Johnson Controls与Armada扩大合作：拟2027年建成500MW分布式AI算力，采用英伟达Vera Rubin/DSX POD预制数据中心，但无收入、订单与融资细节

Johnson Controls ( JCI ) and Armada, a developer of modular AI computing infrastructure, plan to build 500 megawatts of distributed artificial intelligence computing capacity in 2027. The plan expands their existing partnership to include factory-built data centers based on Nvidia's ( NVDA ) next-generation Vera Rubin computing platform. The companies announced Wednesday at Nvidia's GTC conference in Berlin that they will collaborate on engineering, manufacturing, integrating and deploying prefabricated AI facilities using Nvidia's new DSX POD reference design. The approach allows computing equipment, electrical systems, networking and liquid-cooling infrastructure to be assembled and tested before shipment to customers. For Johnson Controls ( JCI ) investors, the planned 500MW deployment represents a potential opportunity to expand the company's role in AI data center construction beyond conventional building systems and cooling equipment. The initiative could create additional demand for its thermal management technology, engineering services and manufacturing capabilities as AI infrastructure developers seek alternatives to lengthy construction projects. However, the companies didn't disclose the project's expected revenue, capital requirements, financing arrangements or binding customer commitments. The latest announcement adds a specific capacity target and deployment timetable to the companies' existing relationship. It also establishes plans to incorporate Nvidia's Vera Rubin architecture and DSX POD design into their modular data center manufacturing program, potentially expanding the scope of equipment and services required from Johnson Controls ( JCI ). Under the proposed manufacturing model, computing racks, power distribution systems, networking equipment and liquid-cooling components would be installed and validated at factories before delivery. The prefabricated units would then be connected to power, water, cooling and networking infrastructure at their destinations. The companies aim to reduce on-site construction requirements and allow customers to expand computing capacity by adding standardized modules. Armada also plans to use NVIDIA's DSX Air software to simulate facilities and test its infrastructure management software before equipment ships. Once deployed, NVIDIA's DSX MaxLPS technology is intended to manage power consumption and computing workloads to improve processing output within each location's available electrical capacity. Johnson Controls Chief Executive Officer Joakim Weidemanis said in a statement that the collaboration combines Armada's experience deploying computing infrastructure in remote locations with Johnson Controls' thermal management technology, engineering resources and manufacturing operations. The companies intend to serve customers seeking computing capacity near available energy sources rather than relying exclusively on large, centralized data centers. The proposed network would include modular facilities located near renewable energy sources and other available power supplies. Armada said the infrastructure would allow customers to retain control of their computing systems, AI models and data without depending on external cloud platforms. The companies didn't specify how much of the planned 500MW capacity is backed by customer orders, where the facilities would be located or when individual deployments would begin. Those details, along with the pace of manufacturing and installation, will help determine how quickly the initiative translates into revenue for Johnson Controls ( JCI ).