AI Data Center Infrastructure Funds: The Biggest One Owns No Data Centers

The largest fund people reach for when searching for an AI data center infrastructure fund holds zero real estate — no data center, no REIT, and no Real Estate sector line at all. Read from its own Schedule of Investments: its ten biggest positions are software, hyperscalers, chipmakers, Netflix, Tesla and SpaceX, and its index has no category that a data center could fit into.

If you searched for an AI data center infrastructure fund, you were probably looking for a way to own the buildings — the facilities being built to run AI models, and the rent they collect.

The largest fund that answers to that search owns none of them. Not a reduced allocation. Zero.

This page shows that from the fund's own filing, explains why its index could not hold a data center even if it wanted to, and then says what the fund is actually useful for — because it is a perfectly reasonable product that is simply not the thing you searched for.

The fund, and the number that matters

The Global X Artificial Intelligence & Technology ETF (AIQ) launched on 11 May 2018, holds 88 positions, and charges a total expense ratio of 0.68% (fund fact sheet).

Its Schedule of Investments at 31 August 2026 reports total common stock of $10,171,551,799, or 99.9% of net assets (Q3 complete holdings).

Here is the fund by sector. The filing reports sector weights within each country; the fund-level totals below are our summation of those figures.

Our sector total comes to 99.9%, which is exactly what the filing states for common stock — so nothing has been dropped or double-counted.

There is no Real Estate line in that filing. Not a small one. The sector does not appear in any of the fourteen countries the fund holds. We searched the schedule for real estate, for REITs, and by name for Equinix, Digital Realty, American Tower and Crown Castle. None of them is there.

What it holds instead

The ten largest positions, each read off the same filing:

Not one of the ten owns a data center as its business. The right-hand column is our characterisation, not the filing's, and it is the whole argument of this page in one column.

Three things in that table are worth stopping on.

The largest holding is a software company. Palantir at 4.12% is the single biggest position in a fund reached by searching for infrastructure.

The fourth largest is SpaceX. "Space Exploration Technologies, Cl A" is a private company held inside a listed ETF, at 3.32%. Whatever one thinks of that, it is not AI data center infrastructure, and it carries valuation mechanics entirely unlike the rest of the portfolio — a private position has no daily market price to mark against.

Netflix and Tesla are in the top six. Both are substantial buyers of compute. Neither is in the data center supply chain in any sense a landlord would recognise.

Sort the rest and the pattern holds:

Microsoft, Alphabet, Oracle, Amazon, Meta and Tencent are the customers. They sign the leases, or they build their own campuses and need no landlord at all. They sit on the other side of the contract from a data center owner. When a data center REIT discloses tenant concentration risk, these are the tenants it means — the numbers are here.

NVIDIA, TSMC, Samsung, Micron, SK Hynix and Broadcom are the suppliers. Their revenue is the data center's capital expenditure — a different line in a different company's accounts.

So the fund gives you exposure to the demand for AI compute and to the equipment inside the building. It gives you none to the building, the land, the power contract or the rent.

Why the index cannot hold a data center

This is not an oversight, and it is published. The index methodology summary defines the theme as companies "positioned to benefit from the development and utilization of Artificial Intelligence (AI) in their products and services, as well as companies that produce hardware used in AI applied for the analysis of Big Data", and sets out the only two categories that exist:

"Category 1 · Artificial Intelligence Developers – develop AI and use AI in their own products · Artificial Intelligence-as-a-Service ("AIaaS") – provide artificial intelligence capabilities to their customers as a service.

>

Category 2 · Artificial Intelligence Hardware - produce semiconductors, memory storage and other hardware that is utilized for artificial intelligence applications · Quantum Computing - develop quantum computing technology."

Then: "The top 60 companies from Category 1 and the top 25 companies from Category 2 will form the final index" (index methodology summary).

There is no category for infrastructure, real estate, facilities or power. A data center REIT does not develop AI, does not sell AI as a service, does not manufacture semiconductors and does not build quantum computers. It is ineligible by definition. The 0.0% in our table is not a market outcome; it is the rulebook working as designed.

Which also means the 0.0% is durable. A weight that is zero because of drift can come back next quarter. A weight that is zero because the index has no category for it cannot.

The word doing the damage is "infrastructure"

Note what the fund is actually called: Artificial Intelligence & Technology ETF. It does not claim to be an infrastructure fund. The gap between the search and the product is created partly by how the category is talked about, where "AI infrastructure" is used for at least four different things:

A fund can be an honest, well-run product and still be the wrong answer to your question, because your question used a word that covers four things and the product covers two of them.

The sister fund makes the same point from the other direction: a fund with "Data Center" in its name turns out to be 41.7% data centers, 28.2% cell towers and 22.3% semiconductors, for reasons written into its index rulebook — that whole calculation is here.

So what should you actually look at?

If you want the buildings and the rent, the listed route is the data center REITs themselves, held directly or through a fund where they are the majority of the assets rather than a plurality. Start with AI data center REITs for what that gets you and what it dilutes, and data center funds for how to check a fund's composition before buying it.

If you want the AI demand story, this fund is a legitimate way to own it, and the honest comparison is not against a data center REIT. It is against a broad technology index fund, because most of its largest holdings are among the largest companies in the world. Before paying 0.68%, work out how much of this you already own through an S&P 500 or global index fund — for most readers the answer is "most of it". The exception worth separate thought is the private holding: a broad index fund will not give you SpaceX, and if that is the reason you are buying, it is a different decision from the AI one.

If you want a specific building or project, no listed fund offers that at all. That is the private route, with the minimums, the lock-ups and the diligence set out in how to invest in AI data centers and reading an offering circular.

Four checks on any fund sold as AI infrastructure

1. Open the Schedule of Investments and look for a Real Estate sector line. If there is none, the fund owns no buildings. This takes one search of one document.
2. Read the index's category definitions, not its name. A category that does not exist cannot be filled later, which makes a zero weight permanent rather than temporary.
3. Sort the top ten into tenants, suppliers and owners. If tenants and suppliers account for all of it, you are exposed to the demand for data centers and not to the economics of owning one. Do this from the filing and read every line, including the ones carrying footnote markers — that is where we nearly missed the largest holding in this fund, and a private company in fourth place.
4. Measure the overlap with the broad index fund you probably already hold, then decide whether the theme is worth the fee difference every year you hold it.

The general method for all of this, and the naming problem across the whole category, is in data center ETFs and funds.

For the engineering of what these companies build and buy, the free vendor-neutral reference is *The Definitive Guide to AI Data Centers*. They cover how it is built; we cover whether to put money into it.

Nothing here is investment advice, and no fund named on this page is recommended. See how this site is funded.