New development of INPOWER smart storage empowered AI computing industry

As a key infrastructure in the era of digital economy, data centers are undergoing technological upgrading and green transformation. In the rapid iteration of artificial intelligence technology, the traditional Internet data center (IDC) is accelerating the evolution of artificial intelligence data center (AIDC) , according to pricewaterhousecoopers analysis, the development of data centres has gone through the following four main stages:

1. Network Center Stage (before 2000) : in this stage, the data center is the main carrier of network traffic, mainly to meet the needs of enterprise informatization and basic data storage, enterprise information, network and storage requirements of the initial growth is its main driving factors;

2. It Center Stage (2000-2010) : in this stage, the service content of data center is diversified, including value-added services such as data storage, backup and security. The popularization of the internet and the complexity of enterprise IT system are the main driving factors

3. The Cloud Center Stage (2010-2020) : large-scale data centers have become the mainstream form of the industry, and the construction entities have diversified patterns such as cloud service providers and telecom operators, the maturity of cloud technology and the explosion of cloud service demand are the main driving factors in this stage

4. Computing Power Center Stage (2020-present) : in this stage, the demand for AI computing power has increased significantly. AIDC has gradually become a new form of data center development, the main driving factor of this stage is that artificial intelligence technology is widely used in the increase of computing power-intensive tasks.

Source: AIDC White Paper, Zeping Macro, Huawei

The core difference between AIDC and IDC

AIDC, also known as AI Data Center or AI-dedicated Data Center, is often referred to as an"Intelligent computing Center", which can be regarded as a"Computing factory" focused on artificial intelligence training. Its core objective is to specifically provide powerful computing power for AI, focusing on training and running various AI large models with extremely high power demand on GPU clusters. In contrast, traditional IDC is more inclined to"Information warehouse" role, the core function is the data storage, management and distribution. Like a vast modern library or warehouse, it ensures that vast amounts of data (documents, videos, user information, transaction records) are stored securely, reliably, and efficiently, and can be quickly retrieved and transmitted when needed.

In terms of electrical characteristics, the difference between the two is significant. IDC's power consumption is relatively"Flat," with the majority of power consumed by IT devices (servers, storage, networking) and cooling systems (air conditioning) , although some large idcs have higher total power consumption, however, the load fluctuation is relatively small, and the power demand curve is relatively flat. AIDC, on the other hand, is an"Energy hog" and its power consumption is"Explosive" and"High-density," with a cluster of AI servers consuming tens or even hundreds of times as much power as a traditional server. It is AIDC's high dependence on electricity and high energy consumption that makes its integration with energy storage technology an inevitable trend and core competitiveness.

Source: Ai DC White Paper Huawei,

"2024 data center industry investment and value insight", zeping macro

From the perspective of industry operation rules, the computing power load of AIDC shows significant"Tidal characteristics", which are deeply bound to the high-frequency demand scenarios of people's livelihood. According to the Meituan Research Institute, the daily peak of food take-away orders is from 11:30 pm to 13:00 pm and from 17:30 pm to 19:30 pm, with peak orders up to 3.2 times that of the off-peak period. Didi travel big data shows that, taxi requests during morning and evening commutes (7:00-9:00 and 17:00-19:00) increased 2.8 times compared to off-peak hours. Affected by such demand fluctuations, the load power consumption may fluctuate by more than 80% of the rated power consumption, that is, the load power consumption of the intelligent computing center may quickly change from 10% to 80% or even 100% , and there will be 3-4 obvious load peaks in a single day, which is a typical"Multi-frequency dynamic load" scenario. (data source: "The White Paper on the construction of computing power factory", "The white paper on the development of intelligent computing center", etc.) 

In contrast, the traditional IDC mainly carries static services such as web storage and e-mail service, and the load fluctuation range is stable at about 5%-10% for a long time. The overall running state shows the characteristics of"Low fluctuation and high stability". This stable load characteristics make the traditional IDC can not achieve cost optimization through"Peak-valley arbitrage"(that is, the use of electricity price peak-valley difference to achieve cost optimization) . AIDC, on the other hand, can use"Peak-valley arbitrage" to increase load to absorb low-cost electricity during low-cost hours and reduce load during peak hours to reduce high-cost electricity consumption, with the help of algorithms such as Shangtang Lingang intelligence computing center's"15-minute prediction + 5-minute correction" scheduling, while ensuring the core computing power demand, it can actively stabilize the peak power consumption and reduce the maximum demand value in the cycle, and then greatly reduce electricity demand, forming a"Peak-valley arbitrage + demand optimization" of the dual cost advantage.

The demand problem of computing power center is essentially the contradictory product of the interaction between high power density load and power grid-the instantaneous fluctuation of core loads such as AI training cluster can reach tens of megawatts, the periodic consumption of auxiliary systems such as air conditioning is easy to form a peak demand far exceeding the average load, which directly pushes up the demand electricity price and threatens the stability of the power grid. The energy storage system can physically stabilize the load fluctuation by"Peak clipping and valley filling", but only by forming a Trinity cooperative mechanism with algorithmic decision-making and EMS (Energy Management System) scheduling, in order to realize the precision and economy of demand control, the essence of this process is the deep coupling of"Cyber-physical-energy" system. In its technical nature, this synergy is a paradigmatic example of energy management enabled by computing power-algorithms translate computing power into decision-making capabilities, and EMS transforms decision-making into physical regulation, the organic combination of the two makes energy storage upgrade from passive standby equipment to active demand management center. With the implementation of the"East Counts West" project, this model will become the core support for the coordinated operation of the Computing Power Center and the new energy base. While reducing its own demand cost, for the new power system to provide millions of kilowatt-level flexible regulation resources.

Global Energy and technology industry spending increases, accelerate AIDC layout

AIDC has become the core infrastructure to support large-scale model training and intelligent reasoning, and its unique operating characteristics are forming a deep symbiotic relationship with the energy storage industry. This closed-loop system, in which computing power demand drives energy change and energy innovation feeds back computing power development, has not only created hundreds of billions of market opportunities, but also reshaped the competitive landscape of the global energy and technology industries.

According to the forecast of a battery head enterprise, data center energy storage battery shipments will soar from 10GWh to 300GWh from 2024 to 2030, a compound growth rate of more than 76.3% , corresponding to 150 billion yuan equipment market, the growth space is very considerable; Third-party estimates suggest that the number of racks in data centres will reach 181.3 m by 2030, thanks to advances in AI technology and AIDC Computing Power, led a surge in global data centre storage capacity from 16.5 gwh at 2024 to 209.4 gwh in 2030, a more than 12-fold increase in six years.

Demand-side expansion stems from both internal and external dynamics. In overseas markets, the four major cloud vendors, including Google and Meta, spent $95.8 billion on Q2 capital expenditure in 2025, up 64% year-on-year. Google raised its full-year expenditure to $75 billion, up about 43% year-on-year, in the United States, the Middle East, Southeast Asia and other places with weak power grids, AIDC supporting energy storage has become a rigid demand to ensure stable operation, and large projects of 1-2 gwh have emerged. In the domestic market, bytedance's AI capital expenditure is expected to exceed 150 billion yuan in 2025, and the three major operators' computing investment is growing by more than 20% year-on-year. In August, a Hong kong-listed energy storage company reported a 113.1% year-on-year jump in revenue from its AIDC business, raising its revenue share to 47.2% , more marks the track from the"Technical pilot" into the"Large-scale commercial" stage, the head of the enterprise has explored a clear path of profit.

Policy escort

The rapid development of the market can not be separated from the precise guidance of policies. Since 2025, our country has intensively introduced a number of policies to build a multi-level support system for the AIDC + Energy Storage Industry. The action plan for the high-quality development of the new energy storage manufacturing industry, released in February, for the first time included AIDC storage as part of the national strategy, and explicitly proposed"Promoting new energy storage configurations for intelligent computing centres", in May, the"Green power direct connection development notice" encouraged AIDC to adopt the"Green Power + energy storage" model, lowering the threshold for the use of clean energy In July, the green data center recommendation mentioned renewable energy and energy storage utilization as one of the evaluation indicators, with a total score of 10, directly improving the economy of storage and distribution.

Successful case: Shangtang Lingang Intelligence Computing Center project

Shangtang Lingang Intelligent Computing Center is an important benchmark of digital energy integration in Shanghai Lingang area, and it is also the first 5-year-level intelligent computing center in the country. It adopts distributed cluster architecture design, the system adopts a three-layer architecture of"Hardware cluster-software scheduling-intelligent coordination" to realize the efficient release and accurate matching of computing resources. As of August this year, the total computing power of sensetime's devices has steadily grown to 25,000 PetaFlops, according to the company's co-founder and president of the large device business group, energy Efficiency (PUE) was reduced from 1.74 to less than 1.3 through computational and electrical synergy.

At the beginning of the project, the integration, stability and adaptability of energy storage equipment were put forward extremely high requirements. According to the characteristics of the project, INP has customized the 5 MW boost integrated cabin scheme, which breaks the limitations of traditional equipment layout with highly integrated design and becomes the core guarantee for the smooth progress of the project.

The one-stop integrated cabin not only greatly saves the installation space and 

construction period on the project site, but also reduces the line loss between the 

equipment and improves the overall energy efficiency. Under the energy storage

 demand of the Shangtang Lingang project with a total scale of 18 MW/36 mwh,

 the scheme effectively simplifies the system debugging process, ensures that the

 energy storage power station is quickly connected to the 10 kv energy storage booster

 station, and seamlessly connects with the energy system of the computing center. 

At the same time, due to the special geographical environment near the sea in Shanghai 

Lingang area, the anti-corrosion performance of outdoor energy storage equipment is 

severely challenged. Inbev intelligent storage integrates C 5 anti-corrosion grade

 standard into the design of core equipment, which builds a reliable barrier for the

 long-term operation of equipment.

In addition, its self-developed PCS control system supports wide voltage range operation and can adapt to the power demand under different working conditions. The built-in intelligent monitoring system of the device can collect operating data in real time, with the shangtang energy model linkage, for the"Calculation with electricity, electricity with the movement" of the two-way closed-loop to provide data support.

From the customized scheme of Shangtang Lingang project, to the detailed polishing of C5 corrosion protection design, to the technical synergy of digital energy integration, inbev inbev think tank has"Hard core" technical strength, become a key player in AIDC Energy Storage. In the wave of rapid development of the industry, inbev smart storage is constantly exploring the boundaries of technology based on project practice, and building safe, efficient and reliable energy storage solutions for more AIDC, help"Green Computing Power" target landing, leading AIDC Energy storage new track.

Prospect: toward the ultimate form of"Computing power is energy storage"

In the long term, the certainty of AIDC's energy-storage track growth is not in doubt. Aijian securities forecasts that AIDC Energy Storage will become the third growth pole after household storage and large-scale storage, with a compound growth rate of over 80% between 2024 and 2030, and a global energy storage market share of 25% by 2030. As the industrialization of sodium electricity accelerates and VPP polymerization technology matures, the integration of AIDC and energy storage will be further deepened, and eventually the ultimate form of"Computing power is energy storage" will be realized -- that is, AIDC is no longer just a power consumer, it will become a"Virtual energy storage station" for the power grid through flexible dispatching, and a core hub connecting AI and the energy zero market, injecting new momentum into the global energy transformation and technological innovation.

About Us:

Beijing INPOWER Electric Co. , Ltd. was established in March 2004, is a high-tech Enterprises, the company is headquartered in Zhongguancun, Beijing high-tech industrial park, at 

present, there are three research and development centers, three production bases and 25 offices/sales outlets across the country, forming a nationwide marketing and service

 network covering the two major business sectors of power quality and energy storage systems, the energy storage system mainly includes 30 ~ 215 kw string energy storage

 converter, 500 ~ 2500 kw centralized energy storage converter, 1000 ~ 5000 kw boost integrated cabin, off-grid switching cabinet, outdoor cabinet energy storage system and 

Energy Storage Project EPC total package service. The company has acquired 260 + independent intellectual property rights (including acceptance) , 50 + invention patents 

(including acceptance) , and 100 + software copyrights (including acceptance) , with power electronics platform and new energy product development, system integration and 

power system application experience. The company has been in the energy storage business since 2014, it has won the honor of"The top 10 PCS of China in 2020", "The best PCS 

supplier of China Energy storage industry in 2022", "The top 10 PCS of China in 2022", "The top 10 PCS of China in 2023" and"The top 10 PCS of China in 2024". To learn more 

about Inbev, please visit inbev's official website: in-power.net