Plain-English version. Over the last 30 days, a narrative has hardened in trade press and political coverage: roughly half of US data center capacity 'planned for 2026' will be cancelled or delayed. The claim comes from a few different sources stacked together: - **Data Center Watch** (an opposition-tracker run by activist groups): claims 75+ projects representing ~$130B in announced capex have been 'successfully blocked' in Q1 2026 alone, with 69+ local government bans enacted as of May 2026. Used as the political headline number. - **Bain & Company supply-chain analysis** (referenced by TechRadar, TechSpot, Futurism): claims nearly half of US 2026 capacity has been pushed because of (a) electrical-equipment supply chain delays (transformers, switchgear, GIS, large transformers from Hitachi/Siemens/GE) and (b) power-securing failures (couldn't get utility commitment). - **Ipsos survey, late-2025 to mid-2026:** 50% of Americans didn't want a data center near their neighborhood late 2025; 70% by mid-2026. The opposition number is moving fast. SemiAnalysis (Dylan Patel's analyst shop, widely followed in the AI infrastructure space) pushed back hard in a June 25 newsletter post titled 'Stop Saying Half of 2026 US Datacenter Capacity Is Canceled.' Their argument: the '50% cancelled' number conflates two very different categories of capacity. Category A is 'announced' capacity — paper announcements of a campus that may or may not have utility allocation, permits, financing, or anchor tenant. Category B is 'under-construction' capacity — projects with shovel in ground and committed delivery dates. SemiAnalysis says under-construction capacity is largely on track for 2026 delivery; what's getting delayed or 'cancelled' is announced-but-not-started capacity, which has always had a high attrition rate (historically 30-50% of announced data center projects never get built; the AI cycle just pushed both the numerator AND the denominator higher). The distinction matters because it determines which narrative is right about the moratorium movement's actual impact on supply: - **Activist narrative:** moratoriums are working; opposition is bending the supply curve; the AI infrastructure buildout is being meaningfully slowed. - **SemiAnalysis pushback:** announced-vs-built attrition was always high; the moratoriums are mostly hitting projects that wouldn't have been built anyway; the marquee under-construction campuses (Kilby 2.67 GW, Stargate, LightHouse) are on track. For Cliff's diligence framework, the distinction matters in a specific way. If activist narrative is right, the diligence stack should weight political-opposition risk very heavily — even greenfield sites with utility commitment can be killed by a county council. If SemiAnalysis is right, the diligence stack should weight supply-chain-and-utility-commitment risk much more heavily than political-opposition risk — the political opposition is mostly hitting projects that already had supply-side risk.
Why it matters
Two implications. (1) The narrative collision is itself a leading indicator. When two credible analyst camps publicly disagree about whether the buildout is slowing, it usually means the data is genuinely mixed at the asset-class level — top-tier hyperscaler campuses moving forward, secondary-market merchant builds stalling. Cliff's segmentation should reflect this: a Project Kilby-tier site (hyperscaler equity + co-located generation + bilateral utility deal) has fundamentally different risk variables than a 50 MW wholesale colo expansion in a non-RTO county. The corpus needs separate tracking for the two tiers; conflating them is what causes the public debate to look unresolvable. (2) The political-vs-supply-chain risk decomposition is exactly the diligence variable hyperscaler in-house teams (Microsoft DC Power Procurement, Google Global Energy, Meta Reality Labs Infra) are now asking external diligence vendors for. They have internal models for the supply-chain side (transformer lead times, utility-commitment letters, PPAs); what they want from external vendors is the political-opposition side (county council composition, recent moratorium adjacent jurisdictions, state-legislative pipeline by jurisdiction). That's a specific GTM read: lead with the political-side diligence in any hyperscaler conversation, because that's the gap they don't already cover internally.
Related filings
NM 4 Democratic lawmakers announced Jul 3 statewide moratorium bill / Reps Chavez Cadena Rubio + Sen Hamblen / January 2027 60-day session / first-in-nation statewide moratorium if passed / paired with microgrid loophole closure + data security / Santa Fe County Jun 30 unanimous 18-month moratorium / 1 MW threshold (lowest in US) / 75x below ERCOT Batch Zero 75 MW / 20x below NY A11560 20 MW / 50x below Lakeland FL 50 MW / 100 MW draft lowered on floor / durability higher than county ordinances / removes state from siting map for duration
Plain-English background on why 'statewide' and '1 MW threshold' are the two most important words in this item. Until this week, every US data center moratorium was either a county ordinance (Linn IA, Washington MD, Prince George's MD, Whitley IN, Clark IN,...
FERC show cause orders Jun 18 / Section 206 / PJM MISO SPP CAISO ISO-NE NYISO / 30-day informational report Jul 20 / 60-day tariff response Aug 17 / 90-day abeyance request Aug 3 / five reform categories / cost-shift prevention is the big one / colocation + BTMG rules / flexible-load transmission services
Plain-English version. On June 18, 2026, the Federal Energy Regulatory Commission (FERC, the federal regulator with jurisdiction over the interstate transmission grid) used Section 206 of the Federal Power Act to issue what are called 'show-cause orders'...
20MW threshold convergence / NY S.10642 + AOC HR + Sanders S.4214 + FERC RM26-4-000 / forming model-bill consensus / sweeps in wholesale colo + hyperscaler leased capacity / not just marquee campuses
Plain-English version. Until June 24, the most common threshold for state and local data-center legislation was either 100 MW+ (hyperscale conventional definition, used by industry groups like AFCOM and Uptime Institute) or no threshold at all (county...
FL moratorium wave / Clay final + Hernando, Lake first reading Jun 23 / Santa Rosa 3-1 to draft / Zephyrhills final / 5-8 FL jurisdictions in 5 weeks / SB 484 backdrop / water-first concern
Plain-English version of what happened. Tuesday, June 23 was the single most active moratorium day Florida has had — three county boards moved separate ordinances forward in a coordinated wave. Clay County (just south of Jacksonville, ~220,000 residents) gave...