Global X Data Center & Digital Infrastructure ETF
Global X (Mirae Asset)
Tracks the Solactive Data Center REITs & Digital Infrastructure Index — data-center REITs plus the semiconductor/memory names that physically enable AI compute (a REIT-plus-silicon mix).
- Concentrated — Top disclosed positions are 50%+ of the disclosed book.
Top holdings
Largest data-center REIT globally
Wholesale / hyperscale data-center REIT
Communications tower REIT
US tower / fiber REIT
Korean memory / HBM maker — AI compute
Memory / HBM
Datacenter optics / transceivers for AI networking
Performance
July 2026: 9.66% price (Jun 30 $30.37 to Jul 31 $27.44, stockanalysis) — July's AI-hardware rout hit the ~46-50% semis sleeve (MU -28.7%, MRVL -37.0%, SMH -17.6%) yet DTCR fell only -9.66%, implying the real-estate half was roughly flat-to-up. Trailing 1Y dropped ~19pp in one month (64.76% Jun 30 NAV to 45.27% Jul 31); YTD ~8-9pp below Sector mix and top-10 weight (67.8%) essentially unchanged; fee unchanged.
Recent moves
Surging on the AI datacenter buildout — hyperscalers projected to spend ~$527-602B capex in 2026 (~75% on AI infra); Equinix posted record Q1 2026 margins and is targeting capacity-doubling by 2029.
Our take
The most direct datacenter-real-estate play, but the ~25% memory/silicon sleeve (SK Hynix, Micron, Marvell) makes it a hybrid REIT-plus-semis bet — that's why it crushed pure-REIT peers, and also why it carries more compute-cycle risk than its name implies.
The interesting question is not whether AI data centers are growing — they are — but whether this wrapper gives clean exposure to that growth.
It does not. Roughly 17% sits in AMT and CCI, tower REITs whose revenue driver is carrier network spend, a mature and decelerating business with no meaningful AI linkage; they are here because the index defines 'digital infrastructure' loosely. Another ~46% is semiconductors that most investors already own via QQQ, SMH or any AI fund — so DTCR's outperformance is largely borrowed beta you may be double-counting. What is genuinely differentiated is the colocation sleeve (EQIX, DLR, NextDC, Keppel DC), about a quarter of assets, and that is the part with the weakest recent contribution because REIT valuations fight the rate curve. Net: 0.50% for a blend where the diversifying piece is small and the returning piece is commoditised. Unlevered, so there is no volatility-decay mechanic — the compounding risk is concentration, not path dependence.
Thesis
DTCR sells itself as the data-center landlord trade, but the portfolio is three different businesses stapled together: colocation REITs (EQIX 13.30% + DLR 11.56%), cell-tower REITs (AMT 9.83% + CCI 6.91%) that lease to wireless carriers and have nothing to do with AI compute, and a ~46% Information Technology sleeve of memory and compute silicon (MU, SK Hynix, MRVL, AMD). The name describes one third of the fund. The record was made by the semis, not the real estate — the single most important fact the fact sheet does not lead with. (Issuer, weights as of Jun 30 2026.)
Market-cap-weighted tracking of the Solactive Data Center REITs & Digital Infrastructure Index — 25 holdings, global (US colocation REITs plus NextDC AU 5.04% and Keppel DC REIT SG). Sector split 51.6% Real Estate / 45.7% Info Tech / 2.8% Comm Services; weights as of Jun 30 2026. Unlevered 1x, no daily reset, no derivatives. Expense ratio 0.50%; inception Oct 27 2020; net assets $2.11B (issuer, Jun 30 2026) to $2.18B (stockanalysis, Jul 21 2026).
Assessment
- Genuinely global colocation access — NextDC (AU, 5.04%) and Keppel DC REIT (SG) are hard for a US retail investor to hold directly.
- Unlevered plain-vanilla index structure: no daily reset, no swaps, no decay drag, no counterparty exposure.
- Liquid for a thematic — ~768k shares average daily volume on ~$2.18B assets (stockanalysis, Jul 21 2026).
- Transparent 25-holding index with published weights; you can see exactly what you own and reproduce it.
- Naming mismatch: 45.7% Information Technology in a fund called 'Data Center & Digital Infrastructure' — the semis sleeve, not the real estate, drove the record.
- ~17% in AMT + CCI tower REITs, a carrier-capex business with no AI-buildout linkage, diluting the thesis buyers think they are buying.
- Top 10 ≈ 67.7% of assets across ~25 holdings (Jul 21 2026) — EQIX/DLR single-name risk nears owning the two stocks directly.
- Heavy overlap with mainstream AI/semi funds (MU, AMD, MRVL, SK Hynix) means the exposure may be redundant alongside QQQ or SMH.
- 0.50% for a 25-name index is priced as a specialist product while ~half the book is large-cap semis available at far lower cost.
Record
Issuer NAV returns to Jun 30 2026: +64.76% 1Y, +32.52% 3Y annualised, +13.77% 5Y annualised, +14.94% since Oct 2020 inception; index +65.96% 1Y, so tracking drag is roughly the fee. The 1Y is a semiconductor result wearing a REIT ticker — memory pricing and AI-accelerator demand carried MU, SK Hynix, MRVL and AMD while rate-sensitive REIT valuations lagged. The cleanest evidence is the trailing-1Y spread against the purer real-estate expression: DTCR +44.34% vs SRVR −5.72% (stockanalysis, Jul 21 2026). The 5Y (+13.77%) is the honest read across a full cycle including the 2022 REIT drawdown — far below trailing 1Y and 3Y. Beta 1.35 to the S&P 500, standard deviation 20.0% (issuer, Jun 30 2026). Headline 1Y ranges ~44–65% depending on as-of date and dividend treatment.
Risks & fit
- Memory-cycle reversal: MU and SK Hynix are commodity-priced; a DRAM/HBM price roll would hit the sleeve that generated the returns.
- Hyperscaler capex is the demand curve for both sleeves — one guidance cut from a handful of buyers reprices most of the book.
- Rate sensitivity in the ~52% REIT half; long-duration cash flows discount badly if the curve backs up.
- Power and grid-interconnect constraints can delay colocation capacity delivery regardless of demand.
- Tenant concentration inside EQIX/DLR: the same hyperscalers are their customers and, increasingly, self-build competitors.
The data-center-real-estate framing fails if forward returns keep tracking semiconductor indices more closely than colocation REITs. Concretely: run a rolling 12-month correlation of DTCR against SMH versus against SRVR or the EQIX/DLR pair. If SMH stays the tighter fit — as the trailing-1Y spread (DTCR +44.34% vs SRVR −5.72%, Jul 21 2026) implies — the fund is functioning as a semis fund with a real-estate hedge and should be measured against semi benchmarks. Conversely, a period where semis fall sharply and DTCR holds up on REIT strength would validate the blended-infrastructure framing.
Neutrally: the exposure suits someone wanting one ticker spanning both the landlords and the silicon of the AI buildout, who accepts ~68% top-10 concentration and 1.35 beta and does not already hold a semiconductor or broad AI fund. Poor fit for pure data-center real-estate exposure (REIT-only peers are cleaner), for REIT-style income (30-day SEC yield 1.19%, semi-annual), or where MU/AMD/MRVL already appear elsewhere.
0.50% (issuer, Jul 2026); peer SRVR 0.49% (stockanalysis, Jul 21 2026). But ~half the book is large-cap semis obtainable near 0.35% (SMH) and the REIT half at 0.08–0.13% — you are paying for the blend, not for access.