Cliff Daily — May 2.

Why it matters

This is the supply-side companion to yesterday's $725B hyperscaler capex sweep. The 51 GW contracted figure is the largest single-utility data center book in the country and the cleanest validation that hyperscaler capex is landing physically in PJM-East. Contract-stack discipline matters: SELOA → CLOA → ESA is a maturity ladder, and ESA carries take-or-pay — i.e., Dominion is not absorbing speculative load. Useful for the firm-equivalent-mw-underwrite.md framework: the SELOA / CLOA / ESA progression is exactly the kind of contract maturity tagging the de-rate calculator needs to model — a 30 GW SELOA pipeline doesn't underwrite the same way as a 10 GW ESA book. Add a row to virginia-data-center-legislation-2026.md showing that 51 GW contracted is the denominator the legislature is fighting over in the HB 1393 / SB 253 cost-shift mechanism (next item). Also: Dominion's contracted-capacity disclosure cadence (quarterly, with ~5% QoQ growth) is something the regulatory-knowledge-graph should ingest as a structured-data feed — every other utility serving hyperscalers is going to be pressured to disclose at the same granularity by mid-2026.

Why it matters

Direct update to strategy/virginia-data-center-legislation-2026.md, which as of Apr 28 had this as an 'open question' — the question is now resolved on the legislature side. Three scenarios remain for May 22: sign as-passed (the maximalist VA cost-shift template), veto (politically costly given residential savings), or no action (still enactment). The Apr 30 K92/WSLC reporting that Spanberger is publicly defending Google data center deals amid Roanoke protests suggests veto is unlikely; the realistic question is whether she signs or lets it become law without signature. Combine with the Dominion Q1 ESA disclosure: the 10.4 GW already at ESA stage *predates* the new cost allocation, so the cost-shift effectively applies to the ~40 GW SELOA/CLOA pipeline coming behind it. That's the population the regulatory-knowledge-graph wedge can immediately price — what's the $/MW lifetime cost difference for a project whose ESA signs after vs before whatever effective date Spanberger lands on. This is a calendar-driven underwriting question and exactly the kind of unstructured-text-to-structured-data work the air-permit case study scaffolding can extend to. Also pair with the Spanberger bill-disposition fight on HB 507 (Tier-4 generator mandate) — same governor, same May 22 deadline, same political dynamic; whether she splits the disposition is a tell on her data-center stance.

Why it matters

Add to the open-investigations queue with a watch on the Supreme Court's discretionary review decision (months out, typically). The legal theory is the leading edge of the local-zoning vs state-policy fight — a state-level Supreme Court reinstatement would reset the playbook for hostile-county hyperscaler sites; refusal would harden the precedent that local zoning challenges can nullify approved data center campuses even years post-approval. This is exactly the site readiness-risk surface the regulatory-knowledge-graph + approval-outcomes corpus is positioned to track. The fact that one of two co-developers (Compass) already walked is itself underwriting-relevant: it confirms 'litigation overhang' is now a real input to data center site valuation, not just zoning approval. Useful counterweight to the Dominion 51 GW contracted figure — load is contracted but the underlying buildable site supply is being eroded by exactly this kind of ruling. The two numbers together tell the supply-demand story better than either does alone.

Why it matters

Reverses the storyline strategy/2026_04_16_how_to_sell_to_people_in_the_datacenter_space.txt has logged from the Apr 14 Reuters reporting that Maine had become the 'first state to approve a moratorium.' That entry should be updated: the moratorium passed the legislature, was vetoed, and the override failed — Maine is *not* a moratorium state as of Apr 29. The lesson is more nuanced than 'state-level moratoria are now real' (Cliff's prior framing): even when one passes, the political economy of a single anchor project ($550M / Jay) can flip an executive veto, and supermajority overrides remain hard. For the regulatory-knowledge-graph, this means moratorium status needs an executive-disposition field, not just legislative passage. Also note: Cassville WI (May 1) just passed a 2-year town-level moratorium — the local-vs-state pattern is what's actually emerging, not a cascade of state-level bans. Update byog-byong-state-patchwork.md to reflect Maine = no statewide moratorium; track local moratorium count separately (~63 local actions according to Robert Bryce's compile, ~54 already passed).

Why it matters

Direct input to hyperscaler-lease-vs-build.md and bound to BTM-product-surface.md. The structural lesson: even the most aggressive AI buildout in the world has retreated from owning data centers to leasing capacity from purpose-built operators (Oracle, Crusoe, Related Digital). That validates the 'site readiness platform serving the dev-and-lease ecosystem' positioning over an 'AI hyperscaler tools' positioning — Cliff's customers are increasingly the developers/operators, not the AI companies themselves. The Saline financing is also a useful data point on the capital-stack reality: PIMCO had to anchor $10B because US banks retreated on AI-data-center risk; that means the capital available to fund $5B+ site readiness risk is constrained, which sharpens the 'underwrite the site, not just finance it' wedge. Add the financing-anchor pattern (PIMCO/Blackstone equity + insurance-style fixed-rate debt) to the developer-financing layer of the wiki — Cliff's de-rate calculator output is feeding into capital-markets diligence, not just utility filings.

Why it matters

The Dominion-Amazon SMR collaboration is in PJM-East, where the 51 GW contracted figure (top item) is concentrated and where the HB 1393 / SB 253 cost-shift fight is happening simultaneously. Three regulatory tracks now intersect on the same physical territory: (1) cost-allocation rule, (2) generator-emissions Tier-4 BACT presumption (yesterday's lead), (3) SMR siting/licensing. This is exactly the multi-axis site readiness-risk surface where the air-permit / interconnection / siting integration matters most. For the wiki: open a new entry tracking the Dominion-Amazon 5 GW SMR collaboration with milestones and link from interruptible-load-derate.md (because 300 MW SMR + curtailable load is a different denominator than 300 MW gas + curtailable load). The fact that NuScale rallied 10% on a competitor (X-energy) deal is a rising-tide signal for the SMR-as-data-center-primary-power thesis; worth a watch on the Q1 SMR developer earnings cycle (Oklo, NuScale next week) for any Cliff-relevant disclosures on data center customer pipeline.

Why it matters

Three signals worth pulling out: (1) the deal is now structurally coupled to DTE's pending Michigan rate case (yesterday's coverage of the 'ransom note' tying ratepayer relief to The Barn's energization by 2027) — that means the Saline financing close is a precondition for DTE's rate freeze pitch to the MPSC. The capital markets just resolved the financing-uncertainty axis; the only remaining uncertainty is regulatory + siting. (2) Oracle-financed BTM battery storage is a structural template for the BTM-product-surface.md page — anchor tenant funds the storage layer, host utility supplies grid power, developer owns the campus. That's three-way risk allocation Cliff's BTM economics calculator can model directly. (3) PIMCO-anchored fixed-rate debt at this scale is the new normal for data center site readiness finance and it's a different counterparty than what most site readiness consultants are used to selling diligence into. The capital-markets channel deserves its own GTM lane in strategy/outreach-channel-playbook.md — insurance-anchored debt financiers need exactly the kind of structured regulatory risk diligence Cliff produces.