Up to half of planned US data centres face delays or cancellation, Kimmeridge says

Kimmeridge warns that as many as half of proposed US data centers could face delay or cancellation due to political backlash and the challenge of building physical infrastructure. The firm expects this to reduce forecasts for US gas demand, given that most gas demand is for Europe-bound supply. Overall, the outlook shifts toward weaker incremental demand for power and gas-linked infrastructure.

Analysis

This is less a clean “AI demand goes away” story than a timing reset, and the market mechanism is in gas forwards, not spot. If a meaningful share of proposed load slips, the first hit lands on names that have been underwriting multi-year demand growth assumptions: US gas producers, LNG exporters, and midstream capacity stories built on incremental baseload pull. The second-order winner is the buyer side of the Atlantic gas trade — lower implied utilization should ease marginal pricing pressure for European importers and energy-intensive industrials.

The most vulnerable equities are the ones where the equity story is predicated on an ever-steeper gas takeaway curve: LNG, EQT, AR, KMI, WMB, and to a lesser extent power-stack beneficiaries that had been leaning on data-center load growth to justify higher long-run capex. If project delays are driven by permitting and local backlash, the damage is not just volume; it also lengthens the payback period on new pipelines, compression, and LNG feedgas expansions, which tends to compress multiples before it shows up in earnings.

The key risk to a bearish gas call is that load does not disappear — it relocates. Hyperscalers can re-site to better-powered markets or shift to behind-the-meter generation, which would preserve gas burn while bypassing the political bottleneck. That makes this a 1-3 month forecast-revision trade rather than a 6-18 month structural call unless we see repeated cancellations across multiple regions and power buyers start cutting capex guidance.

Consensus is probably overestimating how linear the AI power buildout will be, but underestimating how quickly capital can pivot once permits or grid access become binding. That argues for being selective: short the gas-demand narrative, not the entire AI infrastructure complex. The cleanest contrarian read is that existing operators with secured power and interconnects could take share from speculative greenfield projects, while the real loser is the pipeline of projects still trying to win permits and power contracts.

Key Decisions for Investors

• Tactically short UNG or buy KOLD on any weather-driven gas rally; use a 1-3 month horizon for forecast revisions and storage resets, and cover if Henry Hub re-prices above the recent range on a sustained basis.
• Trim or hedge LNG, EQT, and WMB exposure until we see evidence that proposed data-center load is being converted into signed power contracts; the upside case depends on real feedgas demand, not press-release demand.
• Pair trade: long European gas-sensitive equities/consumers (e.g., SHEL or TTE as lower-cost energy beneficiaries) vs. short US gas producers (EQT/AR) if TTF/JKM soften and US LNG feedgas estimates start coming down.
• Avoid adding to data-center infrastructure names that rely on greenfield power access until interconnection and permitting data confirm the buildout; the risk is a valuation air pocket if project timelines slip again.
• Set an alert on US LNG feedgas and power-queue commentary over the next 4-8 weeks; if hyperscalers announce on-site generation or alternate sites, that would falsify the bearish gas-demand thesis and require covering shorts.