The "AIDC Infrastructure Specification (V1.0)", released by the Open AI Infra community of the Global Computing Consortium (GCC), is an operational technical document for the design and construction of Artificial Intelligence Data Center (AIDC) infrastructure. Its core objective is to provide a detailed set of technical requirements and implementation guidelines—from the campus to the computer room, and from construction to operation and maintenance—to support AI liquid-cooled full racks with per-rack power evolving to approximately 300kW over the next 3-5 years. The specification emphasizes "decoupling the computer room from the rack" to ensure compatibility with 2-3 generations of AI equipment evolution.
1. General Requirements and Key Parameters
The specification begins by clarifying the core challenges and response strategies for AIDC construction, with its core philosophy being flexible compatibility and rapid deployment.
Per-rack power rating: Rack power consumption is explicitly divided into three levels: 100kW, 200kW, and 300kW, with corresponding power supply interface formats provided (e.g., 380Vac 63A industrial connectors).
System-level synergy: Requires that building structures, power supply, cooling, and network cabling be planned in a unified manner in conjunction with the generation of computing equipment. For example, the clear building height should not be less than 6.0 meters, and the live load of the main computer room should not be less than 20kN/m².
Rapid deployment: Advocates for the standardization, modularization, and prefabrication of mechanical and electrical systems, and emphasizes conducting collision, hydraulic, thermal field, and energy efficiency simulations through 3D digital modeling, as well as building intelligent operation systems.
2. Building and Environmental Requirements
This section makes rigid provisions for the physical space of AIDCs to accommodate the deployment of ultra-heavy, ultra-tall, and high-density equipment.
Structural design: The seismic fortification category for newly built AIDCs is the key fortification category (Category B), with a fire resistance rating of no less than Grade I. There are clear lower limits for the clear height, load (not less than 20kN/m²), and waterproofing grade (Grade I) of the main computer room.
Logistics channels: To facilitate the transportation of full racks, detailed specifications are provided for the dimensions of freight channels (width ≥ 2.4m, height ≥ 2.8m), elevator load capacity (preferably ≥ 5000kg), and cabin size.
Environmental control: Specific indicators are clarified for computer room temperature and humidity (cold aisle 18-27℃), air cleanliness (ISO 14644-1 Class 8), and gas corrosion grade (ANSI/ISA G1).
3. Power Supply and Distribution System
The power supply system design revolves around high reliability, flexible expansion, and high energy efficiency to accommodate significant load fluctuations.
System architecture: The logical structure of the uninterruptible power supply system for IT and critical infrastructure should adopt a 2N architecture to ensure redundancy. End-point power distribution can use row cabinets or busway distribution, with detailed specifications for current carrying capacity, cooling clearance, and expansion space.
Power quality: Specific regulations are made for the input/output voltage, frequency, and harmonic distortion rate of the UPS (e.g., input THDi < 8%, output voltage regulation accuracy ≤ ±1%).
Backup batteries: Clarifies the backup power time for IT batteries (single-side full load ≥ 7 minutes, 2N full load ≥ 15 minutes), and details the deployment safety requirements for lithium-ion batteries (lithium iron phosphate should be used), such as fireproof compartments, capacity limits, and fire extinguishing systems.
Campus power supply: Encourages integrated planning of "computing-power and electricity synergy", prioritizes the use of clean energy, and simultaneously plans intelligent management and control platforms.
4. Cooling System
The cooling system is the focus of this specification, with the core being air-liquid flexible design to accommodate different proportions of liquid cooling and air cooling heat dissipation requirements.
Liquid cooling system:
Primary side cold source: Various forms can be selected, and an air-liquid common source design should be considered to simplify the system.
Secondary side liquid cooling: This is the core of the specification, requiring pipeline design to be compatible with a ≥90% liquid cooling proportion and supporting flexible expansion (CDU quantity configured as N+1+X). Very detailed technical parameters are proposed for the key specifications of CDU equipment (e.g., cooling capacity ≥ 450kW, temperature control accuracy ≤ ±1℃, liquid supply pressure ≤ 3bar), secondary pipelines (e.g., using 304 stainless steel, flow resistance requirements), and cooling working fluids (material compatibility, composition limits).
Air cooling system: Clarifies the upper limit of air cooling heat dissipation power for the air-cooled part (20kW~40kW), and puts forward requirements for the cold source system (e.g., chilled water supply temperature should be ≥ 18℃) and air cooling terminals (redundancy configuration rules).
5. Cabling, Delivery, and Operation & Maintenance
Cabling: Emphasizes the rational planning of cable tray routes, preferably using overhead cabling, and specifies construction details such as the bending radius, spacing, and lashing of optical fibers and copper cables to ensure the quality of high-speed network transmission.
Delivery and implementation: Requires factory prefabrication of liquid cooling pipelines and prohibits on-site welding to ensure cleanliness and quality. Detailed specifications are provided for the cleanliness of the delivery environment (reaching ISO Class 8), pipeline flushing standards (conductivity ≤ 10μs/cm), and the four stages of testing and acceptance.
Operation and maintenance: Proposes general O&M service requirements (monitoring, emergency response, spare parts management), and provides specific O&M requirements for the liquid cooling system, including regular testing of the coolant, CDU inspection, and preventive maintenance.
Operation optimization: Advocates deploying intelligent operation management systems. Through 3D digital modeling and AI energy efficiency tuning algorithms, it aims to achieve an infrastructure simulation accuracy of ≥95% and reach a target of global energy efficiency improvement of ≥12%.
In summary, this V1.0 specification is a detailed "construction guide" for AIDC infrastructure construction. It transforms the concept of "system-level synergy" from previous framework reports into specific engineering parameters, equipment selection standards, and construction process requirements. Its core value lies in providing the industry with a measurable, implementable, and multi-generation technology evolution-compatible construction benchmark, offering a practical and feasible technical basis for the actual construction of gigawatt-level intelligent computing centers.
Disclaimer: This account focuses on sharing relevant technologies. The content and opinions do not represent the stance of this account. Traceable content is sourced accordingly. If there are any copyright issues with the published articles, please leave a message to contact us for deletion. Thank you.