T+3 holiday-weekend consolidation — the moratorium wave crossed from county-cluster to state-scale for the first time (New Mexico legislators announced Thursday July 3 that four Democrats will introduce a statewide first-in-nation ≥ some-MW moratorium bill for the January 2027 session, and Santa Fe County unanimously passed an 18-month freeze June 30 with the threshold dropped ALL THE WAY DOWN to 1 MW — that's the lowest data center threshold anywhere in the US and roughly 75x smaller than ERCOT Batch Zero's 75 MW floor); the QTS Digital Gateway public-notice-defect litigation template from Thursday now has a companion precedent from June — Monterey Park CA voters banned data centers 86-14 by ballot initiative (Measure NDC, June 2 special election), which is the first VOTER-ENACTED permanent municipal ban in the US and cannot be reversed by city council, only by another ballot; Oracle sued the Wisconsin PSC in Ozaukee County (filed June 19) challenging the ~$100M/year financial-security requirement the PSC imposed on BBB-rated hyperscalers in April — first major hyperscaler litigating a state ratepayer-protection rule head-on; PJM 2028/29 capacity auction bidding closes Monday July 7 with results Tuesday July 14 (governor-negotiated $175-$325/MW-day price band still holding), ERCOT Batch Zero Form W Part A is now T-6 to July 10; three new failure modes stacked in one week — statewide moratorium legislation, ballot-initiative bans, and hyperscaler lawsuits against state PSCs.

Why it matters

Three implications. (1) A first-in-nation statewide moratorium enacted January 2027 would remove New Mexico from every hyperscaler siting map through mid-2028 minimum. New Mexico is not a top-10 data center state today but its low power costs, cool nights in the north, existing DOE lab infrastructure (Los Alamos, Sandia), and land availability had made it a candidate for the next-wave siting cycle. The Meta Los Lunas campus (~600 MW, ongoing) is the load-bearing precedent for NM data center capex; a statewide moratorium blocks the next wave. Cliff should build corpus depth on NM state legislative dynamics — sponsor voting records, committee assignments, chamber math (NM Senate: 27 D / 15 R; NM House: 45 D / 25 R — Democrats control both chambers, Gov. Michelle Lujan Grisham is D and term-limited-out in Nov 2026 with a new D or R governor arriving in Jan 2027). If Lujan Grisham's successor is a Democrat, signature is highly likely. If a Republican wins Nov 2026, veto is plausible. That gubernatorial election is now a first-order input to NM data center siting probability. (2) Santa Fe's 1 MW threshold is the design signal to watch. If the statewide bill uses 1 MW, effectively every data center is banned. If it uses 20 MW (following NY), the ban covers most hyperscale but leaves edge/small-colo open. If it uses 100 MW (Santa Fe's original draft), it only affects mega-campuses. Cliff should publish an analysis piece the week the introduced-text becomes public in January 2027, mapping which specific NM sites the various thresholds would cover. That's a productizable deliverable to every hyperscaler evaluating NM. (3) The statewide vs. county-cluster distinction matters for how Cliff models regulatory-risk probability across states. County-cluster moratoriums are short-term (typically 12-18 month freezes with high renewal rate) and geographically scattered — hyperscalers can route around them. Statewide moratoriums are multi-year (typical 24-month freeze plus another 12-24 for regulatory framework development) and geographically total. If NM enacts and 2-3 other states follow (WA and OR have similar political shapes and pending bills), the aggregate US available-siting geography compresses meaningfully. Cliff should build a state-by-state 'moratorium risk score' capturing (a) governor party, (b) chamber math, (c) prior EO or bill history, (d) county-cluster density, (e) hyperscaler industry political spend. Score by end of Q3 and publish.

Why it matters

Three implications. (1) Ballot-initiative risk is now empirically demonstrated. Before June 2, it was a theoretical concern; after June 2, there is a working template. Every California suburban city with an active-initiative culture is now a plausible ban target. Cliff should build a 'ballot initiative risk' data field on every California siting output, and separately catalog which specific cities have (a) initiative process availability, (b) recent successful ballot activism, (c) proximity of any hyperscaler-considered parcel to residential. That's a real product input, not a research artifact. (2) The 86-14 vote share tells us something about the political ceiling for anti-data-center organizing when the specific facility is close to homes. This is not a 55-45 vote; it's 86-14. That means the specific narrative (data center within 500 feet of residential, drawing 50 MW, in a demographic that votes on quality-of-life issues) is essentially unopposable at the ballot. Hyperscalers cannot outspend that political dynamic; they can only site around it. Cliff should tell hyperscaler siting teams: any parcel <1000 feet from residential in an initiative state is now a high-risk siting choice regardless of local zoning. (3) The combination of failure modes 1-4 changes what Cliff sells. Consulting firms today sell qualitative regulatory risk assessments ('here are the concerns in Virginia'). The four-failure-mode framework converts that into a quantitative, per-site composite score with specific data inputs (initiative history, moratorium calendar, notice-compliance record, state legislative composition). That's productizable in a way that qualitative consultant output is not — it becomes a scorable, comparable, benchmarkable API-accessible field on every siting decision.

Why it matters

Three implications. (1) This is the first time a major hyperscaler has litigated a state PSC ratepayer-protection rule head-on rather than negotiating a settlement or absorbing the cost. That signals a strategic shift: hyperscalers had been passively accepting the 2025-2026 wave of rate-class rules because per-facility costs were manageable and litigation was politically expensive. Oracle broke that pattern — probably because Wisconsin's $100M/year figure was large enough per facility to justify the political cost, and because Oracle's Port Washington project economics couldn't absorb it. Expect other hyperscalers to follow: whenever the per-facility financial-security or rate-class cost exceeds roughly $50M/year, litigation becomes economically rational. Cliff should model 'PSC-rule-litigation risk' as a real regulatory input for every state — states with $50M+ potential per-facility rules are litigation-prone, states with smaller per-facility rules are less. (2) The Wisconsin ruling will effectively decide what states can and can't do to protect ratepayers from hyperscaler-driven cost impacts. That has enormous implications for every state's PSC rulemaking pipeline in 2026-2027. Cliff should build a 'PSC rule vulnerability index' scoring each state's current rule against the specific arguments Oracle is making (statutory authority, discrimination, due process). States with rules that survive the Oracle framework are more durable; states with rules that don't are exposed. (3) The credit-rating dimension of the case is under-appreciated. Oracle is BBB. AWS/Amazon parent is BBB+/A-. Microsoft, Google, and Meta are all A+ or better. If the Wisconsin rule survives (or if similar rules elsewhere use credit-rating gating), the effective competitive dynamics shift: A+-rated hyperscalers face no financial-security cost, sub-A hyperscalers face 6-8 figure annual costs per facility. That's a material moat for Google/Microsoft/Meta specifically against Oracle/Anthropic/OpenAI (as AI compute buyers) and against any private-equity-owned data center portfolio (which typically have BBB or lower credit ratings). Cliff should quantify this credit-rating advantage per facility per state and publish — it's a genuinely productizable investor-facing artifact.

Why it matters

Three implications. (1) The six-deadline calendar is the highest-density regulatory arc Cliff has seen this year. Every deadline lands a concrete signal that can be modeled, quantified, and productized. Form W submissions become an ERCOT-visible dataset by Thursday. PJM's clearing price becomes public Tuesday Jul 14. Lakeland's first-reading vote becomes public Monday. Each is a natural moment to publish a specific-and-timely analysis piece. Cliff should draft a batch of six analysis pieces this weekend, one per deadline, and publish each within 24 hours of the corresponding event. (2) The three-track segmentation (developer, local politics, ratepayer-cost) is a genuinely useful product frame. Prior conversations with hyperscalers have blurred these tracks together ('there's regulatory risk in your target state'); separating them gives specific-and-actionable guidance to specific-and-relevant stakeholders in the customer's org. The developer track is for the site-selection team, the local-politics track is for the government-affairs team, the ratepayer-cost track is for the regulatory-affairs / finance team. Cliff should structure its outbound emails accordingly. (3) The PJM 2028/29 clearing price on Tuesday July 14 is the single most-consequential number arriving next week. If clearing is at or near the $325 cap, that validates the 13-governor pre-negotiation model and cements the political floor under future auctions. If clearing is meaningfully below (say $220-$260), that suggests supply-side response is holding and the 13-governor model is over-designed. Either outcome is meaningful for hyperscaler 2028+ capacity cost modeling; Cliff should publish an analysis piece the afternoon of Jul 14 immediately after results drop.