Real Estate Is Up 13%. The Data-Center REITs Powering AI Are Up 36%.

The Vanguard Real Estate ETF (VNQ) holds 159 positions and $38.2B AUM with a 0.13% expense ratio and a 3.57% dividend yield, but the AI infrastructure sleeve is only a low double-digit share of the fund. By contrast, the Global X Data Center & Digital Infrastructure ETF (DTCR) is concentrated in data-center/tower exposure (e.g., Equinix 13.80%, Digital Realty 12.61%, American Tower 12.25%) and related AI/compute beneficiaries, driving far stronger performance: DTCR is up 36% YTD and 56.52% over the past year vs VNQ’s 12.87% one-year gain. The tradeoff is higher cost (0.50% fee), smaller scale ($1.22B AUM), and recent underperformance (DTCR down 7.37% over the past month vs VNQ up 2.32%), plus capital-application/semiconductor-driven risk rather than traditional REIT income.

Analysis

This is less a broad-real-estate call than a factor-split: passive REIT baskets are structurally forced to own too much rate-sensitive, low-growth property beta and too little of the scarce capacity story. The real edge is in identifying where AI capex converts into pricing power and recurring contracted cash flow; that still looks strongest in digital infrastructure names with bottleneck assets and balance-sheet flexibility, not in broad landlord exposure. The second-order effect is that capital may continue to migrate out of yield-only REIT allocations into higher-growth infrastructure sleeves, compressing VNQ’s relative multiple while widening dispersion inside REITs.

The trade is not cleanly one-way. In the next 1-3 months, the biggest risk to the crowded AI-infrastructure trade is not demand destruction but multiple compression if hyperscaler capex slows even modestly or if investors rotate back to defensives on falling rates. Over 6-18 months, the key question is whether incremental supply in data centers and power interconnects erodes occupancy and lease economics; if it does, the market will punish the pure plays first, while tower assets with less AI hype can look more resilient. That makes AMT and CCI worth watching as a lower-beta way to express the infrastructure theme if data-center valuations get stretched.

Contrarian takeaway: the market may be underestimating how much overlap exists between the AI basket and existing tech holdings. For investors already long NVDA/AVGO/AMD, DTCR can become redundant and the marginal upside from reallocation may be smaller than advertised. The more interesting inefficiency may be a relative-value rotation within listed real estate rather than a full sector switch: own the bottlenecked digital assets, fade the broad basket, and be ready to reverse if rates fall sharply or AI capex guides down.

Key Decisions for Investors

• Initiate a pair trade: long DTCR / short VNQ for 1-3 months, targeting continued dispersion in AI-linked REITs versus broad real estate; cut if broad REITs outperform DTCR by ~5% from here or if hyperscaler capex commentary softens.
• For accounts already long NVDA/AVGO/AMD, avoid doubling down via DTCR; instead buy EQIX or AMT directly on pullbacks to reduce semiconductor overlap and keep the exposure anchored to recurring infrastructure cash flows.
• Watchlist: add CCI as a lower-volatility beneficiary if tower spending accelerates outside pure data-center buildouts; it is the cleaner way to express edge-network demand if power constraints cap new data-center supply.
• Reduce or hedge VNQ if the thesis is AI-infrastructure participation rather than income. VNQ is better treated as a core yield vehicle, not the preferred expression of the current capex cycle.