AI computing power construction shifts from "power shortage" to "talent shortage": modularization rises, Schneider, Vertiv, and Eaton welcome new opportunities.

Bernstein recently released a research report stating that against the backdrop of rapid expansion in AI computing power construction, the primary constraint on data center development is shifting. Grid interconnection challenges are gradually being alleviated through behind-the-meter self-generation, but on-site construction and mechanical/electrical labor shortages have become the new bottleneck. The report notes that modular construction will break through computing power construction bottlenecks, reshaping the market landscape and profit distribution logic of the electrical equipment sector. Construction Constraints Evolve: Labor Shortages Force Modular Transition Over the past several years, the biggest pain point for U.S. data centers has been lengthy grid interconnection timelines, with wait times stretching beyond 5 years. The industry has widely adopted behind-the-meter (BTM) self-generation sources such as fuel cells and reciprocating engines to bypass grid constraints, with BTM power now accounting for 40% of projects in the current planning pipeline. With the power supply bottleneck easing, on-site civil construction and MEP (mechanical, electrical, and plumbing) labor have become the new hard constraint. 70% of data center projects are located in areas with scarce mechanical and electrical workers, and the 800VDC high-voltage architecture further increases construction difficulty, raising on-site installation labor hours by 50%. The U.S. data center construction ceiling constrained by labor is only 35GW by 2030, far below the 70GW potential demand corresponding to GPU computing power. Modularization migrates a large portion of construction work to factory prefabrication, with only assembly completed on-site, which can shorten deployment cycles by 30-60%, reduce comprehensive labor costs by approximately 37%, and shift labor demand from rural construction sites to industrial manufacturing centers, breaking through the labor supply ceiling. Procurement Model Transformation: Integrated Modules Bring Share Gains Modularization is not just a change in engineering processesit is a disruption of procurement logic. Under the traditional model, operators or EPC general contractors procured components separately; in the modular era, customers increasingly prefer to purchase Power+IT integrated prefabricated skids, with OEMs controlling product specifications, system integration, and supply chains, able to supply 80-90% of components internally. Surveys show that 52% of customers want to procure power and IT modules together for future AI projects, and modularization will significantly increase customers' willingness for single-source procurement. Calculations show that rising modular penetration can bring nearly 3 percentage points of net market share gains for leading manufacturers with vertical integration capabilities. In terms of market size, the power module market is $1.9 billion/GW, while the IT module market reaches $1.8 billion/GW; halving deployment cycles can generate $900 million in incremental present value per GW, of which equipment vendors can capture 25%. In the short term, larger order volumes and faster delivery turnover improve corporate order visibility, but capacity ramp-up and increased selling expenses will pressure profits; in the long term, relying on standardization and factory scale effects, EBITA profit per megawatt will continue to rise. Sector Divergence: IT Module Capabilities Determine Competitive Positioning Power module technology is relatively mature, and most electrical equipment manufacturers can supply it; IT computing modules represent scarce capability and serve as the core benchmark for tiering competition. Schneider, Vertiv (VRT.US), and Eaton Corp. Plc (ETN.US) form the first tier, possessing complete product capabilities for both power skids and IT skids. Vertiv has a high product ceiling, with its OneCore architecture capable of supporting campus construction up to 1GW; Schneider has extensive project deployment experience and rapidly expanding prefabrication capacity; Eaton Corp. Plc relies on M&A and external partnerships to fill its IT module gap, with slightly weaker product maturity. Legrand, ABB, and Siemens are in the second tierwhile they have power modules, they lack mature self-developed IT modules and can only supply discrete components. Unless they fill their end-to-end integration capabilities, they will be passive in the wave of integrated module procurement. Non-hyperscale cloud providers and colocation cloud enterprises are the primary DRIVE of modularization, and this group will contribute 60% of new data center additions from 2026-2030, also becoming the core customers of leading integrated OEMs. Overall, modularization is an inevitable choice for AI computing power construction, but there are also potential risks: the turnkey model amplifies OEM supply chain and project execution risks, and if industry overcapacity occurs, price wars will erode industry profitability.