by a factor of 10.

Zhou Hongyi believes that in the future, computing power is likely to become a necessity like water and electricity for businesses and individuals, turning into a bulk commodity that everyone can afford.

Proofreader: Wang Yue

Editor: Xiao Mo

Reviewer: Xu Wen

Massive information, precise analysis, all on Sina Finance APP

责任编辑：Guo Xutong