All themes

Supply-chain research brief

AI Power & Grid Bottlenecks

Research generation, transformers, switchgear, cooling, interconnection, and electrical construction constraints for AI data centers.

26 tracked companiesUpdated 2026-09-20

Market map

Where constraints can shape the market

AI data-center growth is ultimately an electrical-infrastructure build. New compute capacity needs generation, utility interconnection, substations, large power transformers, switchgear, busway, backup power, cooling, and skilled contractors. Project timing is frequently determined by the slowest of these layers rather than by server availability.

Constraint Alpha separates equipment suppliers from power developers and integrators. Transformer factories and qualified high-voltage equipment may have multi-year lead times; turbines and onsite generation can relieve one constraint while creating fuel, permitting, emissions, or service dependencies; liquid cooling and power distribution determine how much compute fits inside a site.

The directory uses backlog, capacity expansion, utility approvals, customer commitments, and official product evidence as validation signals. It does not infer an AI revenue percentage from generic data-center language. Each record states whether the evidence is verified, sourced, or still a hypothesis and identifies the risk that could normalize the bottleneck.

Agent-native research

Versioned notes keep public evidence and member-only interpretation separate while preserving one canonical source.

Agent-assisted · high confidence

AI Power & Grid: A Source-Backed Research Baseline

Constraint Alpha starts the AI power research process with the physical system: firm generation, grid connection, power distribution, cooling, and the qualified equipment needed to deliver usable megawatts to a data center.

Research universe

Supply-chain layers we track

  • Generation & Turbines3 tracked companies
  • Transformers & Switchgear6 tracked companies
  • Cables & Grid Components3 tracked companies
  • On-site & Backup Power3 tracked companies
  • Cooling & Data-center Power4 tracked companies
  • EPC & Interconnection3 tracked companies
  • Utilities & Power Developers4 tracked companies

Public preview

Five representative companies in the research universe

Examples are sampled across supply-chain layers; ordering is not a ranking. Scores, thesis, risks and full evidence remain in the Atlas.

GE Vernova

GEV · NYSE

Supply-chain layerGas turbines and grid equipment

Evidence status: Verified · Reviewed 2026-07-12

ABB

ABBN · SIX

Supply-chain layerElectrification and switchgear

Evidence status: Sourced · Reviewed 2026-07-12

Hubbell

HUBB · NYSE

Supply-chain layerUtility grid components

Evidence status: Sourced · Reviewed 2026-07-12

Bloom Energy

BE · NYSE

Supply-chain layerOnsite fuel-cell generation

Evidence status: Verified · Reviewed 2026-07-12

Modine Manufacturing

MOD · NYSE

Supply-chain layerData-center cooling

Evidence status: Verified · Reviewed 2026-07-12

Full research access

Open all 26 companies, scores, risks, and sources.

Verify your email to use the filterable directory and review the complete evidence-backed research set.

Get full directory by email

Questions

How to read this market

What is the binding AI power bottleneck?

It varies by location and project: generation availability, interconnection, transformers, switchgear, permitting, cooling, or electrical labor can each set the schedule.

Why include utilities and EPC firms?

They validate demand, control interconnection and construction execution, and provide supply-chain context even when they are not pure-play equipment suppliers.

What evidence strengthens a supplier thesis?

Backlog growth, capacity sold ahead, named data-center awards, utility approvals, customer prepayments, and margin-supported pricing are stronger than generic AI marketing.