AI data center stocks: who gets paid to build the buildout
The companies operating, building, powering, and cooling AI data centers, graded by how direct their exposure is.
Updated · newest price in the table below
The companies
Operators: the GPU clouds
The companies whose product is the capacity itself, renting accelerator time or leasing turnkey AI data center capacity to AI labs and hyperscalers under multi-year contracts. The most direct exposure on this page, and the most concentrated: a handful of customers account for most of the revenue at nearly every name in this tier.
| Ticker | Company | Exposure | Why it is here | Price | Day | Market cap | Next earnings | Our coverage |
|---|---|---|---|---|---|---|---|---|
| CRWV | CoreWeave, Inc. | Direct | The largest of the listed GPU clouds, built on long-term take-or-pay contracts with a short customer list. The pure expression of the buildout, in both directions. | $39.51 | -12.02% | $22.0B | - | 100 |
| NBIS | Nebius Group N.V. | Direct | GPU cloud running a data center it owns in Finland, a third-party-built facility in New Jersey, and colocation sites in the US, Europe, and Israel, with owned sites in Missouri and Alabama due to deploy capacity in 2027. | $81.31 | -16.11% | $23.8B | Feb 11, 2027 | 80 |
| IREN | IREN LIMITED | Direct | Bitcoin miner decommissioning its mining fleet as its power-secured sites move to AI cloud. AI cloud revenue passed mining revenue for the first time in the June 2026 quarter and reached $200 million of $228 million in total revenue in the September 2026 quarter. | $11.82 | -17.97% | $5.0B | Feb 4, 2027 | 59 |
| APLD | Applied Digital Corporation | Direct | Builds AI campuses and leases them to CoreWeave and investment-grade hyperscalers, with leases for about 1.41 GW of critical IT load across five campuses as of August 31, 2026, which makes contract terms the thing to read first. | $11.49 | -17.81% | $3.5B | Jan 6, 2027 | 34 |
| CORZ | Core Scientific, Inc. | Direct | Converted miner whose high-density colocation business, still concentrated with CoreWeave, produced $215.4 million of its $232.9 million in third-quarter 2026 revenue. Leases signed in July 2026 with AMD and a neocloud are expected to start producing revenue in 2027. | $11.33 | -15.19% | $3.7B | - | 25 |
| WYFI | WhiteFiber, Inc. | Direct | AI data center and GPU cloud operator funding its buildout with successive convertible raises, which makes the financing window its binding constraint. | $6.35 | -16.99% | $288M | - | 5 |
Miners turning into data centers
Bitcoin miners hold exactly what the buildout is short of, powered sites with grid connections, and every one of them now markets that fact. The grades here separate signed hosting contracts from ambitions: an announcement is not a lease, and a megawatt in a press release is not revenue.
| Ticker | Company | Exposure | Why it is here | Price | Day | Market cap | Next earnings | Our coverage |
|---|---|---|---|---|---|---|---|---|
| RIOT | Riot Platforms, Inc. | Significant | Has leased 1,151 MW of critical IT capacity for about $38 billion of contracted revenue: a 15-year, 910 MW powered land lease covering its entire Corsicana campus, announced on November 4, 2026, a 20-year, 191 MW lease at Rockdale, and an AMD lease at Rockdale. Bitcoin mining was still $116.0 million of its $181.4 million in third-quarter 2026 revenue. | $16.23 | -8.46% | $6.1B | - | 17 |
| GLXY | Galaxy Digital Inc. | Significant | Digital asset and data center company whose Helios campus, a former bitcoin mining site, has its first 800 MW of approved power, Helios I, leased to CoreWeave under 15-year agreements, with another 830 MW of approved capacity at Helios II not yet leased. | $11.22 | -8.93% | $6.6B | - | 6 |
| WULF | TeraWulf Inc. | Direct | Miner whose AI data center leases already produce most of its revenue, about 83% in the September 2026 quarter, with a 20-year Anthropic lease for about 401 MW starting delivery in the second half of 2027. | $9.85 | -14.94% | $4.9B | - | 38 |
| CIFR | Cipher Digital Inc. | Significant | Formerly Cipher Mining, renamed Cipher Digital in 2026, with signed long-term data center leases with Amazon Web Services and with Fluidstack and Google. | $5.43 | -16.13% | $2.5B | - | 25 |
| HUT | Hut 8 Corp. | Significant | Runs a power-first platform with signed 15-year AI data center leases covering 949 MW of IT capacity, while its bitcoin mining is held through a majority-owned arm. | $45.95 | -17.65% | $5.5B | - | 6 |
| BTDR | Bitdeer Technologies Group | Significant | Mining-rig maker and self-miner with a GPU cloud line and a signed 16-year AI data center lease at its Tydal, Norway, site. | $2.96 | -13.34% | $816M | Feb 11, 2027 | 11 |
| HIVE | HIVE Digital Technologies Ltd | Significant | Miner standing up a GPU cloud arm alongside its bitcoin production. | $1.16 | -14.44% | $320M | - | 17 |
| CLSK | CleanSpark, Inc. | Significant | Bitcoin miner that signed a 20-year lease in July 2026 for 175 MW of data center capacity at its Sandersville, Georgia, campus, carrying about $6.6 billion of contracted revenue, with deliveries starting in late 2027. Mining is still the revenue it books today. | $4.89 | -16.27% | $1.3B | Nov 24, 2026 | 30 |
| MARA | MARA Holdings, Inc. | Indirect | One of the largest bitcoin miners. Its majority-owned Exaion unit signed a four-year, $79 million GPU-as-a-service contract in the third quarter of 2026, but MARA has not signed an AI data center lease. | $5.08 | -18.57% | $2.0B | - | 5 |
Landlords
The listed owners of the buildings. AI demand is a growth layer on top of large, diversified colocation businesses, which is exactly why their exposure is graded lower than the operators above.
| Ticker | Company | Exposure | Why it is here | Price | Day | Market cap | Next earnings | Our coverage |
|---|---|---|---|---|---|---|---|---|
| EQIX | Equinix, Inc. | Significant | The largest data center REIT by revenue, at $9.2 billion in 2025. AI capacity is a named growth driver inside a broad colocation and interconnection business. | $968.22 | -4.89% | $95.8B | - | 8 |
| DLR | Digital Realty Trust, Inc. | Significant | Data center landlord leasing capacity to cloud and AI tenants. | $162.46 | -6.64% | $62.5B | - | 12 |
Servers and racks
Someone has to turn pallets of accelerators into working systems. These companies build, integrate, and support the AI servers that fill the buildings, a high-volume, thinner-margin business than the chips inside.
| Ticker | Company | Exposure | Why it is here | Price | Day | Market cap | Next earnings | Our coverage |
|---|---|---|---|---|---|---|---|---|
| SMCI | Super Micro Computer, Inc. | Direct | AI server systems are the business. Ships accelerator racks at volume and rises and falls with that shipment rate. | $21.65 | -9.19% | $14.3B | Feb 2, 2027 | 19 |
| DELL | Dell Technologies Inc. | Significant | One of the largest AI server builders, alongside its PC, storage, and traditional server businesses. | $278.02 | -11.04% | $176.8B | Nov 24, 2026 | 32 |
| HPE | Hewlett Packard Enterprise Comp | Significant | AI systems and supercomputing franchise inside a diversified enterprise hardware company. | $40.91 | -10.07% | $54.3B | Nov 23, 2026 | 13 |
| CLS | Celestica, Inc. | Significant | Builds hyperscale racks and networking hardware as an outsourced manufacturer for the biggest AI buyers. | $201.80 | -16.00% | $25.4B | Jan 27, 2027 | 0 |
Power, cooling, and construction
Every megawatt a data center draws has to be distributed, conditioned, backed up, and turned into heat somewhere. This tier sells what the buildout physically consumes, and it keeps earning on retrofits after the shell is up.
| Ticker | Company | Exposure | Why it is here | Price | Day | Market cap | Next earnings | Our coverage |
|---|---|---|---|---|---|---|---|---|
| VRT | Vertiv Holdings, LLC | Direct | Power distribution and thermal management for data centers is the core business. | $166.24 | -10.14% | $64.0B | - | 25 |
| ETN | Eaton Corporation, PLC | Significant | Electrical equipment giant where data centers are a named growth market, inside a broad industrial portfolio. | $311.33 | -5.82% | $121.0B | Feb 2, 2027 | 6 |
| GEV | GE Vernova Inc. | Significant | Gas turbines and grid equipment. Data center load growth is a named driver of its order book on both the generation and transmission side. | $699.33 | -7.53% | $185.1B | Jan 27, 2027 | 26 |
| MOD | Modexus Solutions, Inc. | Direct | Thermal management company whose Data Centers segment is its largest, with sales up 95% in the September 2026 quarter. It spun off its Performance Technologies business on October 1, 2026. | $124.67 | -7.35% | $6.6B | Feb 3, 2027 | 3 |
| PWR | Quanta Services, Inc. | Significant | Builds the electrical infrastructure, substations, transmission, and site power that connect new campuses to the grid. | $559.03 | -4.83% | $84.2B | - | 4 |
| STRL | Sterling Infrastructure, Inc. | Significant | Site development and electrical contractor whose E-Infrastructure segment, which it calls its growth engine, serves data center campuses, semiconductor plants, and advanced manufacturing, and was 79% of revenue in the first nine months of 2026. | $247.92 | -9.27% | $7.6B | - | 1 |
| AGX | Argan, Inc. | Significant | Engineering and construction firm building the gas-fired power plants that increasingly get planned alongside large AI campuses. | $268.94 | -9.10% | $3.8B | Dec 3, 2026 | 12 |
Exposure grades are our own judgment, defined in the methodology below. Market cap is unavailable from our quote source for a small number of symbols and is left blank rather than estimated. “Our coverage” is the number of articles we have published on that company.
An AI data center is a building engineered around one problem: feeding tens of thousands of accelerators with enough electricity and removing the heat they give off. The AI buildout has turned those buildings into one of the largest private construction booms in the United States, and a whole layer of companies now lives on that spending: the GPU clouds that operate the capacity, the landlords that own the shells, the server integrators that fill the racks, and the electrical, cooling, and construction firms that make the sites work.
Who leads
The buildout has different leaders at each layer, and the biggest builders of all are not investable through this theme.
| Company | How to buy it | Position in the market |
|---|---|---|
| The hyperscalers | MSFT, AMZN, GOOGL, META | Build and own more AI data center capacity than anyone else, for their own use. They set the demand every company on this page lives on, but the buildout is a cost to them, not revenue, which is why they are not constituents. |
| CoreWeave | US (CRWV) | The largest of the listed GPU clouds and the bellwether for the operator tier. Its contract terms, customer concentration, and financing costs are the cleanest read on the economics of renting AI compute. |
| Vertiv | US (VRT) | Data center power and cooling equipment is its core business. |
| Equinix and Digital Realty | US (EQIX, DLR) | The two big listed landlords. Slower and steadier than the operators: long leases and investment-grade tenants, with AI demand arriving as a growth layer rather than a transformation. |
What drives this group
The variables that decide whether these companies have a good year. Mechanisms rather than forecasts.
1. Hyperscaler capital spending
A handful of companies, Microsoft, Amazon, Google, Meta, and a few AI labs, fund most of the buildout, and they reset their capital spending plans quarterly.
Why it matters: Every tier of this page is a derivative of those budgets. A guidance change at one hyperscaler reaches the operators, the landlords, the server builders, and the equipment vendors within days, which is why the group trades together on capex headlines even though the businesses differ.
2. Power and interconnection
New sites need generation capacity and a grid connection, and both are queued years out. Permits and interconnection agreements lead construction by long lead times.
Why it matters: Electricity is widely cited as the binding constraint on how fast capacity can be added. A blocked or delayed interconnection does not cancel demand, it relocates it, but the delay compresses the growth rate the whole group is priced on. Our AI bubble index tracks exactly this gap between announced and permitted capacity.
3. Contract quality: announced, signed, effective
The buildout runs on multi-year leases and take-or-pay capacity contracts, announced long before revenue flows, and sometimes contingent on financing or power that has not been secured yet.
Why it matters: Two announcements of the same dollar size can be wildly different assets. Whether a contract is signed or merely announced, whether it is effective or contingent, and how few counterparties stand behind it decide whether a backlog is future revenue or a press release. This is where the operators differ most, and it is the first thing to check on any lease headline.
4. The financing window
Sites are increasingly built with debt raised against those long-term leases, so the marginal project depends on credit markets staying open at workable rates.
Why it matters: Cheap project finance turns announcements into buildings; expensive finance quietly kills the marginal site without a headline. If this cycle ends the way credit cycles usually do, it will show up here first, which is why the financing terms in operator filings deserve more attention than the gigawatt numbers in their press releases.
5. Rack density and the cooling transition
Each accelerator generation draws more power per rack, pushing sites from air cooling toward liquid cooling and heavier electrical distribution.
Why it matters: Density growth means the equipment vendors get paid twice: once on new construction and again retrofitting the existing fleet. It shifts value toward thermal and electrical suppliers even when the number of new buildings disappoints.
Recent SEC filings
| Filed | Ticker | Form | Company |
|---|---|---|---|
| Nov 20, 2026 | CRWV | 144 | CoreWeave, Inc. |
| Nov 20, 2026 | CRWV | 144 | CoreWeave, Inc. |
| Nov 20, 2026 | CRWV | 144 | CoreWeave, Inc. |
| Nov 20, 2026 | CRWV | 144 | CoreWeave, Inc. |
| Nov 20, 2026 | CRWV | 144 | CoreWeave, Inc. |
| Nov 20, 2026 | NBIS | 4 | Nebius Group N.V. — PRIMARY DOCUMENT |
| Nov 20, 2026 | APLD | SCHEDULE 13D/A | Applied Digital Corp. |
| Nov 20, 2026 | APLD | 4 | Applied Digital Corp. — 4 |
| Nov 20, 2026 | GLXY | 4 | Galaxy Digital Inc. — FORM 4 |
| Nov 20, 2026 | GLXY | 4 | Galaxy Digital Inc. — FORM 4 |
| Nov 20, 2026 | GLXY | 4 | Galaxy Digital Inc. — FORM 4 |
| Nov 20, 2026 | BTDR | 4 | Bitdeer Technologies Group |
Our recent coverage
- Jensen Huang's foundation commits $350 million to CoreWeave (CRWV) GPU time for researchers Nov 19, 2026
- Trump disclosure shows four Nvidia (NVDA) buys, plus SpaceX (SPCX) and CoreWeave (CRWV) stock and bonds Nov 19, 2026
- CoreWeave (CRWV) co-founder Brian Venturo buys $30 million of its bonds due 2031 Nov 19, 2026
- Two-thirds of Leopold Aschenbrenner's latest 13F is calls. His SharonAI (SHAZ) and Nebius (NBIS) filings fill in what it leaves out. Nov 16, 2026
- Circular financing, explained: Nvidia (NVDA) guarantees up to $105 billion of OpenAI's leases and owns OpenAI stock Nov 16, 2026
- CoreWeave reportedly seeks $4 billion loan backed by Meta contracts at about 7.5% Nov 13, 2026
- Nebius (NBIS) beat Q3 revenue estimates but now expects 2026 sales at the low end of its range Nov 12, 2026
- CoreWeave (CRWV) jumps 17% premarket after its Q3 beat, lifting Nebius (NBIS), IREN, and Nvidia Nov 11, 2026
How the group traded
Unweighted average of the 28 constituents that returned a quote in the most recent session. This is a read on how the group traded, not an index and not a tracked return series.
How this list is built
The names on this page are graded by how directly that spending reaches their revenue, because the label covers very different businesses. A GPU cloud lives or dies on the buildout. A diversified electrical vendor merely has a strong segment. Treating them as one trade is how people end up owning the wrong risk, and sorting that out is what this page is for.
A company is on this page when AI data center demand is a named, disclosed driver of its business, in one of four roles: operating AI compute capacity, owning and leasing the buildings, integrating the servers and racks inside them, or supplying the power, cooling, and construction that make a site run. We grade how direct that exposure is rather than presenting every name as a pure play, because most of them are not.
- Operators: GPU clouds and AI-hosting companies whose revenue is the capacity itself, including former bitcoin miners that converted their power-secured sites.
- Landlords: listed data center owners leasing space and power to AI tenants.
- Servers and racks: companies building and integrating the AI systems that fill the buildings.
- Power, cooling, and construction: companies supplying the electrical equipment, thermal management, and site engineering the buildout consumes.
Exposure grades
- Direct. The theme is a primary revenue driver. A change in it changes this company's year.
- Significant. Named by the company as a real driver, but a minority of revenue.
- Indirect. Benefits from the theme inside a diversified business. Do not read it as a pure play.
What is deliberately not here
- The hyperscalers, Microsoft, Amazon, Google, and Meta, build and own more AI data center capacity than anyone. They are still not on this list, because they are the demand side: the buildout is capital spending for them, not a revenue line. They are covered on our company pages instead.
- Chipmakers are the other obvious omission. What goes inside the racks is its own supply chain, covered on our HBM memory page and on company pages.
- Power generation, the utilities, nuclear developers, and independent power producers that sell the electricity, is deliberately held out as its own theme. It shares one driver with this page and nothing else.
- Networking and optics vendors sit between our sector themes and will get their own page rather than a token row here.
- Bitcoin miners appear only when AI hosting is a disclosed, material part of the story, graded by whether the contracts are signed or still ambitions. Miners with nothing beyond an announcement are graded Indirect.
Prices, market caps and earnings dates come from our market data provider and refresh through the trading day. Filings come straight from SEC EDGAR. Membership and exposure grades are reviewed by us as companies change what they disclose, and changes are recorded in the page history.
Use this data
Our figures are free to reuse in articles, newsletters, and research with attribution and a link back to this page. How every number is computed: methodology. Current table as a file: download CSV.
Suggested citation: AIStockWire, “AI data center stocks,” aistockwire.com/sectors/ai-data-centers. Questions about the data: support@aistockwire.com.
Related: HBM memory stocks · AI storage stocks · AI bubble index · AI infrastructure backlog tracker · CoreWeave (CRWV) company page · Vertiv (VRT) company page
Frequently asked questions
What are AI data center stocks?
AI data center stocks are the companies paid by the AI buildout: GPU clouds that operate accelerator capacity, listed landlords that own the buildings, server integrators that fill the racks, and the power, cooling, and construction firms that make the sites work. The label covers very different risk profiles, from pure-play operators whose entire revenue is AI capacity to diversified industrials where data centers are one strong segment.
Which stocks benefit from AI data center power demand?
Inside this group, the most direct beneficiaries of the power problem are the electrical and thermal equipment makers, Vertiv, Eaton, GE Vernova, and Modine, plus the builders like Quanta, Sterling, and Argan that construct substations, transmission, and on-site generation. The power producers themselves, the utilities and nuclear developers that sell the electricity, are a separate theme and are left off this page.
Are Microsoft and Amazon AI data center stocks?
Not in the sense this page uses. The hyperscalers build and own more AI data center capacity than anyone, but they build it for themselves: the buildout is capital spending for them, not a revenue line. Owning them is a bet on what the data centers eventually earn, not on the buildout itself.
Are bitcoin miners AI data center stocks?
Some are becoming so, and the test is contracted hosting revenue rather than announcements. Riot has leased 1,151 MW of critical IT capacity across its Rockdale and Corsicana campuses for about $38 billion of contracted revenue, Galaxy leased Helios I, the first 800 MW of approved power at its Helios campus, to CoreWeave, and CleanSpark signed a 20-year lease at its Sandersville, Georgia, campus in July 2026; those are real conversions. TeraWulf, Cipher, Hut 8, Bitdeer, and HIVE have also signed AI hosting or GPU cloud contracts, at different stages of delivery. MARA remains a bitcoin miner first: its majority-owned Exaion unit has a four-year, $79 million GPU-as-a-service contract, but MARA has not signed an AI data center lease, which is why it carries a lower exposure grade on this page.
What is a neocloud?
A neocloud is a company that rents out GPU capacity without being one of the big general-purpose cloud providers. CoreWeave and Nebius are the best-known listed examples, along with WhiteFiber and IREN, a bitcoin miner moving its power-secured sites to AI cloud. Core Scientific and Applied Digital are related but different: their main business is leasing AI data center capacity to tenants such as CoreWeave rather than renting out GPUs. Their common trait is concentration: long contracts with a small number of customers, which is both the appeal and the risk.
Why is electricity the constraint on AI data centers?
Because generation and grid connections take years to add and the queues are long. Chips can be bought and buildings can go up in quarters, but a new site still needs a utility-scale power supply and an interconnection agreement, and those lead construction by years. That is why announced projects concentrate around existing power, why on-site gas generation keeps appearing in project plans, and why our AI bubble index tracks the gap between announced and permitted capacity.
Published for information only. Nothing here is investment advice or a recommendation to buy or sell any security, and membership of this list is not an endorsement of any company on it.