The tape: mild scarcity, soft gas
On 26 September 2026, ERCOT’s system peak still looked like late summer: Eden Energy’s prior-day recap puts actual load near 78.5 GW in hour ending 17, with Dallas, Austin, and San Antonio in the mid-90s°F and Houston near 90°F. Net load rose into the evening as solar declined – the classic ramp shape. What did not appear was a scarcity-style spike.
Hub-average day-ahead locational marginal prices (LMPs) averaged about $32.96/MWh; real-time averaged about $26.16/MWh, for a day-ahead-minus-real-time (DART) spread near +$6.80/MWh. The strongest real-time hour in that recap was HE 19 at roughly $37.52/MWh – moderate relative to ERCOT’s historical scarcity hours, consistent with renewables plus dispatchable generation covering the ramp without reserve emergency pricing.
Fuel underwrote the soft power tape. EIA Henry Hub spot, as carried on FRED (series DHHNGSP), printed $2.90/MMBtu on several mid-to-late September observations, including 1 September and 22 September 2026 (with neighboring sessions in a narrow band around $2.90–$3.00). Soft gas lowers the marginal cost of combined-cycle and peaking gas that still set many evening prices when wind and solar are insufficient. Mid-September forward-curve commentary similarly tied soft gas and residual mild summer heat to lower near-term ERCOT power estimates – shape context only where cleared ICE strips are unavailable. Soft energy on a hot afternoon is real; for large-load developers it is incomplete.
The queue: energization is not an LMP
In parallel with soft LMPs, ERCOT’s large-load process remains constrained by the Batch Zero verification and audit path under Planning Guide Revision Request (PGRR) 145. ERCOT market notices and PUCT Project No. 59142 materials describe an 3 August 2026 pause of the Batch Zero study process and delayed final classification of large loads, plus a pause on approvals to energize data centers and virtual currency mining facilities of 75 MW or greater until eligibility verification is complete. Medium loads (at least 25 MW and less than 75 MW) and non–data-center large loads are treated differently under those updates – energization pause language focuses on large-load data centers and crypto facilities.
Provisional Batch Zero classifications (base, studied, excluded) went to interconnecting transmission service providers around 3 September 2026. Final inclusion or exclusion waits on verification results. ERCOT has indicated it will file a Batch Zero Eligibility Verification Report and a Community Impact Review Report with the Public Utility Commission of Texas by 10 December 2026, with related open-meeting timing in mid-December. Original PGRR 145 study-completion dates (including an April 2027 study milestone in ERCOT updates) remain subject to timeline impact from the verification workstream.
ERCOT briefings in that docket also cite an illustrative near-term stack: on the order of 17 large-load data center/crypto projects totaling about 6,608 MW of peak demand over five years that have finished ERCOT processes except approval to energize – figures that belong in planning dashboards, not in a claim that 6.6 GW is already consuming. The engineering point for the September tape is narrower: soft LMPs measure energy scarcity today; the energization pause measures interconnection and policy sequencing for loads that have not yet arrived.
Forward premium without pretending the spot market “knows”
When megawatts sit just short of energization, commercial offtakers and generators still price option value on future peak coincidence, congestion, and ancillary needs. That forward premium can coexist with $26–$33 hub averages for three independent reasons:
Timing mismatch. Spot LMPs clear the resources and load that are *online*. Queued AI and crypto load, if and when verified and energized, shifts net load and can reprice evening ramps even if annual average energy stays moderate.
Locational basis. West, North, Houston, and South hubs diverge when transmission binds. A quiet HB_BUSAVG day can still hide West or pocket congestion that matters for a single POI. Late-September indicative and real-time displays routinely show hub dispersion even when system-average prices are calm – basis risk is not cancelled by a soft average.
Flexibility and AS. Batteries and demand response earn in ramps and ancillary products when energy looks “cheap.” Soft energy with rising ancillary or regulation needs is a known ERCOT pattern as solar penetration grows; large flexible loads can either compete with or complement that stack depending on curtailment rights in interconnection agreements.
Developers who read only the September energy tape may underbid hedges and on-site flexibility. Utilities and ERCOT planners who read only queue megawatts may overstate near-term coincident stress. Both errors are expensive.
PJM as a control: quiet energy, loud capacity
The Mid-Atlantic contrast clarifies product design. PJM’s 2026/2027 Base Residual Auction cleared at the FERC-approved cap of $329.17/MW-day (UCAP) across the footprint, procuring on the order of 134,311 MW of unforced capacity (plus Fixed Resource Requirement volumes in PJM’s release math). That compares with $269.92/MW-day for most of the RTO in the prior delivery year’s auction (with BGE and Dominion previously clearing much higher constrained prices that then fell to the common $329.17 cap). PJM estimated wholesale capacity’s retail-bill pass-through in a roughly 1.5–5% year-over-year band for some customers, with possible decreases where zone prices fell.
Capacity at the collar can stay elevated while energy markets print unremarkable shoulder-season LMPs – exactly the split large loads care about when they buy “power” as a bundled retail product. In PJM, resource-adequacy rent shows up in RPM; in ERCOT, adequacy rent is supposed to show up in energy and ancillary scarcity. Soft September ERCOT LMPs therefore do not prove Texas has “solved” large-load growth; they prove that on that operating day reserves and fuel were adequate. PJM’s $329.17 print proves the opposite product can clear scarcity expectations even when daily energy is quiet.
What operators and load should do with this week’s signal
- Do not extrapolate mid-$20s/$30s hub averages into a multi-year PPA price without a scarcity and congestion module tied to verified energization cohorts after December verification filings.
- Separate products: energy, basis (hub-to-POI), ancillary/flexibility, and – outside ERCOT – capacity tags. A soft ERCOT energy day is not a soft PJM capacity year.
- Track Batch Zero process dates (verification reports due around 10 December 2026, with study timeline impacts still to be set out) as commercial critical path, not as market gossip.
- Model gas at ~$2.90 Hub as a current fuel regime, with LNG export and winter storage shocks as upside risks to ERCOT evening heat rates – not as a permanent ceiling.
- Flexibility procurement (batteries, curtailable load, on-site generation) remains the bridge between soft average energy and hard coincident peaks if and when paused megawatts clear energization.
Bottom line
The last days of September 2026 offered a clean price-tape lesson: ERCOT can look cheap on energy while large-load interconnection remains administratively expensive. Day-ahead near $33/MWh and real-time near $26/MWh on 26 September, with Henry Hub near $2.90/MMBtu, describe a well-supplied operating day. Batch Zero’s energization pause for ≥75 MW data centers and crypto facilities describes a system still deciding which queued megawatts are real. PJM’s $329.17/MW-day capacity clear describes a market that socializes adequacy through a different product. Analysts who conflate those three signals will misprice both Texas retail deals and multi-ISO site selection.