Explainers

Data center stocks: who owns, builds, powers and cools AI data centers

Data center stocks are the companies that own, build, fill, power and cool data centers. The layers, the tracked names in each, and how AI-exposed they are.

Written by Supercycle Research AI · How it's made · Published · Data as of

Data center stocks are public companies that make money from data centers. Some own the buildings and rent them out. Some build new campuses. Others sell what goes inside: servers, chips, networking gear and cooling. Others supply the power. Supercycle tracks 63 stocks in its Data Centers & Hardware layer and hundreds more in the layers around it.

The label is wide, so the useful question is how much of each company depends on data centers being built for AI. Supercycle's AI Exposure Score answers that on a 0 to 100 scale. This guide walks through each group, the tracked names in it, and their scores, which update every night.

What makes a company a data center stock?

A company counts if data center spending lands in its revenue. That can happen in five ways:

  1. It owns the building. Data center REITs and operators rent space, power and connections to cloud providers and enterprises.
  2. It builds the campus. Developers secure land and power, then lease the finished halls, often to one or two large tenants.
  3. It sells what goes inside. Servers, chips, memory and networking gear.
  4. It keeps the building cool and powered. Cooling systems, electrical gear, generation and the contractors who wire it all.
  5. It rents out the computing power. Neoclouds fill data centers with GPUs and rent them by the hour or by contract.

The same company can sit in more than one group. Vertiv is in both the data centers and cooling layers. Core Scientific is both a developer and a neocloud. Supercycle lists each stock under every layer it belongs to.

How exposed are data center stocks to AI?

It varies a lot, even inside one layer. A company that sells only AI servers scores near the top. A company where data centers are one customer group among many scores much lower. The table shows the main groups, with live counts and scores.

Group Supercycle layer Tracked stocks Examples (score, tier)
REITs, operators and server makers Data Centers & Hardware 63 Digital Realty (60, Strategic), Super Micro (90, Pure Play)
Cooling Cooling & Thermal 10 Vertiv (85, Core), AAON (58, Strategic)
Networking and optics Networking & Optics 24 Astera Labs (92, Pure Play), Arista (80, Core)
Chips Semiconductors 65 NVIDIA (97, Pure Play), Broadcom (78, Core)
Memory Memory & Storage 9 Micron (78, Core)
Power generation Power & Energy 62 GE Vernova (55, Strategic), Constellation (68, Strategic)
Grid and electrical work Utilities & Grid 41 Eaton (50, Strategic), IES Holdings (62, Strategic)
Rented GPU capacity Neoclouds 23 CoreWeave (97, Pure Play), Nebius (97, Pure Play)

The AI infrastructure explainer covers all twelve infrastructure layers and how each one gets paid.

Which data center REITs and operators does Supercycle track?

These are the landlords. They own data center buildings and earn rent, usually under long contracts. The glossary entries on data center REITs and colocation explain the model.

  • Equinix (EQIX), score 55: rents connected data center space in many cities, and builds larger xScale campuses for AI and cloud tenants.
  • Digital Realty (DLR), score 60: rents both whole halls and smaller spaces to cloud providers and enterprises. AI leasing is now a main driver of its new bookings.
  • Iron Mountain (IRM), score 35: a records storage company with a growing data center business.
  • GDS Holdings (GDS), score 65, and VNET Group (VNET), score 60: data center operators in China.
  • Blackstone Digital Infrastructure Trust (BXDC), score 65: owns newly built data centers leased on long contracts to hyperscale tenants.

REITs tend to land in the Strategic tier. AI demand is growing inside them, but they also serve a large base of older enterprise and cloud customers.

Which data center hardware stocks does Supercycle track?

The same layer also holds the companies that build the servers and racks that fill the halls.

The chips inside those servers sit in their own layers. NVIDIA (score 97) and Broadcom (score 78) lead Semiconductors, and Micron (score 78) is in Memory & Storage.

What are the data center networking stocks?

AI clusters link thousands of chips together, so the switches, cables and optical parts that move data between them are a large part of the bill. Supercycle tracks 24 stocks in Networking & Optics, and it is one of the most AI-exposed layers.

  • Astera Labs (ALAB), score 92, Pure Play: chips that connect processors, memory and switches inside AI servers.
  • Credo Technology (CRDO), score 88: high-speed connectivity chips and cables.
  • Arista Networks (ANET), score 80: the switches that tie AI clusters together.
  • Lumentum (LITE), score 75, and Coherent (COHR), score 65: optical parts that carry data as light between racks and buildings.

What are the data center cooling stocks?

AI chips run hot, and packed racks need far more cooling than older data centers did. Much of it now comes from liquid cooling. Supercycle tracks 10 stocks in Cooling & Thermal.

  • Vertiv (VRT), score 85, Core: power distribution, thermal management and liquid cooling for data centers. It is the highest-scoring name in the layer.
  • AAON (AAON), score 58, Johnson Controls (JCI), score 55, Modine (MOD), score 55, and Trane (TT), score 52: heating and cooling makers whose data center lines sit next to commercial HVAC, vehicle or other thermal products.

The gap between Vertiv and the rest is the point of the score. All of them sell data center cooling, but for most of them it is one product line among many.

What about data center power stocks?

Power is now the main limit on how fast new data centers open. Generation sits in Power & Energy, and the grid gear and electrical contractors sit in Utilities & Grid. Examples include GE Vernova (score 55) for gas turbines and grid equipment, Eaton (score 50) for electrical gear, and contractors such as Quanta Services (score 55), IES Holdings (score 62) and Comfort Systems USA (score 55).

The AI power stocks analysis goes deal by deal through which power companies have signed contracts to supply data centers.

Which publicly traded developers have 1 GW or more of data center capacity?

The buildout tracker logs AI data center campuses with their operator, capacity, status and a source for each. Adding up each public developer's tracked campuses shows ten with at least 1,000 megawatts, or one gigawatt, as of November 8, 2026. The hyperscalers are left out here, since they build mainly for their own use.

Developer Score Tracked capacity Campuses Status
Fermi (FRMI) 92 11,000 MW 1 Project Matador, Amarillo, under construction
Nebius Group (NBIS) 97 3,130 MW 5 Two 1,200 MW US campuses, in Missouri (under construction) and Pennsylvania (announced)
Core Scientific (CORZ) 80 3,000 MW 2, plus Denton Muskogee and Pecos expansions, gross power, under construction
IREN (IREN) 65 2,750 MW 2 Sweetwater and Childress, Texas, both partly running
Cipher Mining (CIFR) 55 2,477 MW 5 Texas sites, the largest two announced, Colchis and Apollo
TeraWulf (WULF) 85 1,839 MW 3 New York and Kentucky, including a 1 GW Kentucky site (announced)
Hut 8 (HUT) 72 1,745 MW 3 Beacon Point, Texas, 1,000 MW under construction
BCE (BCE) 25 1,700 MW 2 Bell AI Fabric in British Columbia and Saskatchewan, under construction
Applied Digital (APLD) 95 1,540 MW 5 Polaris Forge and Delta Forge campuses, one partly running
Riot Platforms (RIOT) 35 1,101 MW 2 Corsicana announced, Rockdale under construction

Three things to read from the table. First, most of this capacity is not running yet. The status column, from announced to under construction to operational, is what turns a plan into rent. Second, many of these developers began as bitcoin miners. They already had the power and land, and they are turning those sites into AI data centers. Third, a megawatt figure is not always the same measure. Some companies report gross power at the site and others report power for the computers alone, so totals across companies are a rough guide.

Several of these developers are also neoclouds. The neocloud explainer covers how that model works.

How can you get exposure to data centers through stocks?

Each group carries a different kind of risk.

  • Landlords (REITs and developers) earn rent under long contracts. Their income moves more slowly than chip sales, in both directions. The risk is tied to tenants: a developer with one or two tenants depends on those few contracts.
  • Hardware sellers (servers, chips, networking, cooling) get paid when a product ships, so their sales follow data center spending closely. They feel a cut in spending first.
  • Power and grid companies earn from long power contracts or construction jobs. For large utilities, data centers are one part of a much bigger business, which keeps scores lower.
  • Neoclouds rent out GPU capacity. They are among the most AI-exposed stocks, and they often depend on a short list of large customers.

All of these groups get paid from the same source: the capital spending of a small number of cloud providers and AI labs. When that spending changes, every layer feels it.

This guide describes how each group earns money. It is not a recommendation to buy or sell any stock.

How do you track data center stocks on Supercycle?

  1. Open the Data Centers & Hardware category for the full list with live scores. The cooling and networking pages list the suppliers.
  2. Sort by score on the AI Exposure Leaderboard to see where a data center stock ranks against the whole universe.
  3. Check the buildout tracker for campuses, capacity and power deals. Each row links to its source, and the status column shows when a campus moves from announced to running.
  4. Use the heatmap to see every layer side by side, with each stock sized by its value and colored by its recent move.
  5. Read the score rationale on any stock page. It says what the company sells into data centers and how much of the business that is.

Frequently asked questions

What are data center stocks?

Public companies that earn money from data centers: the REITs and operators that own the buildings, the developers that build new campuses, and the suppliers of servers, chips, networking, cooling and power that go inside or around them. Supercycle tracks 63 stocks in its Data Centers & Hardware layer alone.

Which data center stock has the highest AI exposure score?

Among the building owners and hardware makers in the Data Centers & Hardware layer, Super Micro Computer scores 90, the Pure Play tier, because it sells AI servers. Among the REITs, Digital Realty scores 60 and Equinix 55. Scores are recomputed each night on the leaderboard.

Are data center REITs AI stocks?

Partly. A data center REIT rents space and power to cloud providers and enterprises, and AI demand is a rising share of that leasing. A large base of older colocation business sits underneath it, which is why Equinix and Digital Realty score in the Strategic tier rather than higher.

Which publicly traded developers have 1 GW or more of AI data center capacity?

On the buildout tracker as of November 8, 2026: Fermi, Nebius, Core Scientific, IREN, Cipher, TeraWulf, Hut 8, BCE, Applied Digital and Riot each have 1,000 megawatts or more across their tracked campuses. Most of that capacity is announced or under construction, not yet running.

How do I track data center stocks on Supercycle?

Open the Data Centers & Hardware category page for the full list with scores, the buildout tracker for campuses and power deals, and the heatmap to see the layers side by side. Each stock page shows the score, the rationale and any tracked campus or deal.

The data behind this post

Sources

  1. Federal Register, Fermi America Project Matador application, October 9, 2026
  2. Data Center Dynamics, Nebius breaks ground on Independence, Missouri campus, May 2026
  3. Data Center Dynamics, Nebius details 1.2 GW Pennsylvania campus, July 2026
  4. Core Scientific, Muskogee expansion to 1.5 GW of gross power, August 14, 2026
  5. Core Scientific, Pecos expansion to 1.5 GW of gross power, April 27, 2026
  6. Data Center Dynamics, IREN energizes first phase of 2 GW Sweetwater campus, May 2026
  7. 24/7 Wall St, Cipher Digital conditional ERCOT status for Texas capacity, September 16, 2026
  8. TeraWulf acquires 1 GW eastern Kentucky campus, May 26, 2026
  9. Hut 8 fully commercializes 1 GW Beacon Point campus, July 20, 2026
  10. CBC, Bell AI data centre in Saskatchewan, September 14, 2026
  11. Applied Digital fiscal Q4 and full year 2026 results, July 27, 2026
  12. Riot Platforms third quarter 2026 results, November 4, 2026

About the author

Supercycle Research AI is not a person. It is the byline on every Supercycle blog post, because the posts are researched and written by AI: Anthropic's Claude models, working from the datasets behind the buildout tracker, the AI Exposure Scores, the investor and congress trackers and the AI Fear & Greed Index, and from the primary sources each post lists. More about Supercycle Research AI.

Researched and written by AI from Supercycle's data and the sources listed above, and checked against them before publishing. How Supercycle is made. Nothing on Supercycle is investment advice. Data is end of day and dated where it appears.

Read next