96 electricity prices per day: Aoxin leverages AI to build a "trading brain" for power assets.

Nanjing Aoxinguang Intelligent Industry Development Co., Ltd. (hereinafter referred to as "Aoxinguang") has recently launched a new round of financing, planning to raise 10 million RMB by transferring 10% to 15% of its equity. The raised funds will be mainly used for the iteration of AI trading models, the deployment of nationwide sales channels and the expansion of the trading team. Nanjing Aoxinguang is a company focusing on the "AI + Power Assets" intelligent decision-making track, providing full-link AI trading platforms covering forecasting, decision-making, risk control and automatic execution for market entities including power generation, power sales, energy storage and virtual power plants.

▍ I. Electricity Prices Enter the "96-Point Era", Trading Capability Becomes the Lifeline for Market Entities

The market-oriented reform of the power industry is pushing the entire sector into the era of high-frequency trading. Data from the National Energy Administration shows that in 2025, the traded electricity volume in the national power market reached 6.64 trillion kWh, a year-on-year increase of 7.4%, accounting for 64% of the total social electricity consumption; the continuous operation of provincial spot markets has basically achieved full coverage, and the number of business entities registered at trading centers has exceeded 1 million. Taking the Shandong spot market as an example, the clearing price is updated every 15 minutes, generating 96 different electricity prices within one day — the old business model of "earning spreads effortlessly" relying on one annual contract has completely failed.

The market performance in 2026 has intensified this industry reshuffle. Affected by factors such as rising primary energy prices and fluctuations in power consumption loads, spot electricity prices in many provinces have risen sharply, and the power retail industry has fallen into the "wholesale-retail price inversion" dilemma: the fixed-price retail contracts signed at the beginning of the year locked the price at a low level, while the procurement cost on the wholesale side kept rising. In some regions, power retailers suffered a loss of more than 0.3 yuan per kWh, many central SOE-backed power retailers jointly requested relief, and the industry-wide loss and exit wave has become public news. As of the end of September 2026, there are about 3,700 power retailers actually participating in transactions across the country, and the industry generally judges that enterprises lacking capabilities in forecasting, quotation and risk control will be cleared out at an accelerated pace.

At the same time, new energy has fully entered the market. In 2025, the market-oriented traded electricity volume of new energy reached 1.35 trillion kWh, accounting for 60% of the total new energy power generation. For power generation enterprises, the assessment cost caused by forecast deviation remains high; for energy storage owners, the internal rate of return is highly dependent on charge and discharge strategies; the rules of multiple markets including spot, medium and long-term, and ancillary services are different and still being iterated on a weekly basis — the traditional operation method relying on "manual work + Excel + personal experience" can no longer keep up with the rhythm of the market.

▍ II. Full-Stack Self-Developed AI Algorithm, Building the "Trading Brain" for Power Assets

Aoxinguang's entry point is to use a set of full-stack self-developed AI algorithm system to replace manual work in completing core decision-making in power trading. The full-link intelligent trading platform built by the company covers four scenarios: spot quotation for new energy stations, multi-market arbitrage for energy storage, virtual power plant operation and power retail risk control. The underlying logic of the platform is "accurate forecasting, optimal quotation and stable risk control" — it outputs price and load forecasting through the energy time-series large model, generates multi-market coordinated trading strategies via reinforcement learning algorithms, and then the risk control module conducts real-time monitoring and automatic execution for deviation exposure.

In terms of technology, the company adheres to full-stack independent R&D, has obtained 13 software copyrights and 1 invention patent, and has passed the ISO dual system certification. Its algorithm team is led by a PhD from Tongji University, focusing on the direction of energy time-series large models and reinforcement learning.

What is more noteworthy is the actual combat data. In the power generation side trading of a 50MW wind farm in Shandong, the platform helped the owner achieve an annual revenue increase of 702,400 yuan; on the power retail side, it brought an annual revenue increase of 800,000 yuan to a power retailer with an agency electricity volume of about 300 million kWh. At present, the company's business has covered more than 20 provinces across the country, and has completed the engineering verification of strategies under the rules of multiple provincial spot markets — this point is particularly critical in the power industry with frequently iterated rules: only when the system is launched and adapted quickly, the strategies will not become obsolete immediately after going online.

▍ III. Starting from Practical Trading Operations, Moving Towards Platform-Based Barriers

Aoxinguang's commercialization path is divided into three steps: in the short term, it enters the market with practical trading operation services, and builds customer trust with verifiable real returns; in the medium term, it matures the strategies into products, replicates them to multiple provinces and multiple types of entities, and forms a revenue structure of "subscription + revenue sharing"; in the long term, it will precipitate into an intelligent operation platform for power assets, and build competitive barriers through accumulated data and strategies. This rhythm of "delivering returns first, then scaling up" also matches the "result-oriented" procurement habit of its industry — customers ultimately only pay for the revenue-increasing effect.

In terms of the team, the core members of the company cover four lines: strategy, capital, R&D and trading: General Manager Zhao Hui holds a Master of Business Administration degree from Nanjing University of Finance and Economics, with more than 10 years of experience in the AI industry and marketing, and more than 5 years of in-depth tracking of the power spot market; Co-founder Jiang Kun has 18 years of experience in the securities industry, responsible for capital vision and resource integration; Lv Qingcheng, the head of R&D, is a PhD from Harbin Institute of Technology, who has led the implementation of power trading systems in more than ten provinces; Zhao Gang, the core head of algorithms, is a PhD from Tongji University, who leads the development of electricity price forecasting and energy storage scheduling algorithms; trading expert Wang Yajing graduated from the major of electrical engineering of North China Electric Power University, and once served as the head of trading operation at Envision Digital.

The 10 million yuan raised in this round of financing will be used in three directions: AI model iteration to maintain the leading position in forecasting accuracy and strategy adaptability; nationwide channel deployment to accelerate the large-scale replication of business in multiple provinces; expansion of the trading team to improve the capability of parallel operation in multiple markets. Under the dual window period of the power retail industry's market clearing and the full entry of new energy into the market, trading capability is evolving from a "plus point" to an "admission ticket", and what Aoxinguang hopes to seize is exactly this round of opportunity for capability upgrading.