The ERCOT Curve Has Broad Shoulders Now

TL;DR

The Texas grid load curve looks nothing like it did five years ago. The sharp 2 PM AC-driven peak has been replaced by a broad afternoon plateau with an evening ramp. Two new flat consumers — hyperscale datacenters and bitcoin miners — run baseload around the clock and have filled in what used to be the overnight trough. Wind dominates the overnight hours at near-zero marginal cost. Solar caps the midday and pushes the peak earlier into the evening. Batteries run the 5–9 PM ramp. ERCOT has hit instantaneous renewable penetration above 70% on multiple days in 2025.¹

Hydrogen, which was supposed to be the next-gen fuel for Texas a decade ago, has effectively died as a grid-generation play. Natural gas dispatches into ERCOT at roughly $30/MWh.² Hydrogen-fueled generation lands closer to $50/MWh on a like-for-like basis, and that gap doesn’t close by chasing electrolyzer scale alone. The reason is physics, not policy.

This post unpacks both shifts because they are connected. The grid that came out of the last five years is a different system than the one we were planning for in 2018, and the materials supply chain feeding it is reorienting accordingly.

§1 — The shoulders, not the peak

The traditional ERCOT load curve had one feature: a sharp afternoon peak driven by residential and commercial air conditioning. Generators were dispatched against that peak. Capacity markets were designed around it. Reliability planning treated the 2–5 PM window as the constraint and everything else as headroom.

Three things changed:

Datacenters arrived as flat 24/7 consumers. Texas now hosts more datacenter capacity under construction than any state except Virginia, with capital expenditure up roughly 3,000% over five years.³ A 100 MW hyperscale facility consumes 100 MW continuously — there is no diurnal pattern to it. Aggregate datacenter draw on ERCOT has crossed several gigawatts of ~constant baseload, and it is still rising fast.

Bitcoin miners arrived as flexible price-takers. Texas became the global capital of grid-connected bitcoin mining around 2021, and the operators run a different pattern: they take cheap power whenever they can find it, which means overnight wind and afternoon shoulders. They curtail aggressively at peak prices because their economics work better as a price-following load than as firm load. The aggregate effect is to fill in the overnight trough and lift the shoulders without contributing to peak stress.

Renewables flooded the middle. Texas wind capacity passed 40 GW in 2025; solar passed 30 GW. Wind blows hardest overnight, solar peaks midday, and the combination caps the midday demand-supply gap. ERCOT prices regularly go negative in the overnight hours when wind output exceeds load.¹

The cumulative result: the 2 PM peak is no longer the binding hour for most days. The evening ramp from roughly 5 PM to 9 PM — when solar comes off and AC load is still high — is the harder hour. That is the window batteries dispatch into. It is also the window gas peakers fight for.

Battery storage on ERCOT is now the marginal generator for the evening ramp on most warm-weather days. The fleet — roughly 14 GW / 23 GWh as of late 2025⁴ — is sized to that specific shoulder. This is why warranty risk on the BESS fleet (covered separately in The ERCOT Battery Warranty Cliff) is not a financial curiosity — it is a reliability question, because the shoulders are where the storage works for its money.

§2 — Why hydrogen didn’t happen in Texas

Ten years ago hydrogen was the consensus next fuel for the Texas grid. Major industrials announced hydrogen hubs. Pipeline companies pitched H₂ as a drop-in replacement for natural gas. The 45V production tax credit in the IRA poured federal money into electrolyzer projects. As of 2025, none of those projects are dispatching meaningful electricity into ERCOT.

The reasons are three layers of physics, in order of severity:

1. Volumetric energy density. By mass, hydrogen carries ~3× the energy of natural gas (120 MJ/kg vs 50 MJ/kg). By volume, the relationship inverts — at equal pressure, H₂ delivers about ⅓ the energy per cubic foot of natural gas. Moving the same MW of fuel through a pipeline requires roughly 3× the volumetric throughput, which existing pipeline geometry and compressor stations cannot deliver. The “blend H₂ into existing NG pipelines” pitch works at low percentages (5–20% by volume) and breaks down above that.⁵

2. Hydrogen embrittlement. Pure H₂ diffuses into the steel of conventional pipelines and pressure vessels, weakening the crystal structure over time. Mitigation requires either polymer-lined pipe, specialty alloys, or running at lower pressures with thicker walls — all of which raise the delivered cost per MWh substantially. The existing US natural gas pipeline network was not built for high H₂ fractions and cannot be cheaply retrofitted.

3. Combustion dynamics in turbines. This is the constraint Scheel’s mention of “thermal qualities” was compressing. Combined-cycle gas turbines installed across Texas were tuned for the flame temperature, flame speed, and acoustic profile of natural gas. Hydrogen burns roughly 7× faster than methane, with a higher adiabatic flame temperature and a tendency to drive acoustic instabilities (combustion “humming”) that crack turbine parts.⁶ Most installed turbines can co-fire 5–30% H₂ without major retrofits. Running 100% H₂ requires new burner cans and sometimes a different turbine architecture. The installed Texas fleet is tuned for natural gas, and turbine OEMs (Mitsubishi, GE Vernova, Siemens) have 100%-H₂ models on the roadmap but not at the deployed scale Texas would need.

Stack those three constraints and you get the cost gap. Natural-gas-fired generation dispatches into ERCOT at roughly $30/MWh.² Grey hydrogen (steam-methane reforming, no capture) generates at roughly $45/MWh equivalent. Blue hydrogen (with carbon capture) lands near $60/MWh. Green hydrogen (electrolysis from renewables) lands above $100/MWh on most days, with the IRA 45V credit needed to bring it down to the $50–60 range — and the credit is currently subject to revision.

In a market where the marginal hour is set by gas peakers at $30/MWh against batteries discharging stored solar, hydrogen at $50–60/MWh has nowhere profitable to dispatch. It is more expensive than gas, less flexible than batteries, and incompatible with the installed turbine fleet without capital reinvestment that nobody is underwriting at scale.

The hydrogen story didn’t lose to a competitor. It lost to physics, and the physics hasn’t changed.

§3 — What it means for materials

Two threads run through this for a materials company.

Battery cathode and anode chemistry sit on the evening ramp. The shoulders the storage fleet works in determine what the cells need to do — deep discharge over 2–4 hours, daily cycling, multi-decade lifetimes under high state-of-charge dwelling. The cell chemistries that survive this are not the same chemistries that won the EV market (which optimizes for energy density first, cycle life second). LFP dominates grid storage for a reason — it cycles cleanly, tolerates high SOC dwell, and degrades predictably. The next-generation grid cell (sodium-ion at scale, solid-state, advanced LFP variants) is being chosen against the duty cycle of the evening ramp, not against the EV cycle. We are watching that chemistry choice closely.

The hydrogen-pipeline materials problem is real but mostly outside our wheelhouse. Polymer-lined pipe, embrittlement-resistant alloys, and turbine alloys that survive H₂ combustion are all materials problems with active commercial supply chains. They are not in our screening pipeline — but they sit adjacent to the superconducting-cable and grid-protection materials that are. If hydrogen comes back as a regional fuel for specific industrial corridors (Houston Ship Channel, Permian), the materials supply chain to enable that will look more like specialty alloys than like commodity steel.

For now: the grid that ERCOT runs in 2026 is wind, solar, gas, and batteries — in that order by hours, with datacenters and miners filling the floor. The shoulders run the system. Hydrogen sits in the corner waiting for someone to underwrite the turbine retrofit. Until somebody does, the storage fleet is what carries the evening ramp. Keeping that fleet on warranty and operational is the binding question.

---

Footnotes

1. ERCOT, Generation Mix Reports and Real-Time System Conditions dashboards, 2025 data. Renewable penetration above 70% recorded multiple times during 2025 spring shoulder months. ↩ ↩²

2. ERCOT, Real-Time and Day-Ahead Market Reports, settlement-point price data, 2024–2025 averages for natural-gas-fired generation. Dispatched cost (not bid) used to align with operator economics. ↩ ↩²

3. U.S. Bureau of Economic Analysis, Texas private fixed investment in information processing and software, datacenter subcategory, 2019–2024. ↩

4. ERCOT, Capacity, Demand and Reserves report, December 2025; battery storage installed capacity figures. ↩

5. U.S. Department of Energy, Hydrogen Pipeline Working Group technical reports, 2020–2023; volumetric energy-density and pipeline retrofit analyses. ↩

6. Lieuwen, T. C. et al., “Burner Development and Operability Issues Associated with Steady Flowing Syngas Fired Combustors” Combustion Science and Technology 180(6), 1169–1192 (2008). Updated H₂-specific dynamics: General Electric / Mitsubishi Power technical white papers on 100%-H₂ turbine development, 2022–2024. ↩