North American data center pipeline rises to 338GW, construction pace slows and stranded capacity climbs to 38GW
AI summary card
North American data center pipeline rises to 338GW, construction pace slows and stranded capacity climbs to 38GW
Bernstein's monthly tracking shows the data center project pipeline increased 14GW month-on-month to 338GW, but capacity under construction rose only modestly to 63.2GW, requiring about 12 years to absorb the pipeline at the current build pace.
- The project pipeline grew 4% from the previous month, increasing by 14GW to 338GW, mainly led by hyperscalers, developers, and colocation data centers.
- Capacity under construction increased by only about 200MW to 63.2GW; hyperscalers and Neocloud together dragged it by about 1GW month-on-month.
- Stranded capacity increased by 4GW to 38GW, accounting for 11% of the total pipeline, with local opposition and NIMBY issues remaining key constraints.
- The behind-the-meter pipeline rose to 131GW, but June additions slowed and represented only 12% of June's total pipeline additions.
- Using the disclosed per-MW opportunity, data center-related TAM includes PWR at approximately $4.6T, VRT at approximately $1T, Schneider at $902B, and ETN at $791B.
Report interpretation
Overview
This report updates Bernstein's monthly data center capacity tracking, covering changes in North American data center projects by development stage, operator type, and geography, and linking pipeline, under construction, stranded, behind-the-meter power, and electrical OEM data center TAM to the listed-company investment implications.
Core views
The core conclusion is that demand remains strong while supply realization is slower: the project pipeline continues to expand to 338GW, indicating AI and cloud computing infrastructure demand still has multi-year support; however, growth in capacity under construction remains limited, and at the current TTM construction pace it would take about 12 years to absorb the pipeline, a further extension from 10 years last month. At the same time, stranded capacity has risen to 38GW, reflecting that grid interconnection, approvals, and local opposition are becoming more explicit bottlenecks.
Analysis framework
The report uses a monthly capacity-tracking framework, splitting data center capacity by pipeline, under construction, active, and stranded stages, and further analyzing month-over-month and trailing-12-month changes by operator type, hyperscaler, Neocloud, colocation data centers, crypto miners, and state/RTO geography; at the same time, it estimates TAM for electrical equipment and infrastructure companies based on current pipeline capacity and company-disclosed per-MW revenue opportunities.
Methodology notes
Track data center capacity by stage, operator, and geography
The report splits project capacity into pipeline, capacity under construction, active capacity, and stranded capacity, and uses month-over-month and TTM changes to assess demand strength, construction pace, and bottlenecks.
Estimate data-center-related TAM based on per-MW opportunity
The report uses current data center project pipeline capacity and combines it with company-disclosed per-MW revenue opportunities to estimate the potential market size for electrical and infrastructure companies such as PWR, ETN, VRT, and Schneider.
Stranded capacity is postponed, canceled, unapproved, or withdrawn capacity
The report uses the stranded capacity ratio to total pipeline to measure how approvals, grid readiness, and local opposition affect realization of data center build-out.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- ETN / HUBB / LGN / J / TRMB / URIU.S. mechanical and electrical equipment beneficiaries
- Strengths
- Data center-related demand is strong and equipment and labor bottlenecks support order flow and backlog.
- Weaknesses
- Valuation and execution pace depend on projects moving from pipeline to under construction and delivery.
- Comparison
- The report maintains these names as Outperform, with a more constructive stance than Market-Perform names such as CAT, CMI, OSK, and PWR.
- Risks
- Construction slowdown, project cancellations, delays in power interconnection, or client capex cuts.
- VRT / NVT / JCIDiverse industrial and data center infrastructure beneficiaries
- Strengths
- Directly benefit from AI data center power, thermal management, and infrastructure spending.
- Weaknesses
- Demand growth is already well-reflected by the market, with higher sensitivity to execution ability and valuation.
- Comparison
- The report maintains VRT, NVT, and JCI as Outperform, and CARR as Market-Perform.
- Risks
- AI buildout pacing below expectations and easing supply-chain bottlenecks reducing pricing power.
- DLR / EQIX / CRWVCommunication infrastructure and colocation/cloud infrastructure names
- Strengths
- Colocation data center and cloud infrastructure demand benefits from AI capacity expansion.
- Weaknesses
- CRWV is rated Underperform, reflecting higher valuation or business-model risk.
- Comparison
- DLR and EQIX are Outperform, while CRWV is Underperform.
- Risks
- Financing costs, energy costs, lease pricing, customer concentration, and project delivery risk.
- AMZN / META / GOOGL / MSFT / ORCLHyperscalers and primary cloud capex issuers
- Strengths
- Amazon and Google drove the month's hyperscaler pipeline additions, while Microsoft continues to have direct buildout additions.
- Weaknesses
- Capex intensity is high; returns could come under pressure if cloud/AI revenue growth does not keep up with spending.
- Comparison
- The report rates AMZN, META, MSFT, and ORCL as Outperform and GOOGL as Market-Perform.
- Risks
- AI monetization slower than capex growth, shifts in leasing versus on-prem strategy, regulation, and power constraints.
- WULF / CIFR / IREN / CORZ / RIOT / CLSK / MARABitcoin miners/emerging AI infrastructure
- Strengths
- Bitcoin miners have about a 30GW planned power footprint and have signed about 6GW with hyperscalers, Neocloud, and AI chip makers, providing a power-resource advantage for transitioning toward AI/HPC.
- Weaknesses
- Business transformation depends on execution of power, site, and client contracts, and still faces execution and financing risk.
- Comparison
- The report rates WULF, CIFR, IREN, CORZ, RIOT, and CLSK as Outperform, while MARA is Market-Perform.
- Risks
- Crypto cycles, client contract failures, delays in grid interconnection, and capex pressures.
- APLD.USDevelopers/data center pipeline-related company
- Strengths
- The report notes Applied Digital added about 640MW to pipeline this month and was one of the key contributors among developers and colocation-related additions.
- Weaknesses
- As a single company, this is not a primary coverage focus; most information comes from project pipeline changes rather than full company fundamentals.
- Comparison
- Listed alongside EdgeConnex, Beale Infrastructure, Beltline, Digital Realty, and QTS as one of the key contributors to monthly additions.
- Risks
- Project approvals, power supply, financing, and buildout realization.
Key data
- Total Project Pipeline338GWIncreased by 14GW from the prior month, up about 4%.
- Capacity Under Construction63.2GWIncreased only about 200MW month-on-month; at the current build pace, it would take about 12 years to absorb the 338GW pipeline.
- Stranded Capacity38GWUp 4GW month-on-month, accounting for about 11% of total project pipeline.
- Behind-the-meter Pipeline131GWIncreased by 1.6GW month-on-month; June additions contributed about 12% of June pipeline additions, below the 36% year-to-date average.
- Hyperscaler Pipeline Additions3.5GWThe month's additions were mainly driven by Google at about 1.8GW and Amazon at about 1.4GW.
- Developer and Colocation Data Center AdditionsDevelopers 2.2GW, Colocation Data Centers 2.0GWEdgeConnex, Beale Infrastructure, Beltline, Digital Realty, Applied Digital, and QTS were important contributors.
- Electrical Equipment TAM EstimatePWR $4.6T; VRT $1T; Schneider $902B; ETN $791B; ABB $676B; LGN $343BEstimated based on current pipeline capacity and the per-MW opportunities disclosed by each company.
- Top 5 Hyperscaler Capex OutlookAbout $727B in 2026, about $916B in 2027The report says consensus Capex for the top 5 hyperscalers is expected to rise about 76% year-over-year in 2026 and grow a further about 26% in 2027.
Impact & implications
The investment implications remain structurally positive over the long term but place more emphasis on execution and bottlenecks. Electrification equipment, construction services, liquid cooling/power, colocation data centers, AI server chains, and miner/new AI infrastructure names with available power resources benefit from multi-year demand; however, project approvals, local opposition, grid interconnection, construction resources, and BTM power realization speed will determine revenue recognition timing.
Risks
- Stranded capacity is rising, and postponed, canceled, unapproved, or withdrawn projects may continue to increase.
- Local opposition and NIMBY political pressure could slow data center approvals and construction.
- There is uncertainty around power interconnection, grid capacity, and behind-the-meter power realization.
- At the current build pace, the pipeline would take about 12 years to digest, so demand realization may be slower than market expectations.
- If hyperscaler capex is revised down, orders for equipment, construction, and colocation data centers would be affected.
- High-valuation AI infrastructure and semiconductor-related names are sensitive to changes in growth expectations.
What to watch
- Whether the monthly project pipeline continues to grow in subsequent months, especially whether TTM additions keep weakening.
- Whether capacity under construction can re-accelerate and whether hyperscalers resume larger-scale ground-breaking.
- Whether stranded capacity as a share of total pipeline remains around 11% or continues rising.
- Power and approval progress in key states such as Texas, Virginia, Ohio, Georgia, and RTOs such as PJM and ERCOT.
- Changes in the share of behind-the-meter projects within newly added pipeline and projects under construction.
- Whether the 2026 and 2027 capex consensus for the top 5 hyperscalers is revised up further.