Cushman & Wakefield Releases 2026 Data Center Development Cost Guide, Citing 21% Rise in Per-MW Construction Costs

Cushman & Wakefield’s 2026 Data Center Development Cost Guide shows data center construction costs have increased an average of 21% per MW since the prior edition (Q4 2024). The guide covers U.S. and Canada markets and highlights rising development economics that may pressure project returns and expansion plans.

Analysis

A higher-cost build environment is mainly a supply story, not a demand story. The first-order effect is fewer speculative starts and a longer path to incremental capacity, which improves the bargaining power of incumbent operators with powered shells and ready-to-lease inventory; that is the cleaner read-through for EQIX and DLR than for any consultant marketing the trend.

The second-order winner is the “picks and shovels” layer that monetizes densification and retrofit spend rather than greenfield dirt: power distribution, cooling, and interconnect upgrades should capture more dollars per deployed MW if hyperscalers keep capex budgets intact. That argues for VRT/ETN over 6-18 months, while pure development-fee models and transaction-driven CRE services face a slower pipeline if projects are deferred or re-scoped.

Contrarian view: the market is likely already aware that data center construction is expensive; the real constraint is utility interconnection and permitting, not hard-cost inflation. If financing eases or AI demand remains elastic, higher costs may simply be passed through, muting the downside for new-build economics. The thesis is falsified if preleasing stays strong, cap rates compress, or developers continue announcing starts despite unchanged IRRs.

Key Decisions for Investors

• Long EQIX / short XLRE on a 1-3 month horizon to express tighter supply and relative pricing power; target ~2:1 upside/downside if data-center scarcity keeps widening rent spreads. Falsify if broader REIT yields re-rate lower and EQIX/DLR lag on operating commentary.
• Add VRT on a 6-12 month basis for retrofit/densification exposure rather than new-build volume; prefer call spreads to cap premium if the market is already pricing AI infrastructure upside. Falsify if hyperscaler capex shifts away from power/cooling and into software/networking instead.
• Avoid chasing CWK into strength: this reads more like lead-generation than an earnings catalyst. If you want to fade the move, do it only after checking whether management cites measurable pipeline conversion on the next call.
• Set an alert for any data-center REIT or developer guidance that shows faster-than-expected lease-up despite higher capex; that would invalidate the slowdown thesis and argue for covering shorts in general CRE proxies.