For PE & infra capital — capital flow, basis, exit windows.

September's deal volume recovered 6.1% month-over-month to 87 transactions totaling $60.7B across 57.96 GW, but the 42.5% decline in deal value signals a significant repricing of the asset class. The headline number masks divergent capital behavior: a single $35B build-to-suit commitment from Anthropic on September 1st—de facto hyperscaler self-supply—accounts for 57.6% of monthly value, while secondary M&A (Equinix acquiring $10B in KKR-held capacity; Stack purchasing GIC's $5B portfolio) reflects sovereign and mega-cap PE rotating out of held infrastructure. Richland Parish and a single asset designated "One" each commanded 5 GW of the month's transactional activity, with Las Cruces (4.5 GW) and Ashburn (4.3 GW, 161 facilities) consolidating coastal and mid-Atlantic bottleneck premium. The geographic concentration and buyer composition—hyperscaler capex + REIT opportunistic buying from PE exits—indicates capital is bifurcating: long-term-committed build capacity flowing to operators with 20+ year power contracts, while stabilized portfolios trading at compressed multiples to clear 2025–2026 origination overhang.

The basis compression and 42.5% month-over-month value decline, despite transaction count growth, signals that exit multiples for 2024–2025 vintage PE-backed development have normalized sharply downward as grid interconnection timelines extend and power-purchase agreement pricing reflects lower AI datacenter utilization curves than 2024 consensus. Expect October–November to show continued portfolio rationalization from non-core holders, with IRRs on stabilized facilities likely resetting toward 7–8.5% unlevered on the back of 7–8 year exit windows rather than 5–6 year turnover. Capacity under construction remains robust at 369 GW in pipeline, but the velocity of capital deployment is decelerating relative to the magnitude of committed megawatt commitments, creating a 18–24 month lag risk between power availability in secondary markets and hyperscaler absorption, particularly outside Virginia, Texas, and the Pacific Northwest.

The capital verdict this quarter turned decisively against European core markets and equally decisively toward U.S. power-constrained regions. Of 333 markets scored, 222 received AVOID verdicts while only 23 earned BUILD ratings—a 90-to-1 rejection ratio that signals structural, not cyclical, repricing. The five BUILD markets cluster in high-power-availability zones: Midland-Odessa benefits from ERCOT's excess capacity and industrial-scale power access; Upper Peninsula Michigan and Williston, North Dakota tap MISO's renewable surplus and lower congestion; Rural SPP and Cheyenne offer sub-$50/MWh power profiles with zero grid constraints. Conversely, Dublin, London, Amsterdam, Frankfurt, and Manchester—the historically premium EMEA gateway markets—have been classified AVOID across the board, reflecting Europe's persistent power scarcity premium, interconnect saturation, and sub-6% IRR prospects on new build. M&A volume of $2.16 trillion across 903 transactions in 90 days confirms capital is rotating, not retreating: Meta deployed $150 million in three separate tranches, CoreWeave committed $96 million, and the hyperscaler-led cohort dominated deal flow, signaling that only operators with sub-$20/MWh all-in power costs and >25 MW contiguous footprints are clearing capital committees.

This window exposes a two-tier market structure hardening into the next four quarters. Hyperscalers are now explicitly buying power-first rather than location-first, evidenced by the concentration of announced and under-construction capacity (15.8 GW and 33.8 GW respectively) flowing to SPP, ERCOT, and MISO regions rather than traditional coastal metros. Concurrently, traditional REIT and growth-PE sponsors are facing compressed basis multiples in EMEA and margin-constrained returns in developed U.S. markets, forcing them into either consolidation plays (acquiring second-tier operators to improve power contracts and utilization) or geographic pivot strategies toward build-to-suit partnerships with regional power utilities. The verdict distribution—88 CAUTION markets sandwiched between BUILD and AVOID cohorts—represents the arbitrage zone where regional developers with utility relationships and <3-year timelines can still execute at 12–15% unlevered IRRs. Expect M&A to shift from asset acquisitions to power-contract roll-ups through Q4 2026.

Monitor Midland-Odessa's capacity absorption rate and interconnect timeline closely; if hyperscalers begin announcing >500 MW net deals in that market, it signals a structural shift from MISO/SPP premium realization to Permian-basin power-cost dominance across AI workload placement. Simultaneously, track Dublin and London vacancy rates in Q4: any facility operator offering >15% year-over-year rate cuts signals panic selling, which would validate AVOID verdicts and create acquisition entry points at 6–8x EBITDA for consolidators with balance-sheet capacity. Finally, watch announced-stage projects in Rural SPP and Upper Peninsula MI for financing announcements; if hyperscalers or mega-REITs begin equity commitments to greenfield builds in sub-$3 billion-scale projects, it confirms the sub-$20/MWh power thesis has become the binding constraint on all new supply, and legacy operators in PJM and California should expect multi-year margin compression.