Modular data centers could shift AI infrastructure value toward integrated electrical OEMs

Overview

This industry study examines how modular, factory-integrated data-center infrastructure can relieve construction bottlenecks in AI build-outs and reallocate value from standalone components toward integrated electrical systems. Bernstein expects the structural shift to favor OEMs with broad Power, IT and cooling capabilities.

Core views

Bernstein’s central argument is that the AI data-center build-out is moving from a power-availability bottleneck to a construction bottleneck. US interconnection timelines increased from about 1.5 years in 2005 to more than five years in 2025, prompting greater use of behind-the-meter power: 22% of projects under construction and 40% of the pipeline, compared with 6% of active data centers. As engines and fuel cells reduce the generation constraint, the report argues that stick-built construction, EPC execution and scarce mechanical, electrical and plumbing labour become the binding constraints. Bernstein estimates an MEP labour ceiling of up to 35GW of US data-center additions by 2030, versus a GPU-manufacturing-implied ceiling of roughly 70GW including retrofits; 70% of projects are in regions with only 30% of the MEP labour pool. The shift to 800 VDC could initially raise fit-out MEP labour hours by about 50%. Modularisation addresses these constraints by shifting integration and testing from the site to factories while site preparation proceeds in parallel. The report cites Schneider’s reduction in lead times from 88 weeks for stick-built projects to 34 weeks, Vertiv’s indication of 10 weeks on site for modular deployments versus 70 weeks otherwise, and ABB’s estimate of a 50% deployment-time reduction using predesigned prefabricated modules. Schneider, ABB and Vertiv collectively expect roughly a 60% reduction in engineering, manufacturing and testing lead times under modular configurations. Bernstein also estimates that the field-to-factory shift could reduce labour costs by about 37%, although some gross on-site labour savings are offset by factory work. The addressable market expands as modular construction takes share. Bernstein screens a credibility-adjusted 2026-30 pipeline of approximately 126GW and forecasts modular additions to grow at a 24% CAGR, 10 percentage points faster than the 14% wider-market CAGR. Modular penetration rises from about 40% currently to just under 60% of additions by 2030, driven particularly by non-hyperscalers, which Bernstein expects to account for 60% of 2026-30 additions. Its model assumes modular adoption rises from 45% today to 65% by 2030 for non-hyperscalers, versus 35% to 45% for hyperscalers. The report notes that this forecast excludes projects announced in the next one to two years that may not be completed before 2030 and makes no further construction-delay adjustment beyond expected activation dates. The economic mechanism is procurement consolidation. Integrated module providers can specify, source and integrate 80-90% of module content internally, replacing component-by-component procurement with larger OEM-led systems. IT modules are especially important because Power-only prefabrication does not fully accelerate delivery unless IT and cooling are also integrated. In Bernstein’s survey of 50 respondents, 52% preferred a combined Power-and-IT-pod approach for future AI deployments, versus 16% each for Power-only pods, IT-only pods and stick-built construction. While current sourcing remains mixed—38% single-source and 26% multi-vendor—the report expects modularisation to increase willingness to single-source. Bernstein estimates that the largest integrated OEMs could capture around 3 percentage points of net market share. Faster deployment also creates value for customers and suppliers. Bernstein sizes barebones Power and IT modules at $1.9bn/GW and $1.8bn/GW, respectively. Using a 30-month conventional build, $150m/GW/month of revenue, a 50% FCF margin and a 10% WACC, it estimates that halving deployment time could generate about $0.9bn/GW of incremental present value. It assumes electrical OEMs capture roughly 25% of that value, or about $0.2bn/GW. The report expects larger and earlier orders, quicker backlog conversion and improved earnings visibility. Near-term EBITA per MW should benefit from content and pricing but may be diluted by capacity ramp-up and selling costs; longer term, standardisation, factory scale, and SG&A and R&D leverage should strengthen margins. On competitive positioning, Bernstein places Schneider, Vertiv and Eaton in an integrated camp with the clearest opportunity. Schneider has the broadest product coverage, proven customer wins, a hybrid strategy combining partnerships with a 270% expansion of prefabrication capacity in two years, and an established EcoStruxure platform. Vertiv is highlighted for the breadth and ambition of SmartRun and OneCore, including IT pods above 6MW and a modular architecture that can scale from 12.5MW deployments to 1GW campuses; however, greater project scope also creates supply-chain and execution risk, illustrated by Q2 2026 issues. Eaton has improved its position through acquisitions and partnerships, including Fibrebond and Flexnode, but Bernstein sees its offering as less mature than Schneider’s and Vertiv’s. Legrand, ABB and Siemens are viewed as relatively disadvantaged by narrower or less integrated portfolios. Legrand has white-space capabilities, growing grey-space exposure and assets from Girtz that could support a fuller offering, but currently relies more on best-in-breed procurement. ABB has credible Power-pod components but no IT-pod direction, limiting cross-sell. Siemens has addressed gaps through partnerships, including Delta, Cadolto and Legrand, but reliance on partners reduces the value capture available to more vertically integrated peers.

Analysis framework

Bernstein combines channel checks, proprietary surveys, a credibility-adjusted data-center pipeline model, customer-cohort adoption assumptions, product-content sizing and competitive mapping. It links deployment-time reductions to labour savings, customer revenue generation, module-level procurement, order conversion and OEM margin potential.

Methodology notes

• Industry AnalysisSupply-demand framework

  Data-center deployment bottleneck and modular-capacity sizing

  The report traces the constraint from grid access and generation to construction, labour and equipment lead times, then estimates how modular supply can relieve those constraints.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  Shift from component procurement to integrated Power, IT and cooling modules

  Bernstein explains how factory integration changes who controls specification and sourcing, allowing integrated OEMs to capture more content and market share.

• Valuation methodsDCF (Discounted Cash Flow)

  Present-value calculation for faster data-center deployment

  The report discounts the incremental cash flow from shortening a 30-month project timeline, using a 50% FCF margin and 10% WACC, to estimate value per GW.

• Corporate Fundamentals and FinanceOperating and Financial Leverage Analysis

  EBITA per MW and backlog-conversion economics

  The analysis separates near-term pricing and ramp-up costs from longer-term benefits of standardisation, factory scale and fixed-cost leverage.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Schneider

  Integrated modular OEM expected to be a leading beneficiary.

  Strengths

  Broadest product coverage, proven customer wins, established partnerships and prefabrication capacity expanded 270% in two years.

  Weaknesses

  Its IT pod is described as less powerful than Vertiv's.

  Comparison

  Alongside Vertiv, it is one of only two providers with a joint Power-and-IT offering including cooling.

• Vertiv (VRT)

  Integrated modular OEM with the most ambitious full-campus modular strategy.

  Strengths

  SmartRun IT pods above 6MW and OneCore architecture spanning 12.5MW deployments to 1GW campuses.

  Weaknesses

  Vertical integration does not eliminate reliance on external components, including transformers.

  Comparison

  Its IT-pod capability exceeds Schneider's c.1.2MW pod scale.

  Risks

  Broader project responsibility exposes the company to execution and external supply-chain risk.

• Eaton (ETN)

  Integrated OEM positioned to benefit, though less mature than the leading pair.

  Strengths

  Expanded its offering through the Fibrebond acquisition and Flexnode partnership.

  Weaknesses

  Its own-brand whitespace modular offering is described as underpowered and less mature.

  Comparison

  Bernstein places it behind Schneider and Vertiv in modular-offering maturity.

• Legrand

  Potential integrated-provider beneficiary if it combines acquired and existing capabilities.

  Strengths

  Growing grey-space exposure, white-space heritage and Girtz integration capabilities.

  Weaknesses

  Currently relies on best-in-breed procurement and may lose share before developing a fuller Power-plus-IT solution.

  Comparison

  Less integrated than Schneider, Vertiv and Eaton.

• ABB

  Power-module supplier with more limited modular upside.

  Strengths

  Has Power-pod components including compact secondary substations and eHouses.

  Weaknesses

  No IT-pod offering or clear IT-module direction, constraining cross-sell.

  Comparison

  Power coverage is competitive, but the portfolio is less complete than integrated peers.

• Siemens

  Modular participant whose partnership-led approach may limit value capture.

  Strengths

  Partnerships address UPS and edge-data-center product gaps; it also offers BIM, digital twins and fire-safety solutions.

  Weaknesses

  Dependence on partners for Power and IT modules restricts integrated-system upside.

  Comparison

  Considered a relative laggard versus more vertically integrated peers.

Key data

• Modular additions CAGR, 2026-3024%10 percentage points above the 14% wider-market CAGR.
• Modular share of data-center additionsc.40% currently to just under 60% by 2030Driven by higher expected adoption among non-hyperscalers.
• Credibility-adjusted 2026-30 additionsc.126GWBernstein pipeline sizing phased using consensus hyperscaler and neocloud capex.
• Potential integrated-OEM share gainc.3 percentage pointsEstimated effect of modularisation and greater single-source procurement.
• Incremental present value from halving deployment time$0.9bn/GWBernstein estimates OEMs could capture c.$0.2bn/GW.
• Power and IT module TAM$1.9bn/GW and $1.8bn/GWBarebones module-content sizing.
• Estimated modular labour-cost reductionc.37%Gross estimate before allowing for additional factory work.
• Preferred combined Power and IT pods52%Bernstein proprietary data-center survey, n=50.

Impact & implications

The report argues that faster time-to-power and time-to-first-token make modularisation a structural source of value creation for integrated electrical OEMs. The expected effects are earlier and larger orders, faster backlog conversion, higher internal content, improved pricing power and better long-run EBITA per MW, with Schneider, Vertiv and Eaton most exposed to those benefits.

Risks

• Vertiv’s broader project responsibility can create execution and external supply-chain risk, as highlighted by Q2 2026 issues.
• Near-term pricing and content benefits may be partly offset by capacity ramp-up and higher selling costs.
• The report’s customer survey evidence for neoclouds and colocation providers is based on small samples of five respondents each, limiting strong conclusions.
• Modular-market forecasts depend on pipeline-credibility adjustments, capex phasing and construction-timing assumptions. განსხვავ