PJM EIT / Stargate restructure

Why it matters

Two macro updates. (1) The PJM EIT May 28 deadline is the PJM-side analog to ERCOT's TAC May 19-20 — both are last-stop governance decision points that lock the procedural rule set under which Cliff's submission-validator runs in each market. Cliff's product surface for PJM applicants currently models a 'standard' interconnection timeline; if EIT is approved on May 28 with July 31 effective, the product needs an additional 'EIT-eligible' classification check (does this site qualify for the expedited track) added to the validator before August 1. The criteria are reasonably bounded — co-located generation supporting an identified large load — but eligibility is a binary input that materially changes timeline math. (2) OpenAI's Stargate JV unwind is the first observable case where the hyperscaler-DC financing market structure shifted *back* from joint-venture toward bilateral. The original Stargate JV pooled risk across OpenAI, Oracle, SoftBank, MGX. The new structure pushes site-level risk back onto whichever counterparty signs each bilateral — Oracle holds 4.5 GW / $300B alone now. This matters for Cliff's competitive-landscape analysis on the financing layer (green-loan-data-center-financing.md, suggested two days ago) because it makes single-site debt structures (like Meta's $13B El Paso package) *more* important relative to JV equity, not less. The next $13M/MW debt anchor will land on a bilateral deal, not a JV, and Cliff customers underwriting individual site economics should be modeling counterparty-specific creditworthiness rather than JV-pooled credit. Also worth tracking: 'Stargate paused the UK site' is the second observable AI-DC site cancellation in a major OECD market this quarter (Microsoft's Aberdeen-area pullback being the first), suggesting hyperscalers are pruning marginal-return sites at the same time they're scaling top-tier US sites — barbell, not uniform expansion.