Bernstein believes North America's data center pipeline needs a substantial discount, with only about one-third more likely to actually be built

Overview

This is an industry research report on the credibility of U.S./North American communications infrastructure and data center development pipelines. Bernstein points out that there are many data center development announcements in the market, but announced capacity is not equal to deliverable capacity. Its Data Center Pipeline Tracker shows a nominal pipeline close to 500GW, covering existing facilities, announcements, land banks, canceled projects, and projects at different development stages; after excluding operating and known canceled/withdrawn projects, the report assigns completion probabilities project by project to the remaining 5,324 projects and estimates that about 135GW of North American capacity is more likely to come online over the next 5-10 years.

Core views

The core judgment is that data center demand remains strong, but industry forecasts cannot directly count most announced capacity as supply. Truly credible capacity is concentrated in projects with actual construction progress, pre-leased to investment-grade hyperscale tenants, where developers have delivery records and regulatory risk is lower. The report continues to favor DLR and EQIX for their ability to deliver construction and believes many supply forecasts about intensifying competition in core metro areas are exaggerated; at the same time, data center capacity tracking continues to support medium- to long-term demand for electrification equipment and engineering services.

Analysis framework

The report uses a bottom-up, project-level probability model. First, it uses the Aterio-supported Data Center Pipeline Tracker as the underlying dataset, covering 7,544 projects in North America; after excluding projects already operational, canceled, and withdrawn, it probability-weights land bank, announced, and under-construction projects. Probability is determined jointly by base project-stage probability, a tenant multiplier, developer quality score, and a state regulatory multiplier, with particular emphasis on distinguishing projects that have already broken ground from those that remain only in press releases or financing presentations.

Methodology notes

• Project-level probability scoringData Center Pipeline Probability Score

  Assign a completion probability to each data center project based on project stage, tenant quality, developer credibility, and state-level regulatory risk.

  Projects under construction receive higher base probabilities because construction progress already exists; announced projects and land bank projects are discounted more heavily. Investment-grade hyperscale tenants, reliable developers, clear power pathways, and capital support raise probability; states with higher regulatory resistance reduce the probability of projects that have not yet broken ground.

• Developer quality scoringCompany Credibility Score

  Developer credibility is weighted based on power pathway, capital, offtake/tenants, and execution track record.

  The weights are Power Path 35%, Capital 30%, Offtake 25%, and Sponsor Execution 10%. Amazon AWS, Google, Facebook, Microsoft, Digital Realty, Equinix, and QTS Data Centers receive top scores of 5.00; announced capacity from lower-scoring developers such as O'Leary Ventures and Vermaland is heavily discounted.

• Regulatory risk discountState-Level Haircut

  Apply state-level multipliers for anti-data-center legislation or policy risk in certain states.

  Virginia and New York have multipliers of 70%; Texas, Ohio, Arizona, Illinois, and Oregon are 85%; Indiana is 90%; Georgia and Washington are 95%; other states are assumed to have no state-level risk by default. Projects already under construction are assumed to benefit from grandfathering, while announced or land bank projects are more affected.

Asset mapping & comparison

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

• DLR (Digital Realty)

  Core beneficiary/covered company

  Strengths

  Long development history, credible construction capability, Bernstein maintains Outperform, target price $232.

  Weaknesses

  The report does not provide a standalone weakness in the excerpt, but the industry as a whole faces supply, regulatory, and execution risks.

  Comparison

  Compared with long-tail developers that have only announcements, DLR is more credible because of its historical delivery record.

  Risks

  Competitive supply forecasts, construction costs, power availability, regulatory changes, and project delivery pacing.

• EQIX (Equinix)

  Core beneficiary/covered company

  Strengths

  Long development history and high-credibility developer score, Bernstein maintains Outperform, target price $1,222.

  Weaknesses

  The report does not provide a standalone weakness in the excerpt.

  Comparison

  Compared with other announcement-driven developers, EQIX is classified as a blue-chip colo, and the model assigns it high credibility.

  Risks

  Data center construction cycles, regional regulation, power constraints, and industry valuation volatility.

• BXDC (Blackstone Digital Infra)

  Data center theme beneficiary

  Strengths

  Bernstein maintains Outperform, target price $26; the Blackstone-related platform background strengthens market confidence in pre-leasing and capital capability.

  Weaknesses

  Still in the early profit ramp-up stage in financial forecasts.

  Comparison

  Compared with long-tail developers lacking capital and execution records, institutional capital backing is an advantage.

  Risks

  Early project execution, customer offtake, capital expenditures, and post-listing valuation volatility.

• ETN, HUBB, LGN, J

  Electrification and engineering services beneficiaries

  Strengths

  Data center capacity growth supports equipment and labor demand, bottlenecks remain, and order backlogs are healthy; all remain Outperform.

  Weaknesses

  Compared with data center developers, the benefit chain is more indirect and is affected by equipment cycles, costs, and project pacing.

  Comparison

  Relative to CAT, CMI, and PWR, the report assigns more positive ratings to these names.

  Risks

  Capex delays, supply chain easing, margin pressure, and cancellation of end projects.

• CAT, CMI, PWR

  Covered companies related to data centers/electrification

  Strengths

  Supported by multi-year end-market growth.

  Weaknesses

  Bernstein maintains Market-Perform, with less positive conviction than ETN, HUBB, LGN, and J.

  Comparison

  They are also part of the data-center-related industry chain, but their ratings are below the Outperform group.

  Risks

  Valuation, demand realization, construction cycles, and macro capex volatility.

• QTS Data Centers, Vantage Data Centers, STACK Infrastructure

  Private developers/contributors to credible capacity

  Strengths

  QTS has 7.6GW of credible capacity, Vantage 5.4GW, and STACK 4.4GW; all have strong execution and/or institutional backing.

  Weaknesses

  Some tenant disclosure is insufficient, and STACK's project stages are relatively earlier.

  Comparison

  Clearly superior to lower-credibility announcement-driven developers such as O'Leary Ventures, Joule Capital Partners, and Vermaland.

  Risks

  Incomplete pre-lease disclosure, construction execution, power, and regulatory approvals.

Key data

• North American nominal pipeline capacity492GWThe current capacity scale in the report's dataset includes many announced or early-stage projects that will not be realized.
• Credible construction capacity135GWAbout 33% of the nominal pipeline, viewed as a more reasonable realized scale over the next 5-10 years.
• Number of projects7,544 projects; probability-weighted sample of 5,324 projectsThe underlying tracker covers 7,544 projects; after excluding operating, canceled, and withdrawn projects, 5,324 projects are scored.
• Target deliveries implied by raw data before 2028about 88GWThe report considers this timeline highly unrealistic and therefore does not make a strong judgment on the specific realization timing.
• Hyperscaler self-build shareabout 43% of credible construction capacityHyperscaler self-build is an important component of the credible pipeline.
• Top 10 developer concentrationabout 60% of credible capacityCredible capacity is highly concentrated among leading developers.
• Contribution from the four largest cloud providersAmazon, Google, Microsoft, and Meta together account for about 42% of credible MW, about 78GWHyperscalers are the core source of credible capacity.
• QTS Data Centers7.6GW credible capacity, 10.6GW nominal capacity, 72% survival rateThe report describes it as one of the largest credible pipelines among non-hyperscale cloud developers, with a credit score of 5.0.
• Vantage Data Centers5.4GW credible capacity, 6.6GW nominal capacity, 82% survival rate3.9GW is under construction and supported by Oracle + OpenAI tenancy.
• STACK Infrastructure4.4GW credible capacity, 6.1GW nominal capacity, 72% survival rateThe report believes its institutional backing and hyperscale-grade delivery record enhance model confidence.
• O'Leary Ventures15.0GW nominal announced capacity, about 489MW recognized by the model, 3.3% survival rateThe report views it as representative of the gap between announced capacity and actual construction.
• Vermaland2.2GW announced capacity, 4% survival rate, credibility score 0.75The report believes it is essentially closer to a land banker than a data center developer.

Impact & implications

The investment implication is that the data center demand theme remains strong, but on the supply side the focus should be on credible capacity rather than nominal announced capacity. For data center REITs and private developers, the market should place higher weight on companies with historical delivery, capital, power pathways, and pre-lease advantages; projects that rely only on press releases and lack tenant or construction evidence should be heavily discounted. For electrification equipment, engineering services, and labor supply chains, even after discounting there is still a credible 135GW pipeline, indicating support for years of end demand and order backlog.

Risks

• Announced capacity may be mistaken by the market or industry forecasts as certain supply, distorting judgments of oversupply.
• Early announced and land bank projects may fail to materialize due to lack of power, capital, tenants, or approvals.
• Legislative and policy risks in states such as Virginia, New York, Texas, Ohio, Arizona, Illinois, and Oregon may affect projects that have not yet broken ground.
• Tenant information in the dataset may be incomplete, and some contracts are not publicly disclosed, affecting tenant quality judgments.
• Development timing is highly uncertain, and the report explicitly does not make a strong judgment on the specific realization pace of the 135GW.
• Long-tail neoclouds, miners, or tenants lacking data center operating track records reduce project credibility.

What to watch

• Whether the share of projects actually under construction continues to rise, especially whether announced projects convert into projects with construction progress.
• New disclosures of hyperscaler self-build projects and pre-leases by investment-grade tenants.
• Changes in data center legislation, permitting pauses, and tax policies in high-risk states such as Virginia and New York.
• Execution progress of Oracle + OpenAI-related projects at developers such as Vantage, Related Digital, BorderPlex, and Crusoe.
• Construction execution, capital expenditures, lease signings, and earnings forecast changes at DLR, EQIX, and BXDC.
• Whether orders, backlog, margins, and delivery bottlenecks at electrification equipment and engineering services companies continue.