2025 Data Center (IDC) Special Report: The AIDC Cycle Has Arrived, with All Players Competing for Market Share

Release Date:

2025-07-10

AI is advancing rapidly, ushering IDC into a new cycle of development.

IDC, as the backbone of the digital economy, is entering a new phase of rapid development.

Data centers serve as the computational infrastructure underpinning next-generation digital technologies such as AI and cloud computing. An IDC (data center) is a dedicated facility that provides tenants or customers with essential infrastructure—including server racks, storage systems, networking equipment, power supply, cooling, security, and monitoring—to host, operate, and deliver applications and services, as well as to store and manage the data associated with those applications and services. As the central hub for data and the primary platform for computational power driving the development of 5G, AI, cloud computing, and other cutting-edge digital technologies, data centers play a critical role in accelerating the growth of the digital economy.

The data center industry has entered a new phase of rapid growth. In recent years, driven by the rapid advancement of information technologies such as cloud computing, artificial intelligence, and big data, demand for data center computing power has surged. According to Huawei’s “Intelligent World 2030” report, by 2030 humanity will usher in the YB era of data—when the total volume of newly generated global data each year reaches the YB level—and global general-purpose computing power will reach 3.3 ZFLOPS (FP32). Meanwhile, demand for AI computing power is expected to skyrocket, reaching 864 ZFLOPS (FP16) by 2030. As a result, the global data center industry is now entering a new round of rapid development.

The rapid development of AI has become the primary driver of market expansion.

The advancement of AI is driving up demand for computing power, making it the primary driver of current IDC market expansion. First, scaling laws and emergent phenomena remain the foundation for enhancing large-model capabilities; improvements in AI model performance depend on larger training datasets and greater parameter counts, which in turn require higher levels of computational power. Second, large models are accelerating their evolution from text-based applications toward multimodal capabilities encompassing images, videos, audio, and more. In December 2024, OpenAI unveiled its video-generation model Sora, while Google released Gemini 2.0, both supporting multimodal inputs such as images, videos, and audio. As training data has evolved from solely textual content to a diverse mix of images, videos, and other media, the need for more powerful computing resources has grown accordingly. Third, new AI application scenarios are continuously emerging, including AI agents, VR/AR, and humanoid robots. Human–model interaction is no longer limited to text-based exchanges; users can now upload images or videos to ask content-related questions or interact with models via voice commands. As these models increasingly integrate into everyday work and life settings, the associated demand for computing power is rising. Fourth, the focus of scaling is shifting toward inference: OpenAI has launched inference models such as o1 and o3, allocating more computational resources to the inference stage to enable models to engage in deep reasoning and solve complex problems. At the same time, the rapid proliferation of newly released models and the substantial increase in the number of users are spurring fast growth in inference-side computing demand.

Intelligent computing centers have become the primary model for IDC development.

Based on application scenarios, IDCs can be categorized into general-purpose, intelligent computing, and supercomputing data centers: 1) General-purpose data centers leverage the computing power provided by CPU-based servers and are primarily used for traditional data storage, processing, and management tasks, with relatively balanced requirements for computational performance that include adequate compute, storage, and network transmission capabilities; 2) Intelligent computing data centers rely on accelerated computing platforms powered by AI chips such as GPUs, FPGAs, and ASICs, and are mainly employed in the fields of artificial intelligence and machine learning, where large-scale data training is used to develop models for delivering intelligent applications; 3) Supercomputing data centers utilize the computing power delivered by high-performance computing clusters, including supercomputers, and are primarily dedicated to cutting-edge scientific research areas such as planetary simulation, astrophysics, and genomic analysis.

With the advancement of AI, the demand for intelligent computing power will become the primary driver of future data center development. According to estimates by the China Academy of Information and Communications Technology, global intelligent computing power stood at 335 EFLOPS at the end of 2023 and is projected to reach 52.5 ZFLOPS by 2030. The rising demand for intelligent computing power will spur substantial growth in data center capacity; SemiAnalysis forecasts that global data center capacity will increase from 49 GW in 2023 to 96 GW in 2026, with newly built intelligent computing centers accounting for 85% of the incremental capacity.

IDC is progressively shifting toward higher density and greater energy efficiency.

Intelligent computing centers are driving the transformation of data centers toward higher density and greater energy efficiency. Prior to the widespread adoption of generative AI, data center operators typically designed and deployed facilities based on a standard rack power capacity of 2–10 kW per rack. However, with the rapid advancement of AI technologies and the corresponding surge in demand for computational power, data centers have begun deploying more AI servers equipped with GPUs to handle compute-intensive workloads, leading to a substantial increase in per-rack power density. According to the “Investment and Value Insights in the Data Center Industry” report, the global average rack power density has risen from 5.6 kW per rack in 2017 to 12.8 kW per rack in 2023; supercomputing and intelligent computing centers even exceed 30 kW per rack, and this trend is expected to continue, with per-rack power densities set to rise further in the future.

The AI era is underpinned by surging demand for computing power, which in turn relies on massive energy consumption. The expansion of AI computing capacity requires substantial electricity and generates significant carbon emissions—issues that are now unavoidable. A single ChatGPT search consumes roughly six to ten times more electricity than a conventional Google search. According to the International Energy Agency (IEA), data center electricity use could double by 2026, reaching between 650 and 1,050 terawatt-hours. In response to the rapid growth in global data center energy consumption, governments worldwide are promoting the transition of data centers toward greener, more sustainable operations and imposing limits on data center Power Usage Effectiveness (PUE).

Rapid Expansion of the IDC Market in China and Globally

China’s AI market is rapidly expanding.

The number and scale of large AI models in China are growing rapidly, with daily active users increasing at a swift pace. According to the China Academy of Information and Communications Technology’s Global Digital Economy White Paper (2024), there are currently 1,328 large AI models worldwide, with the United States accounting for 44% and China for 36%, placing China second globally. The global market for large AI models is projected to reach US$109.5 billion by 2028, with China’s share reaching RMB 117.9 billion. Meanwhile, data from the AI Product Rankings show that in December 2024, the Dabao app recorded 71.16 million monthly active users, making it the second-largest AI application globally after ChatGPT, and indicating that the gap between domestically developed large models and leading overseas counterparts is steadily narrowing.

Large-scale models are accelerating their penetration into industry-specific applications, empowering a wide range of use cases. From an industry perspective, the financial sector leads in AI adoption, with AI-plus-finance solutions now spanning product design, marketing, risk management, and customer service to deliver end-to-end, full-lifecycle solutions. In the healthcare sector, AI project development is diversifying, with large AI models driving innovation across multiple domains, including disease diagnosis, drug discovery support, and treatment optimization. In industrial manufacturing, AI is enabling greater automation in production management, thereby boosting efficiency and streamlining business processes. According to Sullivan’s forecast, driven by the widespread deployment of AI across industries, China’s market for industry-specific large models reached RMB 10.5 billion in 2023 and is projected to grow to RMB 62.4 billion by 2028, representing a compound annual growth rate of 42.82% over the period.

Intelligent Computing Centers Become AI New Infrastructure

Multiple policy measures have been introduced to continuously optimize the deployment of computing power. In recent years, China has attached great importance to the development of the computing-power industry, successively issuing a series of relevant policy documents, including the Implementation Plan for the Computing-Power Hub of the National Integrated Big-Data-Center Collaborative Innovation System, the Opinions on Deeply Implementing the “East Data, West Computing” Project and Accelerating the Construction of a Nationally Integrated Computing-Network, and the Action Plan for the High-Quality Development of Computing-Infrastructure. At present, China has established a data-center layout comprising “Eight Major Hubs and Ten Major Clusters,” which has significantly alleviated the imbalance in the spatial distribution of computing resources between the eastern and western regions and continues to optimize the overall arrangement of computing centers.

Against the backdrop of surging demand for computing power, scarcity of data and model resources, and the widespread deployment of AI technologies, intelligent computing centers have emerged as a new infrastructure pillar for regional AI-driven development. According to IDC data, local governments are rapidly advancing the construction of the intelligent computing industry and actively establishing large-scale computing nodes with capacities ranging from 300 to 1,000 PFLOPS (FP16) to provide efficient and stable computational services for large AI models. As reported by Kechi Consulting, by the end of 2024, the total number of operational or under-construction intelligent computing centers in China had exceeded 500, with the pace of such infrastructure development accelerating and robust market demand remaining strong.

China’s data center market is experiencing rapid growth.

Supported by national policies such as new infrastructure development, digital transformation, and the Digital China initiative, the data center market in mainland China has maintained rapid growth. According to data from the CIC Research Institute, the market size reached approximately RMB 363.6 billion in 2023, up 14.08% year on year, and is projected to reach RMB 543.7 billion by 2028, with a compound annual growth rate of 8.83% over the next five years. In 2023, the number of operational server racks in China totaled 8.42 million, representing a year-on-year increase of 29.54%, and is expected to rise to 20 million by 2028, reflecting a five-year CAGR of 17.48%. As digital transformation gains momentum across regions and industries and AI applications are progressively deployed, the data center market in mainland China is poised to sustain its growth trajectory.

Driven by the capital and scale advantages of leading players, China’s data center operators are increasingly consolidating around these top firms. According to IDC data, the combined market share of the top five data center service providers in China rose from 48.8% in 2022 to 51.3% in 2023. Leading companies are typically able to undertake larger-scale data center construction and technology investments, while also allocating greater funding and resources to R&D. At the same time, the data center market tends to exhibit high customer stickiness, and these incumbents, with their broader client base, enjoy a significant competitive edge in market expansion.

Improving Global Data Center Supply-Demand Balance

Overall demand in major global data center markets has grown markedly, driving pre-leasing rates to historic highs. Surging demand for intelligent computing centers—particularly hyperscale cloud services—has led to declining vacancy rates across multiple markets. According to Cushman & Wakefield data, by the end of 2023, vacancy rates in most mature markets had fallen below 10%, with core areas seeing rates under 5%. The share of pre-leased space has reached an all-time high, with 60% to 70% of leasing activity in key markets already secured before facilities are even completed.

Power has become a critical challenge for data center deployments, driving continued increases in data center leasing rates. Given constrained power supply and supporting infrastructure, most tenants opt to renew their existing leases rather than seek new facilities, resulting in high customer stickiness for data center operators. In U.S. dollar terms, Singapore leads the pack for leases targeting 250–500 kW, with rates ranging from $315 to $480 per kW per month. Meanwhile, in North America’s major data center leasing markets, average monthly service fees for 250–500 kW capacity have risen by 18.6% year over year, reaching $163.44 per kW per month. For data center developers, power supply remains a central consideration. Progress in renewable energy, the development of transmission and distribution infrastructure, and the availability of affordable power continue to shape data center operators’ site-selection decisions.

The global data center market is expanding rapidly.

The global pipeline of data centers—comprising both projects under construction and those in the planning stages—is expanding rapidly, ensuring that the data center sector will continue to experience robust growth. The size of this pipeline serves as a key indicator of future market supply, with the capacity slated for delivery over the next three to five years expected to translate into actual available capacity. According to DCbyte data, as of 2023, the total planned capacity of data centers worldwide reached 37.8 GW—double the current installed base—signaling that the data center industry will maintain rapid expansion over the next five years.

Driven by AI and high-end computing, the global data center market is expanding rapidly. According to Astute Analytica, by 2030 AI workloads are expected to account for one-third of total global data center electricity consumption, and AI-powered data centers will make up 70% of the overall data center footprint. Meanwhile, data from Fortune Business Insight indicate that the global data center market reached US$219.23 billion in 2023 and is projected to grow from US$242.72 billion in 2024 to US$584.86 billion by 2032, representing a compound annual growth rate (CAGR) of 11.6%. Among the segments, colocation services constitute the largest share, accounting for 35% of the market.

IDC Key Company Review

Digital Realty Trust: Rising Downstream Demand Drives Price Increases

Digital Realty Trust is a global leader in data center, colocation, and interconnection solutions. The company’s data centers provide secure, highly connected, and continuously available environments for the exchange, processing, and storage of critical data, enabling customers to connect with their own customers and partners through the PlatformDIGITAL platform. As of Q3 2024, Digital Realty serves more than 5,000 customers across over 50 major markets worldwide, operating more than 300 data centers with a total operational capacity exceeding 2,700 megawatts.

Driven by strong downstream demand, renewal rates have risen sharply. In 24Q3, the company secured $521 million in new lease commitments, with $450 million attributable to projects exceeding 1 MW—primarily due to a rapid increase in large-scale, multi-megawatt bookings in North America and a nearly 30% rise in pricing. To meet this robust demand, the company has undertaken significant data-center expansion, with 644 MW of capacity currently under construction, while maintaining over 3,000 MW of buildable IT load capacity. Total customer pre-orders now stand at $859 million, up more than 60% quarter-over-quarter, with projected recurring annual revenue from these orders reaching $450 million in 2025. In 24Q3, the company also executed $258 million in renewal leases, with cash renewal rental rates increasing by 15.2%; over the first three quarters of 2024, the overall cash renewal spread expanded by 10.5%.

Equinix: Global Leader in the Data Center Market

Equinix is the global leader in the data center market, committed to providing enterprises with equitable access to data centers. As a neutral, multi-tenant data center (MTDC) provider, Equinix was founded at a time when most data centers were either built or leased by telecom operators and internet companies themselves, facing challenges such as high construction and maintenance costs and limited flexibility to scale in response to business growth. To address these pain points, Equinix has focused on delivering large-scale, third-party internet data centers that offer enterprises equal access, neutrality, and interconnectivity. In 2015, the company applied to restructure as a REIT (Real Estate Investment Trust), leveraging the tax advantages of the REIT structure to enhance profitability and further expand its global footprint. According to IDC, Equinix was the world’s largest data center service provider in 2023. Additionally, according to Beacon Consulting, the company held a 10.69% global market share in 2024.

What is a Power Distribution Unit, and what is the development of intelligent PDUs?

How to Choose a Smart PDU? What Factors Should You Consider?