AGI Sector Scan — Data Centers, Cloud & Physical Infrastructure

Executive Summary

AGI requires physical space, power delivery, cooling, and structural systems at a scale the world has never built before. The current data center construction boom — $200B+ annually by hyperscalers alone — is just the opening salvo. Recursive self-improvement means each generation of smarter models demands larger training clusters with denser racks, higher power per rack (50-100+ kW vs. the legacy 10-15 kW), and exotic cooling that legacy facilities cannot retrofit. This module covers the full physical stack from the concrete-and-steel shell to the last PDU in the rack.

Highest conviction: DC REITs (especially wholesale/hyperscale), liquid/immersion cooling, electrical switchgear & transformers, and cloud hyperscalers are the primary beneficiaries with the tightest supply constraints and strongest pricing power.
Strong but broader: UPS systems, power distribution (PDUs/busway), DC construction/engineering, and modular/prefab data centers ride the capex wave with solid but more competitive positioning.
Moderate: Rack & enclosure vendors, cable management, fire suppression, and physical security benefit from volume growth but face commodity economics.
Lower: Traditional HVAC (air cooling) faces secular headwinds as the industry shifts to liquid; water treatment is niche and fragmented.

Narrow Sectors

Companies

High Verdict

Medium Verdict

Low Verdict

High — Direct, high-multiplier AGI demand driver; supply constrained or pricing power intact

Medium — Clear AGI tailwind but shared with secular trends; moderate supply elasticity

Low — Indirect or minimal AGI impact; competitive or declining dynamics dominate

Data Center REITs

High

How It Works

Data center REITs own and operate facilities that lease power and space to hyperscalers, enterprises, and cloud providers. Revenue comes from long-term (10-15 year) leases priced per kilowatt of committed power, with annual escalators. Wholesale/hyperscale REITs lease entire halls (1-50+ MW); retail/colocation REITs lease individual cabinets or cages with cross-connect revenue.

Supply / Demand

AI Demand

Hyperscalers are pre-leasing data center capacity years in advance for AI training and inference. Backlog at the major DC REITs has grown 3-5x since 2023. AI workloads demand higher power density (40-100 kW/rack vs. 10-15 kW legacy), forcing REITs to build purpose-built AI-ready facilities. Every major AI lab (OpenAI, Anthropic, xAI, Google DeepMind) is a tenant or co-builder.

Supply Constrained?

Severely. Power availability is the binding constraint — utility interconnection queues are 3-5 years in key markets (Northern Virginia, Dallas, Phoenix). Vacancy rates in top-tier markets are below 3%. Land with power entitlements commands premium prices. REITs with existing power contracts and land banks have enormous competitive moats. Rental rates have increased 20-40% since 2023 in constrained markets.

Key US-Listed Companies

EQIXEquinix

DLRDigital Realty Trust

CBRECBRE Group (DC dev/mgmt)

QTSQTS Realty (Blackstone; limited float)

Cloud Hyperscalers

High

How It Works

Hyperscalers (AWS, Azure, GCP, Oracle Cloud) own and operate the largest data center fleets on Earth, providing compute, storage, and AI-as-a-service to the world. They design custom servers, build their own facilities, and are the single largest buyers of GPUs, networking, and power infrastructure. Their cloud platforms are the delivery mechanism for most AI inference globally.

Supply / Demand

AI Demand

AI is the fastest-growing revenue segment for every major cloud provider. Combined AI-related cloud revenue is growing 50-80% annually. Enterprise AI adoption is driving a second wave of cloud migration (companies need GPU access they cannot build on-prem). Each hyperscaler is spending $50-80B+ annually on capex, with the majority going to AI infrastructure. AGI-scale training requires infrastructure only hyperscalers can afford to build.

Supply Constrained?

Yes, at multiple layers. GPU supply (NVIDIA allocation), power availability, and physical data center capacity all constrain how fast hyperscalers can grow their AI fleets. GPU cloud instances have waitlists. The constraint is not demand but the physical ability to deploy infrastructure fast enough. This is the most capex-intensive buildout in corporate history.

Key US-Listed Companies

AMZNAmazon (AWS)

MSFTMicrosoft (Azure)

GOOGLAlphabet (GCP)

ORCLOracle (OCI)

Neoclouds & GPU Cloud Providers

High

How It Works

Neoclouds are GPU-first cloud providers purpose-built for AI training and inference, without the legacy enterprise software stacks of traditional hyperscalers. They lease or own GPU clusters (typically NVIDIA H100/B200/GB200), offer bare-metal or managed access, and compete on price, availability, and cluster networking quality. Revenue is usage-based (per GPU-hour) with some reserved-capacity contracts.

Supply / Demand

AI Demand

AI startups, mid-tier labs, and enterprises that cannot get direct GPU allocations from NVIDIA or build their own clusters are the core customers. Demand massively exceeds supply. The neocloud market barely existed in 2022 and is now a multi-billion-dollar segment growing 100%+ annually. Recursive self-improvement means training runs get larger every cycle, expanding the total GPU-hours demanded.

Supply Constrained?

Extremely. GPU supply from NVIDIA is the bottleneck. Neoclouds that secured early GPU allocations have enormous backlogs. The risk is that hyperscalers crowd them out over time as GPU supply normalizes. But near-term (2-3 years), this is one of the tightest demand/supply imbalances in tech infrastructure.

Key US-Listed Companies

CRWVCoreWeave

AIC3.ai (AI platform/cloud)

Data Center Liquid Cooling (Direct Liquid & Immersion)

High

How It Works

Liquid cooling removes heat from high-density server racks by circulating liquid (water or engineered fluids) directly to cold plates mounted on CPUs/GPUs (direct-to-chip or DLC) or by fully submerging servers in dielectric fluid (immersion cooling). This replaces or supplements traditional air cooling, which cannot handle rack densities above ~30 kW. Coolant distribution units (CDUs), rear-door heat exchangers, and facility piping are key components.

Supply / Demand

AI Demand

This is the single most transformative shift in DC physical infrastructure driven by AI. NVIDIA's B200 and GB200 NVL72 racks dissipate 60-120 kW per rack and require liquid cooling. Air cooling simply cannot scale to these densities. Every new AI data center being built today includes liquid cooling from day one. The market is growing from ~$2B in 2024 to a projected $15-20B+ by 2028. This is not optional — it is a prerequisite for deploying next-generation AI hardware.

Supply Constrained?

Yes. The industry is scaling from a niche specialty to a mass-market requirement in under 3 years. CDU manufacturing capacity, specialized piping, and engineering talent are all bottlenecks. Vertiv reported liquid cooling orders up 50%+ YoY in 2025. The shift from air to liquid is happening faster than the supply chain can adapt.

Key US-Listed Companies

VRTVertiv Holdings

CARRCarrier Global (Abound liquid cooling)

NVTnVent Electric

GNRCGenerac (liquid cooling via acquisitions)

Electrical Switchgear, Transformers & Power Distribution

High

How It Works

Switchgear, transformers, and power distribution equipment form the electrical backbone of data centers. Medium-voltage switchgear (15-35 kV) receives utility power; transformers step it down to usable voltages; low-voltage switchboards distribute power to UPS systems and PDUs. A single 100 MW data center campus requires dozens of transformers and hundreds of switchgear panels. Lead times for large power transformers now exceed 2-3 years.

Supply / Demand

AI Demand

Every megawatt of AI data center capacity requires corresponding electrical infrastructure. The $200B+ annual hyperscaler capex translates directly into transformer and switchgear orders. AI data centers are also more power-dense, requiring higher-capacity equipment in smaller footprints. The electrification of everything (EVs, heat pumps, AI) is creating a broad supercycle for electrical equipment, but AI is the sharpest demand spike.

Supply Constrained?

Severely, especially for large power transformers. Global transformer lead times have stretched from 12 weeks (pre-2022) to 100-150+ weeks. The transformer manufacturing base was hollowed out over decades of underinvestment. Eaton, Schneider, and ABB are adding capacity but it takes 2-3 years to build a new transformer factory. This is one of the most acute bottlenecks in the entire AI buildout.

Key US-Listed Companies

ETNEaton Corporation

SBGSFSchneider Electric (ADR)

ABBNYABB Ltd (ADR)

HUBBHubbell Inc

GEGE Vernova (grid equipment)

Uninterruptible Power Supply (UPS) Systems

High

How It Works

UPS systems provide battery-backed power conditioning that protects servers from outages, sags, and surges during the seconds-to-minutes gap between a utility failure and backup generator startup. Modern DC-scale UPS units are 1-3 MW modular systems using lithium-ion batteries. They sit between the utility feed and the IT load, providing "five nines" (99.999%) power availability. Every watt delivered to a server passes through a UPS first.

Supply / Demand

AI Demand

UPS demand scales linearly with data center power capacity. Every new MW of AI compute requires a corresponding MW of UPS. Higher power densities in AI racks mean more concentrated UPS deployments. The shift to lithium-ion batteries (vs. legacy VRLA) increases ASP per unit but improves energy density and footprint. Total UPS market for data centers is growing 15-20% annually, with AI as the primary driver.

Supply Constrained?

Moderately. Lithium-ion battery cell supply (shared with EVs) and power semiconductor components (IGBTs, SiC MOSFETs) are the bottlenecks. UPS lead times have improved from 40+ weeks in 2023 to 20-30 weeks in 2025 but remain elevated vs. historical norms. Vertiv and Eaton dominate with ~60% combined market share globally.

Key US-Listed Companies

VRTVertiv Holdings

ETNEaton Corporation

SBGSFSchneider Electric (ADR)

Power Distribution Units (PDUs) & Busway

Medium

How It Works

PDUs distribute power from the UPS output to individual server racks. Rack-mount "intelligent" PDUs monitor per-outlet power consumption, enable remote power cycling, and provide environmental sensing (temperature, humidity). Overhead busway systems replace traditional cable trays with rigid copper bus bars running above or below racks, simplifying power distribution in high-density environments and enabling faster reconfiguration.

Supply / Demand

AI Demand

AI racks at 50-100 kW need higher-amperage PDUs and busway than traditional 5-15 kW racks. Intelligent PDUs with per-outlet monitoring are becoming standard for managing power budgets in AI clusters. Busway adoption is growing because it is faster to deploy and reconfigure than traditional conduit-and-cable. Volume tracks data center construction linearly, but ASP per rack is increasing with density.

Supply Constrained?

Not severely. PDUs are relatively straightforward to manufacture compared to UPS or switchgear. Multiple vendors compete. However, high-density PDUs (60A+ per circuit) are a newer product category with fewer qualified suppliers. Lead times are manageable at 8-16 weeks.

Key US-Listed Companies

VRTVertiv (Geist PDUs)

ETNEaton (ePDUs, busway)

SBGSFSchneider Electric (ADR)

LEGLegrand (ADR: LGRDY; PDUs/busway)

Backup Generators & Engine Systems

Medium

How It Works

Diesel and natural gas generators provide backup power for data centers when the utility grid fails. A typical Tier III/IV data center maintains N+1 or 2N generator redundancy, meaning 1-2 MW generators for every MW of IT load. Generators start within 10-15 seconds of a utility outage and can run for hours to days on stored fuel. Automatic transfer switches (ATS) manage the switchover between utility, UPS, and generator feeds.

Supply / Demand

AI Demand

Every MW of new data center capacity requires corresponding generator backup. AI-driven DC buildout is a direct volume driver. However, generators are mature technology with no AI-specific innovation premium. Some hyperscalers are exploring alternative backup strategies (battery-only, fuel cells, grid redundancy) that could reduce generator dependency over time. Near-term demand remains strong.

Supply Constrained?

Moderately. Large (2+ MW) generator sets have 20-40 week lead times. Engine manufacturing (Caterpillar, Cummins, MTU/Rolls-Royce) has limited surge capacity. But this is a mature supply chain that is better able to respond to demand signals than, say, transformer manufacturing.

Key US-Listed Companies

CATCaterpillar

CMICummins

GNRCGenerac Holdings

Data Center Construction & Engineering

High

How It Works

Data center construction encompasses the design, engineering, and physical building of data center facilities — from site preparation and foundations through mechanical/electrical/plumbing (MEP) systems to commissioning and handover. General contractors, specialty MEP subcontractors, and design-build firms execute projects that range from $500M to $5B+ per campus. Construction timelines are 18-36 months for a new facility.

Supply / Demand

AI Demand

Data center construction spending is the highest it has ever been, driven almost entirely by AI. The US alone has 5+ GW of data center capacity under construction or permitted. Every hyperscaler is simultaneously building multiple campuses. AI data centers are more complex to build (liquid cooling plumbing, higher power density, specialized structural requirements for heavy GPU racks), increasing revenue per project for construction firms.

Supply Constrained?

Yes. Skilled labor (electricians, pipe fitters, controls technicians) is the binding constraint. Data center construction competes with other electrification projects (EV factories, grid upgrades, semiconductor fabs) for the same trades. Construction costs have inflated 15-25% since 2022. Firms with established DC construction practices and workforce pipelines have pricing power.

Key US-Listed Companies

PRIMPrimoris Services

EMEEMCOR Group

MTZMasTec

FIXComfort Systems USA

Modular & Prefab Data Centers

Medium

How It Works

Modular data centers are factory-built, containerized or skid-mounted units that integrate power, cooling, and IT infrastructure into standardized modules. They are manufactured in a controlled factory environment and shipped to site, reducing construction time from 18-24 months to 6-12 months. Configurations range from single-rack edge pods to multi-MW modules that are assembled into full-scale campuses.

Supply / Demand

AI Demand

Speed-to-deployment is critical for AI infrastructure. Modular approaches let hyperscalers deploy capacity 6-12 months faster than traditional stick-built construction. Microsoft and others have adopted modular approaches for some AI deployments. The labor shortage in traditional construction further tilts economics toward factory-built modules. However, the very highest-density AI racks (GB200 NVL72) require custom liquid cooling that does not always fit standard module form factors.

Supply Constrained?

Moderately. The modular DC market is still relatively fragmented with a mix of established players and startups. Factory capacity is being built out but the market is smaller than traditional construction. Standardization challenges limit how quickly modular approaches can capture the highest-density AI deployments.

Key US-Listed Companies

VRTVertiv (SmartMod, prefab power trains)

ETNEaton (modular power skids)

FLEXFlex Ltd (modular DC manufacturing)

Traditional HVAC & CRAC/CRAH Systems

Low

How It Works

Computer Room Air Conditioning (CRAC) and Computer Room Air Handler (CRAH) units cool data centers by circulating chilled air through raised floors or overhead ducts to server racks. Chillers produce chilled water that feeds CRAH coils; economizer modes use outside air when ambient temperatures are low enough. This is the legacy cooling approach used in data centers built before 2020 and for racks under 20 kW.

Supply / Demand

AI Demand

Air cooling cannot scale to the 50-100+ kW/rack densities required by next-generation AI hardware. New AI data centers are being designed with liquid cooling as the primary cooling system, with air handling relegated to ambient room conditioning only. While the installed base of air-cooled data centers is massive and will require maintenance and replacement parts for years, new builds are shifting away from air-only designs. This sector faces secular headwinds from the liquid cooling transition.

Supply Constrained?

No. HVAC equipment is a mature, competitive market with ample manufacturing capacity. Trane, Carrier, Johnson Controls, and others can ramp production readily. Pricing power is limited. The growth opportunity is in hybrid air+liquid systems, where traditional HVAC vendors are pivoting, but the pure air cooling value proposition is declining.

Key US-Listed Companies

TTTrane Technologies

CARRCarrier Global

JCIJohnson Controls International

Racks, Enclosures & Server Cabinets

Medium

How It Works

Server racks and enclosures are the standardized (19-inch or 21-inch) steel frames that house servers, switches, storage, and cabling inside data centers. Modern AI-optimized racks are reinforced to handle heavier GPU servers (a single GB200 NVL72 rack weighs 1,300+ kg), integrated with liquid cooling manifolds, and designed with optimized airflow or sealed compartments for rear-door heat exchangers. Open-frame racks, enclosed cabinets, and custom OEM designs serve different market segments.

Supply / Demand

AI Demand

Volume tracks data center construction linearly — every server needs a rack. AI racks command higher ASPs due to structural reinforcement, liquid cooling integration, and custom form factors. However, racks are fundamentally sheet-metal products with limited differentiation. AI does increase revenue per rack (2-3x ASP for liquid-cooled AI racks vs. standard) but margins remain competitive.

Supply Constrained?

Not significantly. Rack manufacturing is relatively straightforward and multiple vendors can ramp capacity quickly. Custom AI racks with integrated liquid cooling are newer and have fewer qualified suppliers, but the overall rack market is not a bottleneck in the AI buildout.

Key US-Listed Companies

VRTVertiv (DCL racks)

NVTnVent Electric (Hoffman, Schroff)

ETNEaton (rack enclosures)

Cable Management & Structured Cabling

Medium

How It Works

Cable management systems organize, route, and protect the thousands of power, fiber, and copper cables running through a data center. This includes cable trays, ladder racks, overhead cable runways, under-floor pathways, and in-rack cable management accessories. Structured cabling (Cat6A, fiber trunks, patch panels) provides the physical connectivity layer. Proper cable management enables airflow, simplifies maintenance, and supports higher density deployments.

Supply / Demand

AI Demand

AI data centers have significantly more cables per rack (multiple high-speed fiber connections per GPU, plus power and liquid cooling hoses). Cable density is 3-5x higher than traditional enterprise DCs. This drives higher revenue per square foot for cable management vendors. However, the products themselves are commoditized (steel trays, plastic organizers) with limited pricing power. Volume growth is strong; margin expansion is modest.

Supply Constrained?

No. Cable management products are widely manufactured. The structured cabling market is competitive with multiple global vendors. Lead times are normal (4-8 weeks). Not a bottleneck.

Key US-Listed Companies

HUBBHubbell (Wiegmann/cable tray)

NVTnVent Electric (cable management)

LGRDYLegrand (ADR; Ortronics/cabling)

Fire Suppression & Detection Systems

Medium

How It Works

Data center fire protection uses clean-agent suppression systems (FM-200, Novec 1230, inert gas) that extinguish fires without damaging electronics, combined with very early smoke detection (VESDA laser-based systems) that can detect smoke particles at concentrations 1000x below what conventional detectors sense. Pre-action sprinkler systems provide backup. Fire protection is mandated by building codes and insurance requirements, making it non-discretionary for every data center.

Supply / Demand

AI Demand

Fire suppression demand scales with data center square footage and power, so AI-driven DC construction is a direct volume driver. Higher power densities and liquid cooling (with flammable dielectric fluids in some immersion systems) create new fire risks that may require more sophisticated detection and suppression, potentially increasing ASP per deployment. But fire protection is typically 1-2% of total DC construction cost — material to the vendor, not to the hyperscaler's budget.

Supply Constrained?

Not significantly. Fire suppression is a mature industry with established manufacturing. Clean-agent chemicals have faced some supply disruption due to environmental regulations (HFC phase-down), but alternatives (Novec, inert gas) are available. The market is a duopoly (Carrier/Kidde + JCI/Tyco) in many segments.

Key US-Listed Companies

JCIJohnson Controls (Tyco fire)

CARRCarrier Global (Kidde/Edwards)

APGAPi Group (life safety)

Physical Security & Access Control

Medium

How It Works

Data center physical security encompasses perimeter fencing, bollards, vehicle barriers, CCTV surveillance, biometric access control (fingerprint, iris, facial recognition), mantrap entry vestibules, and security operations centers. These systems protect the billions of dollars of hardware and the sensitive data inside data centers. Compliance requirements (SOC 2, HIPAA, government contracts) mandate specific security standards.

Supply / Demand

AI Demand

More data centers = more physical security deployments. AI data centers handling frontier model training or government AI workloads require elevated security (SCIF-level for some contracts). The increasing value of AI hardware ($1B+ in GPUs per campus) justifies higher security spend per site. AI-powered video analytics are also improving security monitoring efficiency. However, physical security is a small fraction of total DC cost (1-2%) and highly fragmented.

Supply Constrained?

No. The physical security market is fragmented with hundreds of integrators and equipment vendors. Components (cameras, card readers, fencing) are commodity products. Labor for installation is the main cost driver. Not a bottleneck.

Key US-Listed Companies

JCIJohnson Controls (Tyco security)

ALLEAllegion (access control)

BRKSAxcelis Technologies (security)

Water Treatment & Cooling Towers

Low

How It Works

Data centers that use evaporative cooling (cooling towers, adiabatic coolers) require water treatment systems to prevent scaling, corrosion, and biological growth in cooling loops. Cooling towers reject heat to the atmosphere through evaporation, consuming millions of gallons of water annually for large facilities. Water treatment chemicals, filtration systems, and monitoring sensors keep cooling water within specification. Some operators use closed-loop dry coolers to eliminate water use entirely.

Supply / Demand

AI Demand

Water use in data centers is a growing sustainability concern. AI data centers use more water per MW because of higher cooling loads. However, the industry trend is toward water-free cooling (air-cooled chillers, dry coolers, closed-loop liquid cooling) precisely because of water scarcity concerns. Hyperscalers have pledged to be "water positive" by 2030. This creates offsetting pressures: more cooling load but less water-based cooling. Net effect is modest for water treatment vendors.

Supply Constrained?

No. Water treatment is a mature, fragmented market. Cooling tower manufacturing has adequate capacity. Products are largely commoditized. The sector benefits from the volume of DC construction but has no AI-specific pricing power or supply constraints.

Key US-Listed Companies

XYLXylem Inc

ECLEcolab (water treatment chemicals)

SPXSPX Technologies (cooling towers)

Building Management Systems (BMS) & DCIM Software

Medium

How It Works

Data Center Infrastructure Management (DCIM) software and Building Management Systems (BMS) monitor and control the mechanical, electrical, and environmental systems in data centers. DCIM tracks power usage per rack, cooling efficiency (PUE), capacity planning, and asset inventory. BMS controls HVAC, fire systems, and lighting. Together they provide real-time visibility into the physical plant and enable automated responses to environmental changes or equipment failures.

Supply / Demand

AI Demand

AI data centers are more complex to manage than traditional facilities — higher power densities, liquid cooling loops, heterogeneous hardware (GPUs, TPUs, custom ASICs), and tighter thermal tolerances all demand better monitoring and control. AI itself is being used within DCIM tools to optimize cooling, predict equipment failures, and manage capacity. Every new DC needs DCIM/BMS. Revenue scales with the number of managed facilities, but software margins are high.

Supply Constrained?

No. Software scales elastically. The DCIM market is competitive with both pure-play vendors and modules from Schneider, Vertiv, and Siemens. Hyperscalers build their own monitoring systems, limiting the addressable market to colocation providers and enterprises. Medium-conviction because the market grows but is not supply-constrained.

Key US-Listed Companies

SBGSFSchneider Electric (EcoStruxure)

VRTVertiv (Trellis/LIFE)

SIEGYSiemens (ADR; Desigo CC)

DC Site Selection, Land & Real Estate Services

Low

How It Works

Data center site selection involves identifying parcels with access to high-voltage utility transmission, fiber connectivity, water, and favorable permitting/tax environments. Real estate services firms advise hyperscalers on site acquisition, negotiate utility contracts, and manage entitlements. Land values near key substations and fiber routes have appreciated dramatically. Some firms specialize in rezoning agricultural or industrial land for data center use.

Supply / Demand

AI Demand

The scramble for power-ready DC sites is intense. Hyperscalers are acquiring land adjacent to nuclear plants, hydroelectric dams, and newly built gas-fired generation. Brokers and advisory firms benefit from transaction volume. However, the real estate advisory segment is fragmented and project-based, with limited recurring revenue. Most of the value accrues to the landowners (often private) and the DC REITs that develop the sites, not to the services firms.

Supply Constrained?

Land with grid power entitlements is extremely constrained, but the publicly traded advisory firms are not the bottleneck — they are service providers. The constraint is in the utility grid and permitting process, which is captured in the Electrical Switchgear and DC REIT sectors. Advisory firms benefit from deal flow but have limited pricing power or moats.

Key US-Listed Companies

CBRECBRE Group (DC advisory/brokerage)

JLLJones Lang LaSalle

Master Company List — 50 Tickers

#	Ticker	Company	Narrow Sector	Verdict
1	EQIX	Equinix	Data Center REITs	High
2	DLR	Digital Realty Trust	Data Center REITs	High
3	AMZN	Amazon (AWS)	Cloud Hyperscalers	High
4	MSFT	Microsoft (Azure)	Cloud Hyperscalers	High
5	GOOGL	Alphabet (GCP)	Cloud Hyperscalers	High
6	ORCL	Oracle (OCI)	Cloud Hyperscalers	High
7	CRWV	CoreWeave	Neoclouds / GPU Cloud	High
8	AI	C3.ai	Neoclouds / AI Platform	High
9	VRT	Vertiv Holdings	Liquid Cooling / UPS / PDUs / Racks / Modular / DCIM	High
10	CARR	Carrier Global	Liquid Cooling / HVAC / Fire Suppression	Medium
11	NVT	nVent Electric	Liquid Cooling / Racks / Cable Mgmt	High
12	ETN	Eaton Corporation	Switchgear / UPS / PDUs / Modular	High
13	SBGSF	Schneider Electric (ADR)	Switchgear / UPS / PDUs / DCIM	High
14	ABBNY	ABB Ltd (ADR)	Switchgear / Transformers	High
15	HUBB	Hubbell Inc	Switchgear / Cable Mgmt	High
16	GE	GE Vernova	Switchgear / Grid Equipment	High
17	GNRC	Generac Holdings	Backup Generators / Liquid Cooling	Medium
18	CAT	Caterpillar	Backup Generators	Medium
19	CMI	Cummins	Backup Generators	Medium
20	PRIM	Primoris Services	DC Construction & Engineering	High
21	EME	EMCOR Group	DC Construction & Engineering	High
22	MTZ	MasTec	DC Construction & Engineering	High
23	FIX	Comfort Systems USA	DC Construction & Engineering	High
24	FLEX	Flex Ltd	Modular DC Manufacturing	Medium
25	TT	Trane Technologies	Traditional HVAC / CRAC	Low
26	JCI	Johnson Controls	HVAC / Fire / Security	Medium
27	LGRDY	Legrand (ADR)	PDUs / Busway / Cabling	Medium
28	APG	APi Group	Fire Suppression / Life Safety	Medium
29	ALLE	Allegion	Physical Security / Access Control	Medium
30	XYL	Xylem Inc	Water Treatment	Low
31	ECL	Ecolab	Water Treatment Chemicals	Low
32	SPX	SPX Technologies	Cooling Towers	Low
33	SIEGY	Siemens (ADR)	BMS / DCIM	Medium
34	CBRE	CBRE Group	DC REITs / Site Advisory	Medium
35	JLL	Jones Lang LaSalle	DC Site Advisory	Low
36	PWR	Quanta Services	DC Electrical Construction	High
37	APH	Amphenol	DC Power Connectors / Busbar	High
38	WCC	WESCO International	Electrical Distribution / DC Supply Chain	Medium
39	AYI	Acuity Brands	DC Lighting / Controls	Low
40	AAON	AAON Inc	Precision Cooling Units	Medium
41	BMI	Badger Meter	Water Flow / Cooling Monitoring	Low
42	IEX	IDEX Corporation	Pumps / Fluid Handling for Cooling	Medium
43	REXR	Rexford Industrial Realty	Industrial RE (DC-adjacent land)	Low
44	POWL	Powell Industries	Switchgear / Custom Power Solutions	High
45	GEV	GE Vernova (alt ticker)	Grid / Transformers / Switchgear	High
46	STRL	Sterling Infrastructure	DC Site Work / Foundations	Medium
47	DY	Dycom Industries	DC Fiber / Electrical Contracting	Medium
48	ACM	AECOM	DC Engineering / Design	Medium
49	TTEK	Tetra Tech	Environmental / Water for DC Sites	Low
50	MEG	Montrose Environmental	DC Environmental Compliance	Low

Methodology & Caveats

Verdict criteria: "High" = the company's core revenue is directly accelerated by AGI-scale data center buildout with defensible positioning (supply constraints, market share, technical moats). "Medium" = clear AGI tailwind but the company has large non-DC segments, faces commodity competition, or has moderate supply elasticity. "Low" = AI impact is indirect, the sector faces secular headwinds (air cooling displacement), or the market is fragmented with no pricing power.
Multi-sector companies: Vertiv (VRT), Eaton (ETN), Schneider (SBGSF), and Carrier (CARR) appear across multiple narrow sectors. Their master-table verdict reflects aggregate exposure. VRT and ETN are particularly well-positioned because their product portfolio spans the entire DC power and cooling stack.
ADRs: SBGSF (Schneider), ABBNY (ABB), SIEGY (Siemens), LGRDY (Legrand) are foreign-domiciled but US-traded. Liquidity varies; SBGSF is OTC pink sheets with lower volume than the Paris-listed primary.
Private companies excluded: Several important players in this space are private, including Compass Datacenters, Stack Infrastructure, Vantage Data Centers, DataBank, QTS (Blackstone), and most specialty DC construction firms. The publicly investable universe captures the ecosystem but misses some pure-play DC developers.
Overlap with other modules: Cloud hyperscalers (AMZN, MSFT, GOOGL) also appear in Module 02 (Semiconductors) and Module 06 (Software/AI). Power equipment (ETN, GE) overlaps with Module 01 (Energy). The company universe is intentionally cross-referenced across modules for completeness.
Data vintage: Analysis based on publicly available information through May 2026. Supply chain dynamics and company positioning can shift rapidly.