Cliff Daily — May 5.

Why it matters

Two structural updates worth adding to byog-byong-state-patchwork.md and to a new 'data-center-litigation-tracker.md' page if it wasn't created yesterday: (1) the Hermantown sequence is now the cleanest test case of the litigation-pause mechanism overlapping with the moratorium-pause mechanism — opponents filed suit five days before the council vote was scheduled, the suit didn't formally enjoin the council from acting, but the legal-uncertainty + public-comment combination produced a tabling that's functionally identical to a moratorium without any board needing to vote one in. The 'NDA without council approval' procedural angle is also new — if it survives motion-to-dismiss, that procedural defect cuts at every hyperscaler / city pre-deal in Minnesota state law and is likely to spread to other Midwest states' open-meetings statutes. (2) The $80M / 28-year / $33.5M PV abatement structure is the kind of negotiated tax-incentive package that NC HB 1004 (today's bill section below) would explicitly outlaw — meaning Hermantown is currently negotiating the exact-shape deal that the most aggressive state-level data-center bill would prohibit. Worth tracking: if NC HB 1004 passes, MN legislators have signaled they may follow with a parallel bill (the 2026 MN session ended with no data-center law; carryover is on the table). For Cliff's submission validator product surface: tabling-after-litigation is now a discrete pre-decision risk state that submission-readiness needs to model, distinct from approved / denied / moratorium. The 'sue-then-table' play is replicable and has now succeeded in MN, OH (Wilmington), and IL (Yorkville) — three different states inside ~30 days.

Why it matters

Two new ingestion candidates for the wiki: (1) the Naval Submarine Base Kings Bay commander letter is the first observable case of an installation commander formally intervening in a hyperscale data-center entitlement decision — distinct from the Pentagon's general data-center-near-base policy guidance. The Kings Bay basis (aquifer / electrical / ecosystem) is overlapping but distinct from typical NIMBY arguments and is loadbearing because installation-commander objections route directly to Office of the Secretary of Defense and to the relevant congressional delegation, which has a much harder political ceiling than local opposition. Open question for a new entity page or a section on existing federal-permitting-council-fast-41.md: how many of the 380+ existing US military installations are within 50 miles of an active hyperscale data-center entitlement, and what's the precedent for installation-commander-veto-by-letter? This is a federal-overlay risk vector the de-rate calculator doesn't currently model. (2) The Tovey withdrawal collapses the moratorium-vote political economy — Camden County and Kingsland may proceed with the 9-month / parallel moratoria anyway, but with no live application the political case for the moratorium weakens (no acute threat to defend against), which means today's Camden vote is now actually a leading indicator: if it passes despite no active proposal, the local-moratorium pattern has fully decoupled from any specific developer threat and has become a baseline-prophylactic posture, which is a much more durable de-rate input than the application-specific moratoria covered through April. If it fails, the moratorium-as-defense-mechanism interpretation holds. Either outcome is informative for the byog-byong-state-patchwork.md GA section.

Why it matters

This is a structurally distinct legislative template from anything in the VA / TX / OH / VA cluster currently tracked in virginia-data-center-legislation-2026.md, ercot-puc-58317-sb6-implementation-workshop.md, or byog-byong-state-patchwork.md — and worth opening a dedicated north-carolina-hb-1004.md wiki page. Three load-bearing observations: (1) the 25% on-site clean-generation mandate is the first time a state has written the BTM share into statute. Until now, BTM share has been negotiated bilaterally between developer and utility (the Fleet Storey / Joule Utah / Hyperion Louisiana stack) or driven by ERCOT/PJM market design (the FERC ER26-1088 + ERCOT BYOG patchwork). HB 1004 codifies it as a precondition of the load itself — that's a different legal posture entirely. If 'clean energy' is interpreted narrowly to exclude natural gas (the bill text is not yet final on that question), it eliminates the transitional-BTM template that Fleet just deployed in Nevada and forces solar / battery / SMR / fuel cell as the only viable on-site stack. The fuel-cell language in particular is the read-closely item — Bloom solid-oxide fuel cells running on natural gas may or may not qualify under NC's clean-energy framework, and that determination is the real product-relevant dial. (2) The ban on local tax incentives is the kill-shot for the negotiated-tax-abatement playbook running concurrently in Hermantown ($80M / 28-year, top section above), in Mount Pleasant WI (yesterday's news), and in the typical Henrico / Loudoun VA template. If NC HB 1004 passes, it forces hyperscalers in NC to compete on operating economics rather than tax arbitrage — a structural change to data-center site-selection economics that Cliff's de-rate calculator regulatory-risk module should treat as an entirely separate state-level risk axis. (3) NC is the canary because it has both the existing hyperscaler pipeline (Google, Apple, Meta in Lenoir / Maiden / Forest City clusters) AND the political composition (Democratic governor + Republican legislature) that historically passes consumer-protection bills. If HB 1004 clears committee in 2026, the template is highly portable to OH / GA / IN / SC. Specifically watch the next-30-day committee referral in the Senate Commerce or Public Utilities committee — that's where the on-site-generation share gets tested.

Why it matters

Two pieces of signal — both worth adding to the financing / capital-markets layer of competitive-landscape-by-layer.md and to a new strategy/green-loan-data-center-financing.md page if the pattern repeats. (1) Per-MW debt: $1.5B / 114 MW = $13.16M/MW of senior secured construction debt for fully-leased hyperscale shells in NoVa. That's an empirical anchor for the senior-debt component of the 'firm-equivalent MW underwrite' frame in firm-equivalent-mw-underwrite.md, and notably high relative to the $7-9M/MW that consensus underwriting implies for fully-leased Tier III hyperscale shells. The 60-70% LTV implied here suggests Apterra / GIP / ING are pricing in a significant premium above replacement cost — meaning fully-pre-leased NoVa shells are now priced like infrastructure assets rather than real-estate assets, and the capital structure is increasingly bifurcated between pre-leased (infra-style) and speculative (real-estate-style) financing tiers. (2) Green Loan structuring is the more interesting signal. The Green Loan Principles framework requires the borrower to (a) identify Use of Proceeds tied to environmental categories, (b) describe a Process for Project Evaluation and Selection, (c) ring-fence Management of Proceeds, (d) provide ongoing Reporting on environmental KPIs. For a data-center construction loan, that's a meaningfully harder commitment than a vanilla green-bond gesture — EdgeCore had to justify AS01 specifically as a Green Loan project, which means the design-stage embodied-carbon, water-usage, and operational-PUE / WUE metrics are now contractually load-bearing for debt service. This is the first observable case of green-financing covenants becoming a constraint on data-center design rather than a marketing layer on top of standard financing. If the pattern spreads to other Partners Group / GIP / ING-tier financings, every hyperscale design package suddenly has a financing-covenant-driven KPI module that submission-readiness needs to model alongside the regulatory module. Worth a 30-minute pull-through of the 2025 EdgeCore Green Finance Framework PDF (referenced in the press release) to see what the actual covenants are and which design dials are constrained.

Why it matters

Tomorrow's vote outcome (verifiable by Wednesday morning) is the next observable point on the moratorium-cadence map and worth tracking specifically because Sedgwick is the largest county west of the Mississippi to be on its second extension — most county moratoria covered in byog-byong-state-patchwork.md run a single 6-12 month period and either resolve into a permanent ordinance or sunset. If Sedgwick extends a second time, the rolling-90-day moratorium pattern (3-4 month chunks rather than 6-12 month chunks) is the new template for any county that wants to keep optionality without committing to a full ordinance redesign. That has implications for how the de-rate calculator scores moratorium duration as a regulatory-risk input — discrete 6-12 month periods are easier to underwrite around than rolling-90-day windows, which create permanent application-pipeline freeze risk. Wichita is also adjacent to several active large-load interconnection prospects in SPP territory; SPP's HILLGA / LLRIS proceeding (in byog-byong-federal-dockets.md) is the natural FERC-side hedge against the Sedgwick moratorium pattern. Watch tomorrow.

Why it matters

The King George case is structurally the inverse of the Compass / PW Digital Gateway suit covered yesterday: in PW Digital Gateway, residents sued the county and won a procedural-defect ruling that forced the developer to walk; in King George, the developer (AWS) is suing the county to enforce a prior approval that subsequent leadership wants to renegotiate. Both are Virginia-state cases, both turn on procedural validity of prior county action, but the political-economy outcome is opposite: PW Digital Gateway shows that local opposition can void approvals; King George shows that hyperscalers will sue to prevent approval reversal. Worth adding to ai-paperwork-automation-market-research.md as the canonical 'developer-side enforcement' precedent — a different path than the 'opposition-side voiding' path of PW. The Virginia Supreme Court (already taking the QTS PW Digital Gateway appeal per yesterday's coverage) will eventually need to rule on the procedural-validity standard that governs both cases; whichever case lands first sets precedent for the other. For the de-rate calculator's regulatory-risk module: VA-county political-leadership-change is now a discrete risk input — the King George precedent tells underwriters that a board-flip can void a prior performance agreement and force years of litigation, even with a fully-executed 2023 deal in hand. Diligence has to extend backward to map the political-flip risk of every county in the Fredericksburg / Loudoun / Prince William / Henrico cluster, and the underwriting math has to discount accordingly.