How to Invest in Data Centers

Last updated June 2026

Short answer

To invest in data centers, pick a way to get exposure to the physical backbone of the AI build-out: data-center REITs that own the buildings, power and cooling and equipment suppliers, the chip and infrastructure names that fill the racks, sector or thematic ETFs, or a self-built thematic basket across those layers. Each is best for a different job and carries different trade-offs, and the theme as a whole carries real risks (capex cycles, power constraints, and concentration). Size it as one slice of a diversified portfolio. Walnut is an AI investing assistant that can build a data-center basket you approve; it is not an investment adviser.

“Data centers” sounds like a single trade, but there is no one data-center stock. What people mean is the physical stack behind AI: the buildings full of racks, the power feeding them, the cooling keeping them stable, and the chips doing the work. Each of those layers is a different way to invest, with a different profile. This guide covers the main routes (REITs, power and cooling and equipment suppliers, chip and infrastructure names, sector ETFs, and a thematic basket), describes each on the same fields, and is honest about the risks and about where Walnut fits as one option among several.

Why data centers are an AI-era theme

Data centers used to be a quiet infrastructure category. The AI cycle changed that. Training and serving modern models happens inside these facilities, so the build-out that follows AI demand shows up as a chain of physical needs, and each link is investable:

Compute. More AI workloads mean more accelerators, processors, and networking gear filling the racks. This is the layer most people picture first.
Power. Dense compute draws a lot of electricity, so the utilities, generators, and electrical gear behind a site matter as much as the chips, and power availability can cap how fast new capacity comes online.
Cooling. All that power throws off heat, so thermal management and cooling systems have become a theme of their own inside the theme.
Real estate. Someone owns and leases the buildings, the floor space, and the connectivity, which is where the data-center REITs come in.

Because the theme spans real estate, utilities, industrials, and semiconductors at once, how you invest in it depends on which layer you want to own. For the broader idea of buying a theme rather than a single stock, see thematic investing.

The ways to invest

Each route below touches a different layer of the stack. None of them is the single right answer; they are different exposures, and many people blend a few. Here is what each one is, who it is best for, and the honest trade-off.

Data-center REITs

Real estate investment trusts that own and lease the buildings themselves, the racks, floor space, power, and connectivity that tenants rent. They are the most direct listed proxy for data-center real estate and typically pay dividends as REITs.

Layer: Real estate.
Best for: Direct exposure to the physical facilities and their lease economics, with income from dividends.
The trade-off: Sensitive to interest rates and financing costs, and their fortunes ride on leasing demand and how quickly new capacity is absorbed.

Power, cooling, and equipment suppliers

Companies that supply the electrical gear, generators, transformers, thermal management, and cooling systems a dense compute site needs. As racks draw more power, the picks-and-shovels layer that keeps them running and cool becomes a theme of its own.

Layer: Infrastructure.
Best for: Exposure to the power and cooling bottleneck rather than the chips or the buildings.
The trade-off: These are often diversified industrials, so data centers may be only part of the business, and orders can move with broader capex cycles.

Chip and infrastructure names

The semiconductor and networking companies whose accelerators, processors, and switching gear fill the racks. This is the compute layer most people picture first when they think about the AI build-out driving data-center demand.

Layer: Compute.
Best for: Exposure to the compute demand at the center of the AI story.
The trade-off: Can be volatile and richly valued, and a handful of names carry a lot of the theme, so this layer often concentrates risk rather than spreading it.

Sector and thematic ETFs

Exchange-traded funds that bundle data-center, infrastructure, or adjacent AI names into a single ticker. Some track a data-center or digital-infrastructure index; others are broader technology or infrastructure funds that hold the theme as part of a wider mix.

Layer: Diversified.
Best for: One-ticker, diversified exposure without picking individual companies.
The trade-off: Holdings and weights vary widely between funds, expense ratios apply, and a broad fund may dilute the theme with names that are only loosely related.

A thematic basket

A hand-assembled group of names across the layers (real estate, power and cooling, compute) held at weights you set, so you control exactly which parts of the stack you own instead of accepting a fund’s fixed recipe.

Layer: Your own mix.
Best for: Choosing your own mix across the layers and adjusting the weights to your own view.
The trade-off: You own the research and the maintenance, and a self-built basket can drift toward concentration if you do not watch the weights.

For a closer look at individual names across these layers, see the best data-center stocks rundown, and the data-center power theme for the energy angle specifically.

At a glance

Way to invest	Best for	Trade-off
Data-center REITs	Direct exposure to the physical facilities and their lease economics, with income from dividends	Sensitive to interest rates and financing costs, and their fortunes ride on leasing demand and how quickly new capacity is absorbed.
Power, cooling, and equipment suppliers	Exposure to the power and cooling bottleneck rather than the chips or the buildings	These are often diversified industrials, so data centers may be only part of the business, and orders can move with broader capex cycles.
Chip and infrastructure names	Exposure to the compute demand at the center of the AI story	Can be volatile and richly valued, and a handful of names carry a lot of the theme, so this layer often concentrates risk rather than spreading it.
Sector and thematic ETFs	One-ticker, diversified exposure without picking individual companies	Holdings and weights vary widely between funds, expense ratios apply, and a broad fund may dilute the theme with names that are only loosely related.
A thematic basket	Choosing your own mix across the layers and adjusting the weights to your own view	You own the research and the maintenance, and a self-built basket can drift toward concentration if you do not watch the weights.

The risks to weigh

A theme with strong tailwinds still carries real risks, and the data-center story has a few specific ones worth naming before you buy:

Capex cycles. The build-out is funded by large capital spending that can slow, pause, or overshoot. Demand that looks certain in one year can be revised in the next, and the names most tied to it can move sharply with the cycle.
Power constraints. Energy availability, grid capacity, and siting can cap how fast new capacity comes online. The same power angle that makes the theme interesting is also a bottleneck that can limit growth.
Concentration. A handful of large companies carry much of the theme, so a portfolio that leans into data centers can end up concentrated in a few names or one layer without meaning to.
Valuation and rates. Some names are richly valued on expectations, and the REIT layer is sensitive to interest rates and financing costs, so the theme is not insulated from broader market moves.

Spreading across the layers of the stack can reduce single-name risk, but it does not remove the risk of the theme itself. Being honest about that before you size the position is the point.

Sizing the theme

Data centers are one theme, not a whole portfolio. How large a slice it should be depends on your goals, time horizon, and risk tolerance, which is a personal decision rather than a number anyone can hand you. A few practical habits help:

Decide the size before you buy. Choosing a share of your portfolio you are comfortable with in advance is more useful than reacting after the fact.
Treat it as one slice. Many people hold any single theme as part of a diversified portfolio rather than the whole thing, so a bad year for the theme is not a bad year for everything.
Watch for drift. A theme that runs up can quietly grow into a larger share than you intended, which is a form of concentration you did not choose.
Spread across layers if you want less single-name risk. Owning some real estate, some power and cooling, and some compute is a different bet than owning one chip name.

How Walnut fits

To be upfront, since this is our site: Walnut is one option among the several above, not the number-one way to invest in data centers. It is an AI investing assistant you chat with on the broker you already own. You can ask it about the data-center theme and, if you want, have it build a thematic data-center basket across the layers, which you approve before anything happens.

Walnut connects your existing brokerage through SnapTrade, read-only by default, and frames each holding against the S&P 500 so you can see how the theme is doing relative to the broad market. It is not a data terminal and not a stock picker: it helps you research the theme, assemble a basket you control, and track it so it stays close to your intended mix. Every trade needs your approval, and Walnut is informational and is not an investment adviser. The research can inform your decision, but the decision and any trade are yours.

The bottom line

There is no single data-center stock, so investing in the theme means choosing a layer of the stack: REITs for the buildings, power and cooling and equipment suppliers for the infrastructure, chip and infrastructure names for the compute, ETFs for one-ticker diversification, or a self-built basket across all of them. Each is best for a different job, and the theme carries real risks in capex cycles, power constraints, and concentration. Size it as one slice of a diversified portfolio, be honest about what you are buying at each layer, and be wary of anything promising guaranteed returns. Walnut is one way to research the theme and build a basket you approve. Walnut is not an investment adviser.

Try Walnut on top of your broker

Walnut connects any major US broker in a few clicks, then lets you ask about a theme like data centers and build a basket you approve, with each position framed against the S&P 500. Read-only by default; you approve every trade.

Create a free Walnut account

FAQ

How do you invest in data centers?

There is no single data-center stock, so most people get exposure through one of a few routes: data-center REITs that own the buildings, power and cooling and equipment suppliers, the chip and infrastructure names that fill the racks, sector or thematic ETFs, or a self-built thematic basket across those layers. Which fits depends on how direct and how diversified you want to be. Walnut is not an investment adviser.

Why are data centers an AI-era theme?

Modern AI models are trained and served in data centers, so the build-out that follows the AI cycle shows up as demand for compute, the power to run it, and the cooling to keep it stable. That has turned the physical backbone of the internet, once a quiet infrastructure category, into a theme that touches real estate, utilities, industrials, and semiconductors at once.

What are data-center REITs?

They are real estate investment trusts that own and lease the facilities themselves: the floor space, racks, power, and connectivity that tenants rent. They are the most direct listed way to own data-center real estate and, as REITs, typically pay dividends. Their results ride on leasing demand and financing costs, so they are sensitive to interest rates as well as to how fast new capacity fills up.

Is there a data-center ETF?

Yes. Several ETFs bundle data-center, digital-infrastructure, or broader AI-infrastructure names into a single ticker, and some broad technology or infrastructure funds hold the theme as part of a wider mix. Holdings, weights, and expense ratios vary a lot between funds, so read what a given ETF actually owns before assuming it matches the exposure you want.

What are the risks of investing in data centers?

The main ones are capex cycles (the build-out can slow or overshoot), power constraints (siting and energy availability can cap growth), valuation (some names are richly priced), and concentration (a few large companies carry much of the theme). Interest rates matter for the REIT layer. Spreading across the stack can reduce single-name risk but does not remove the risk of the theme itself.

How much of my portfolio should be in data centers?

That depends on your goals, time horizon, and risk tolerance, and it is a personal decision rather than a number anyone can set for you. Data centers are a single theme, so many people treat any theme as one slice of a diversified portfolio rather than the whole thing. Deciding a size you are comfortable with before you buy is more useful than reacting after. Walnut is not an investment adviser.

Are data-center stocks a good investment?

No one can tell you that, because it depends on price, your time horizon, and your view of the AI build-out, none of which anyone can promise. The theme has real tailwinds in compute demand, but it also carries capex, power, valuation, and concentration risks. The honest answer is to understand what you would be buying at each layer, size it sensibly, and be wary of anything promising guaranteed returns.

What is the difference between data-center REITs and chip stocks?

They sit at different layers of the same stack. REITs own the buildings and earn lease income, so they behave more like real estate and are rate-sensitive. Chip and infrastructure names supply the compute that fills the racks, so they track the AI demand cycle more directly and tend to be more volatile. Owning both spreads exposure across real estate and compute rather than betting on one layer.

Can I build a data-center basket myself?

Yes. You can assemble a group of names across the layers (real estate, power and cooling, compute, and ETFs) at weights you choose and hold them in your own brokerage. The trade-off is that you own the research and the upkeep, and a self-built basket can drift toward concentration if you do not watch the weights. Some tools help you assemble and track a basket so it stays close to your intended mix.

How does Walnut help with a data-center theme?

Walnut is an AI investing assistant you chat with on the broker you already own. You can ask it about the data-center theme, and it can build a thematic data-center basket that you approve before anything happens. It connects your brokerage through SnapTrade, read-only by default, and frames each holding against the S&P 500. Walnut is informational and is not an investment adviser; the decision and any trade are yours.

What is the power angle in the data-center theme?

Dense compute draws a lot of electricity and throws off a lot of heat, so power availability and cooling have become as central to the theme as the chips. That is why some investors look past the data centers themselves to the utilities, generators, transformers, and thermal-management suppliers behind them. Power constraints can also cap how fast new capacity comes online, which is a real risk to the growth story.

Should I buy an ETF or individual data-center stocks?

An ETF gives you one-ticker, diversified exposure and spares you picking companies, at the cost of accepting the fund’s fixed recipe and expense ratio. Individual names or a self-built basket give you control over exactly which layers you own, at the cost of more research, upkeep, and single-name risk. Many people blend the two. The right mix depends on how hands-on you want to be.

Walnut is informational and is not an investment adviser. Products, holdings, and availability change; verify current details before deciding. Nothing on this page is a recommendation to buy, sell, or hold any security or to use any particular product.