Potential AI slowdown is not ‘end of the world’ for data center real estate, says Digital Realty CEO

AI could drive roughly 70% of global data-center capacity demand by 2030, requiring nearly $7 trillion in total capital spending, including an estimated $3 trillion of real-estate investment over the next five years. Digital Realty and Equinix shares fell after major AI developers called for a slower pace of advancement, but Digital Realty argues cloud growth, inference adoption and supply-constrained core markets should sustain demand. Digital Realty has expanded its development pipeline to $20 billion from $10 billion at year-end 2023 and says it has strengthened liquidity, reduced leverage and added private-capital and joint-venture funding options.

Analysis

The key underwriting variable for DLR and EQIX is not AI-model release cadence but contracted power availability in constrained metros. A moderation in frontier-model training could actually improve tenant mix: hyperscalers may redirect scarce energized capacity toward commercial cloud and inference workloads, which are broader-based and potentially less vulnerable to a single customer’s capex cycle. DLR’s larger development exposure offers greater upside to stabilized rents if absorption remains intact, but also creates more funding and lease-up sensitivity than EQIX’s interconnection-heavy, more mature footprint.

Near-term equity weakness should be evaluated against private-market clearing prices rather than AI sentiment. BX, KKR and BLK benefit if institutional allocations continue shifting toward digital infrastructure vehicles, but the second-order risk is that abundant private capital bids down development yields and raises the cost of winning assets. JLL is the cleaner advisory/transaction-volume expression if capital formation converts into site selection, financing and asset-sales activity; its earnings response will lag initial fund-raising by roughly 1-3 quarters.

The contrarian risk is power, not demand. Grid interconnection delays, transformer shortages, and local permitting constraints can prevent announced pipelines from becoming revenue-generating inventory, while forcing developers to carry construction costs longer. A sustained decline in hyperscaler capex guidance or evidence that pre-leasing is being deferred—not merely AI training delayed—would invalidate the capacity-tightness thesis over the next 1-3 months. Over 6-18 months, rising rates or tighter credit spreads reversing could pressure REIT NAVs and private-infrastructure valuations even if utilization stays high.

Key Decisions for Investors

• Prefer long EQIX versus short DLR over the next 3-6 months: EQIX has lower development/funding beta and greater recurring interconnection economics, while DLR carries more execution exposure on its construction pipeline. Reassess if DLR reports accelerating pre-leasing with development yields materially above its incremental cost of capital.
• Use any further AI-headline-driven DLR/EQIX selloff as a staged-entry opportunity only after checking quarterly bookings, renewal spreads, committed power, and hyperscaler capex guidance; absent those datapoints, treat this as a watch item rather than a directional recommendation.
• Long JLL as a 6-12 month second-order digital-infrastructure capital-markets trade, sized modestly given cyclical brokerage exposure. Thesis requires growth in data-center financing/transaction mandates; cut if broader CRE transaction activity or corporate services backlog deteriorates.
• Avoid chasing BX, KKR and BLK solely on data-center fundraising. Their upside depends on fee-bearing AUM deployment at attractive yields, while aggressive competition for powered land could compress returns before management-fee benefits are reflected in earnings.