Taiwan III launches national team to deliver DCBBS computer cabinet
Generative AI is fueling accelerated growth in worldwide computing power requirements, yet the industry encounters significant challenges. Although the product cycle for GPUs and AI servers has been condensed to approximately 18 months, the conventional data center construction process remains at two to three years, resulting in a structural gap where the deployment of computing capacity lags behind technology advancements.Supermicro, in collaboration with Easy Field Corporation (EFC) and Chicony Power, has unveiled the world’s first containerized compute solution built on Data Center Building Block Solutions (DCBBS). This solution is designed to expedite engineering delivery timelines and offers an alternative approach to the construction of AI infrastructure. Originally developed by Supermicro, DCBBS is a comprehensive, modular data center architecture aimed at accelerating large-scale compute deployments.In this DCBBS solution, Taiwan's Institute for Information Industry (III) acts as a coordinator for industry integration, connecting resources from IT, infrastructure, engineering consulting, and manufacturing to ensure that all stakeholders comply with a unified delivery timeline.Wang Rui-min (transliterated from Chinese), Senior Strategy Director of the Market Intelligence & Consulting Institute (MIC) at III, explained that AI computing power initiatives involve various components and parts, such as servers, networking devices, power supplies, cooling infrastructure, structural engineering, testing procedures, and international certifications. Historically, inadequate coordination methods often led to asynchronous delivery within these categories.Simon Hsu, COO of Taiwan's National Applied Research Laboratories (NARLabs), observed that traditional data center construction methodologies are inadequate in accommodating the progress of AI technology. Computing power has become a crucial resource for scientific research and industrial development, with the speed of deployment significantly impacting the evolution of both research and applications. In this context, modular and rapidly replicable computing power supply models are garnering attention.The DCBBS solution, collaboratively developed by Supermicro and several Taiwanese manufacturers, utilizes a methodical design approach. It first develops the overall data center architecture, then deconstructs tasks such as power supply, cooling, networking, fire safety, and administration into modular components.Huang Shih-chin (transliterated from Chinese), Product Manager for Supermicro's Rack Solutions, stated that this solution explicitly demonstrates computing power density in its technical specifications, addressing the requirements of various GPU generations, and introduces the notion of "Container Scale," enabling the expansion of computing power on a container-by-container basis. The engineering design prioritizes uninterrupted water, electricity, and network connectivity.The generator, transformer, chiller, and UPS are arranged in a containerized configuration, equipped with wireless environmental control and remote monitoring systems, facilitating mobility and rapid deployment of the computing units. The container is designed according to weight, stress absorption, external circumstances, and transportation certification standards, rather than conventional container specifications. The test line construction and burn-in verification have been completed, and the DCBBS system has entered the final acceptance and shipment phase.In the DCBBS project, EFC used an engineering methodology to complete system integration tasks for computing resource deployments. Chairman Paul Luo stated that, while AI servers are expensive, the primary elements determining computing power competitiveness are deployment timeframes and engineering dependability; delays or system failures result in expenditures that much surpass hardware expenses. Thus, EFC used a modular approach to organize the structural design, vibration mitigation, transportation, and certification requirements for the outdoor installation of large, high-density equipment in atypical environments, before transferring these components to the factory for assembly and verification. Luo indicated that, moving forward, computing containers will primarily be deployed in areas rich in energy resources.Consequently, what is truly finite is not electricity itself, but the ability to swiftly replicate and maintain the level of engineering execution. EFC, possessing extensive expertise in the production of large-scale structures and high-reliability systems, ensures that its computer cabinets are thoroughly prepared for acceptance and delivery before shipment. They are considered the fundamental framework and structure of the entire computing center, enabling future expansion within a managed and foreseeable technical timeline.In terms of work division, Supermicro is in charge of coordinating the full rapid delivery process, drawing on its significant experience with high-density computing clients. In recent years, the company has obtained AI processing power contracts, including those from Tesla, gaining engineering experience in high-density configurations and fast delivery, which has become a critical basis for minimizing deployment delays. Chicony Power oversees the wireless environmental control system, employing sensors, digital twins, and remote-control technology to mitigate the effects of conventional extensive cabling on deployment efficiency, hence facilitating expedited scheduling and replication of computing cabinets.EFC monitors the structural engineering design and, using its expertise in ship door manufacture, has ensured the structural integrity, vibration mitigation, and marine certification compliance for large, high-density equipment employed in outdoor settings. On the power infrastructure side, a modular Power Interface Unit (PIU) developed under the Taiwan Busway (TBC) project enables flexible and scalable power configuration. Sinotech Engineering Consultants also contributed to the end-to-end planning of JBL’s large-scale computing center in Indonesia. Overall, the DCBBS-based initiative has now entered the active implementation phase.Hsieh Chieh-shou (transliterated from Chinese), a director on Viscor Computing's board, stated that the DCBBS project is not intended as a conceptual demonstration but as an engineering solution based on the principles of deliverability and replicability.Unlike other teams that are still engaged in architectural design and strategic planning, the DCBBS project, led by Supermicro, EFC, and Chicony Power, has completed the setup of operational test lines, which include physical equipment integration, power configuration, and burn-in verification, and is ready to advance to the replication and delivery phase. As this model evolves, the team has been emulating the collaborative strategy of "national second-tier teams," thereby broadening the participation of Taiwanese firms possessing engineering and manufacturing expertise. This enables the metal and electromechanical sectors, which have faced challenges due to shifts in the global industrial landscape in recent years, to venture into computing infrastructure and pursue an additional transformation opportunity.Wang noted that the delivery approach of the DCBBS project offers significant advantages for implementing enterprises in both financial and engineering aspects. In engineering, factory modularization and pre-integration—including wiring, labeling, performance verification, and burn-in testing—conducted prior to shipment can substantially diminish the uncertainty and risk associated with repeated modifications during onsite construction abroad. Financially, acceptance and payment can come quickly after testing, shortening the project payback period and reducing foreign deployment and travel costs. He stated that this design allows computing power efforts to evolve into a business model with predictable delivery and cash flow patterns, rather than long-term investment ventures with low returns.Building on this foundation, the DCBBS project has evolved from a single compute cabinet into a sustainably scalable, modular system architecture. The team will subsequently release modules pertaining to generators, transformers, UPS, chiller units, control rooms, and coatings, while including a new 800V DC power architecture and cooling-in-rack design to improve the overall system's integrity and adaptability. Essential technical tasks, such wiring, labeling, integration, and testing, are consistently performed in Taiwan, thereby mitigating the unpredictability of international construction and expediting acceptance and payment processes. Wang emphasized that the core value of the DCBBS initiative lies not only in improving engineering and financial efficiency, but also in retaining the majority of high value-added integration and verification work in Taiwan, laying a sustainable foundation for long-term industrial competitiveness in the AI era.Taiwan-built DCBBS delivers deployable, scalable compute architecture. Credit: Easy Field Corporation
