ERCOT paused Batch Zero. Package quality is now an audit survival question.
Texas paused ERCOT's Batch Zero process and ordered an audit of large-load projects. What that means for developers and consultancies, what auditors will check, and what a discrepancy audit costs versus rework.
In August 2026, Texas paused the ERCOT Batch Zero study process and ordered an audit of data-center interconnection projects, with the regulator stating that noncompliant projects should be denied grid access. If you build, model, or review interconnection packages, that sentence re-priced your deliverables overnight. Package quality stopped being a convenience issue and became a compliance artifact with denial attached to it.
The queue math behind the panic
The backdrop is scale. LBNL's 2026 queue research counts roughly 8,200 projects in US interconnection queues: about 1,312 GW of generation and 749 GW of storage waiting. ERCOT's own queue hit roughly 470 GW in July 2026, around 421 percent of its peak-load record. Study capacity is the scarce resource, and a package that forces re-study burns the one thing nobody can manufacture more of.
Where packages actually fail
Almost every failure we find in delivered packages traces to the same root: the model no longer matches the documents it was built from. Concretely:
- Ratings conflicts: a transformer shown at one MVA on the single-line diagram and another in the equipment schedule.
- Topology drift: breakers or buses present in the drawings but missing from the model, or the reverse, because a revision was never propagated.
- Conductor and cable mismatches between the study assumptions and the schedule of record.
- Study staleness: the model reflects the site as designed two revisions ago.
- Blind spots from manual transcription, which is still how most junior engineers build models today.
What an audit-oriented QA pass checks
A package audit that stands up to a regulator reads the pile the way an adversary would. Five checks carry most of the weight:
- Model to SLD consistency: every component in the drawings exists in the model, and vice versa, revision by revision.
- Ratings reconciliation: nameplate values in the model match the equipment schedules and spec sheets of record.
- Topology completeness: connectivity, breaker numbering, and bus arrangements are internally consistent.
- Assumption traceability: every load, impedance, and setting traces to a source document rather than an engineer's memory.
- Revision control: the package states which drawing revisions it was built from, and that statement is verifiable.
The economics of finding it early
A discrepancy audit on a delivered package runs about a week and five figures. A re-study forced by a bad package costs weeks of schedule and, on data-center timelines, real money per month of delay; a mismatch found by the auditor instead of by you adds the cost of credibility. The math that sells QA has never changed: finding the mismatch before submission is always the cheap version.
| Stage | Who finds it | Typical cost |
|---|---|---|
| Internal QA audit before submission | You | One week, fixed price |
| Study rejection and rework cycle | The ISO or consultant | Weeks of schedule, re-study fees |
| Regulator audit finding | ERCOT or the auditing body | Denial risk, legal and re-filing costs |
How GridLens runs the check
GridLens ingests the SLDs, schedules, spec sheets, and prior studies, parses them into a canonical network graph, reconciles every asset across sources, runs physics sanity checks, and emits a discrepancy report with every value linked to its sheet, revision, and coordinate frame. A professional engineer reviews the findings before delivery. The report is the product: it is the document you hand your client, or your regulator, when the question is whether the package matches reality.
When was the last time your delivered model actually matched your SLDs and equipment schedules?
Common questions
What was the ERCOT Batch Zero pause?
In August 2026 Texas paused the ERCOT Batch Zero study process, which grouped large-load data-center interconnection requests, and ordered an audit of those projects after concerns that study inputs did not reflect actual project data. The regulator stated noncompliant projects should be denied grid access, making package data quality a gating compliance issue.
How much does an interconnection package QA audit cost?
A fixed-price discrepancy audit on a delivered package typically runs $8,000 to $15,000 over about a week, PE-reviewed. Full model rebuilds run $15,000 to $40,000 depending on pile size. Both are an order of magnitude below one forced re-study cycle on a large-load project.
Can AI review a power system model without a PE?
The engine can parse, reconcile, and flag inconsistencies at machine speed, but the deliverable that matters carries a professional engineer's review. GridLens separates the two deliberately: the machine produces traceable findings, the PE signs the judgment.