AI Computing Power Construction Shifts from "Lack of Electricity" to "Lack of Talent": Modularization Rises, Schneider, Vertiv, Eaton Embrace New Opportunities

Zhitong Finance APP has learned that Bernstein recently released a research report indicating that, amid the rapid expansion of AI computing power infrastructure, the main bottleneck restricting data center development is undergoing a shift. The challenge of grid connection is gradually being alleviated by behind-the-meter (BTM) self-supplied power, but on-site construction and shortages in mechanical, electrical, and plumbing (MEP) labor have become new obstacles to development. The report points out that modular construction methods will break the bottleneck of computing power infrastructure, reshaping the market landscape and profit distribution logic of the electrical equipment sector.

Evolution of Construction Constraints: Labor Shortage Drives Modular Transformation

In recent years, the biggest pain point for US data centers has been the lengthy grid connection cycles, with waiting times extending beyond five years. The industry generally adopts fuel cells, reciprocating engines, and other behind-the-meter (BTM) self-supplied power sources to bypass grid restrictions, and BTM sources now account for 40% of projects in the current pipeline.

After the power supply bottleneck is mitigated, on-site civil construction and MEP labor have become new hard constraints. 70% of data center projects are located in regions with scarce MEP workers, while the adoption of 800VDC high-voltage architecture further increases construction difficulty, raising on-site installation man-hours by 50%. By 2030, the US construction cap for data centers due to labor limitations will be only 35GW, far below the potential 70GW GPU computing demand.

Modularization shifts many construction processes to factory prefabrication, with only assembly required onsite, reducing deployment cycles by 30‑60%. This cuts total labor costs by about 37%, shifting labor needs from rural sites to industrial manufacturing centers and breaking through the ceiling of labor supply.

Procurement Model Reform: Integrated Modules Bring Market Share Dividends

Modularization is not just a change in engineering processes; it also revolutionizes procurement logic. Under the traditional model, operators or EPC contractors separately source components; in the modular era, customers prefer to purchase integrated Power+IT prefabricated modules, with OEMs mastering product specifications, system integration, and supply chain, able to supply 80‑90% of components internally.

Research shows that 52% of customers hope to purchase both power and IT modules for future AI projects, and modularization will significantly increase their willingness to single-source purchases. Calculations show that increased penetration of modularization can bring nearly a 3-percentage-point net market share increase for top vertically integrated vendors. Market-wise, the power module market size is $1.9 billion/GW, while IT modules reach $1.8 billion/GW; halving deployment cycles can create additional present value of $900 million per GW, with equipment vendors able to capture 25% of those gains.

In the short term, larger order volumes and faster delivery turnover will enhance order visibility for companies, but ramp-up of capacity and higher sales expenses may suppress profits. In the long term, relying on standardization and factory scale effects, EBITA profit per megawatt will continue to rise.

Track Differentiation: IT Module Capabilities Determine Corporate Competitive Position

Power module technology is relatively mature, with most electrical equipment manufacturers able to supply; IT compute modules, however, are a scarce capability and a key benchmark for defining competitive tiers.

Schneider, Vertiv (VRT.US), and Eaton (ETN.US) form the first tier, each with a complete product portfolio for both power and IT modules. Vertiv leads in product capability, with its OneCore architecture supporting up to 1GW campus builds; Schneider brings extensive project deployment experience and rapid prefabrication capacity expansion; Eaton supplements its IT modules through mergers and external partnerships, with slightly less mature products.

Legrand, ABB, and Siemens are in the second tier. Despite having power modules, they lack mature, self-developed IT modules and can only provide discrete components. Unless they build end-to-end integration capabilities, they will be disadvantaged during the surge toward integrated module purchasing. Non-hyperscale cloud providers and colocation cloud companies are the main drivers of modularization, expected to account for 60% of data center growth from 2026 to 2030, thus becoming core customers of leading integrated OEMs.

Overall, modularization is the inevitable choice for building AI computing power, but also entails potential risks: turnkey models amplify OEM supply chain and project execution risks, and if industry capacity is oversupplied, price wars may erode profitability.