Terafab / Grimes TX

Why it matters

Two items. (1) Terafab is a chip-manufacturing fab, not a data center, so it doesn't directly fit Cliff's submission-readiness wedge — but it's load-bearing for ERCOT load-realization scoring and for the firm-equivalent-mw underwrite. A $119B full-buildout chip fab on a former coal plant site in ERCOT North Central adds substantial new electrical load, water draw, and downstream gas-pipeline demand to a load zone Cliff's calculator already models for hyperscaler siting. Three direct read-throughs: (a) Gibbons Creek already has retired-coal-plant transmission interconnection assets in place, which is the second instance this week (Stratos UT via Ruby Pipeline tap, Sopaipilla El Paso via existing trunk infrastructure) where AI-adjacent capex is preferentially landing on stranded retired-fossil-asset sites — this is now a pattern, and 'retired coal plant with intact interconnect rights' is a discrete site-selection input worth modeling formally. (b) Adjacent demand: SpaceX/xAI/Tesla will need *more* AI training capacity to serve the chips Terafab produces, plus support data centers near the fab for testing/validation/IT. There's a high probability that Terafab anchors a downstream 1-2 GW data-center campus on adjacent parcels within 24 months — Cliff customers underwriting College Station / Brazos Valley sites should watch this. (c) The June 3 vote is a discrete event-driven catalyst worth scheduling a follow-up entry for. (2) ERCOT PRS/PCLR vote outcome still needs verification. The May 19-20 TAC vote is the structurally locked decision point regardless of how the May 6 PRS vote went, so Cliff's submission-validator product needs to be feature-complete against either possible PCLR variant (as-filed or as-amended) by close-of-business May 18. Pull the ERCOT market notice and PRS meeting recording overnight; if the PCLR LPC math was modified at PRS, the validator's curtailment-economics module needs to update before TAC.