Tafel Power

Insights

Original analysis for power infrastructure transactions, large-load strategy and investment decisions across US power markets. Each brief ends in a decision you can act on, states what would prove it wrong, and shows its working.

40 briefs · January 2026 to present

Most recent

Latest

The newest brief, published in full with its chart and its finding.

September 13, 2026

Batteries, Generators, or Demand Response? The Transmission Determinant Decides

Transmission exposure runs from one peak hour a year to every megawatt hour consumed. The shape of the determinant decides which instrument can reach it, and sometimes none of them can.

Sustained work

Research series

Two runs of briefs where the same screen is applied repeatedly, one inside ERCOT and one across markets. Written in order, so the method compounds rather than repeats.

The ERCOT gas queue

The gas queue reads 69.7 GW. Only 8.8 GW of it is signed and near-term. Eight briefs, in order, on how that gap was established and why the headline number cannot be underwritten.

  1. Part 1
  2. Part 2
  3. Part 3
  4. Part 4
  5. Part 5
  6. Part 6
  7. Part 7

    July 11, 2026

    How Much of ERCOT's 69.7 GW Gas Queue Is Credible?

    Apply the historical post-IA completion rate to signed gas and a conservative request-to-operation base-rate weight to unsigned gas, and 69.7 GW becomes about 17 GW on a completion-weighted basis.

  8. Part 8

    July 13, 2026

    The ERCOT Queue Cannot Distinguish Captive Gas from Speculation

    For anyone screening ERCOT gas by interconnection status: a dedicated plant can be under contract to a named customer and still appear in the queue as unsigned and in study. Project Kilby shows why captive generation needs a separate diligence framework.

Signed, near-term gas across US markets

The same screen run across five markets between April and July: how much firm gas is really available, and where the answer is hiding. Ohio built its already, MISO put it in the utility plan, Northern Virginia is leaning on the existing fleet, West Virginia's lead is one project, and the second-largest signed queue is in Oklahoma.

  1. Part 1
  2. Part 2
  3. Part 3
  4. Part 4
  5. Part 5

Lane

Develop

Where you can actually build, what gates it, and how long it really takes.

September 3, 2026

The Utility Publishes the Process. Not the Price of Finding Out.

Wisconsin's public record tells a 100 MW load which tariff it must take, how long the term is, what security it must post and the gates it must pass. It does not say what the next gate costs, and the first answer is directional rather than a megawatt and date.

July 19, 2026

Can Entergy Serve the Next Gigawatt?

Why Entergy is one of the few utilities able to build firm power at AI scale, what customers must fund, and how its regulated model compares with merchant ERCOT.

July 12, 2026

What Actually Powers a 24/7 Data Center?

For hyperscalers and infrastructure investors: the interconnection queue contains nearly 2 TW of proposed capacity, most of it solar, wind, and storage. The scarce product for a 24/7 load is firm capacity, not annual clean energy.

Lane

Invest

What the asset earns, who captures it, and what the contract should say.

September 1, 2026

The Demolition Is Priced. The Carbon Is a Study.

Not owning the plant is ordinary. What is not ordinary is one contract that assigns the construction cost, the exit value and the demolition bill to a named company, and leaves the carbon at commercially reasonable efforts and a funded study.

August 16, 2026

Who Holds the Fuel Risk?

Gas moved by $39 to $59 a megawatt-hour over three years. Where that lands is a drafting choice, not a market fact.

August 16, 2026

Fifty Percent of What?

Every large-load power deal sets a minimum the customer owes whether or not it draws. Everyone compares the percentage. What the percentage is a percentage of is what decides whether it protects anyone.

August 15, 2026

Large-Load Power Contracts: The Record

Four structures, read from the tariffs, the commission orders and the filings. Facts and sources only. The arguments are in the two briefs this supports.

August 6, 2026

Gas Turbines Are Scarce. The Turbine Maker Reports a Thinner Margin.

What every part of the data center power supply chain actually reports, from utilities to software. The turbine maker, holding the acknowledged bottleneck, reports 18.8 percent on an EBITDA basis, against 27.5 percent operating at Eaton. Four parts of the chain disclose no separable return at all.

July 29, 2026

Everyone Saw Mozal Coming. It Closed Anyway.

South32 spent six years trying to replace the power and wrote off $372 million seven months before the plant stopped. So the date was never the risk. The power simply could not be replaced, and no filing has a field for that.

July 23, 2026

ERCOT Fleet Revenue Rose 8 Percent. Revenue per Megawatt Fell 37 Percent.

Total storage revenue keeps rising, but largely because capacity is added faster than per-unit revenue falls. The underwriting error is treating a market revenue forecast as if it described your asset. What still earns depends on which edge an asset holds and how fast it erodes.

July 15, 2026

Clean Power Earns Nothing Where It Is Used Most

Europe charges importers for the carbon in their products. In the industries that run on electricity, it ignores the carbon in the electricity. That exemption is borrowed from a European subsidy and ends when the subsidy changes, which makes it a drafting problem for anyone signing a fifteen year power contract.

July 14, 2026

Cheapest Energy Isn't Cheapest Power

An illustrative benchmark of what a 500 MW load at a 95 percent load factor pays for delivered power across the seven organized markets. Wholesale prices cluster between 30 and 55 dollars per MWh. Delivered, the displayed totals run from about 36 to about 87.

July 2, 2026

When Faster Power Stops Being Expensive

The best way to energize early is not fixed. It moves with what your load earns, and on the illustrative case it changes twice: once at about $49 per MWh and again at about $114.

June 15, 2026

LCOE Is Not a Deal Price

Published cost curves rank technologies by dollars per megawatt-hour at the plant gate. A firm-power buyer pays for a firm megawatt, delivered at a site, on a date. Three adjustments reshape that comparison and reorder the ranking.

April 15, 2026

What a Legacy Biomass Plant Is Actually Worth

Value it as a bundle: existing cash flow, generation interconnection and electrical infrastructure, a permitted industrial site near transmission, and large-load development potential. The operating economics set the base case; the site and its options are the upside.

Lane

Intelligence

How the market is structured, and what that changes about your next decision.

September 11, 2026

A Texas Power Plan Can Cost 9.8 Cents. Or 22.2.

Across the 57 plans priced this way, that is the median at 1,000 kWh against the median at 500. The difference is how much you use that month. I downloaded all 1,053 plans on the state's Power to Choose site to find out how common this is.

August 9, 2026

The Data Center Buildout Has a Toilet Paper Problem

GE Vernova has more gas turbine capacity reserved than sold: 63 GW of held production slots against 53 GW of signed orders. Holding a slot is not the same as buying a turbine, and that difference is what the headline number hides.

June 30, 2026

The Real AI-Power Risk Sits Downstream of the Megawatts

For anyone siting or financing AI infrastructure: new AI hardware arrives on a two-year cycle while the delivery infrastructure built for it is financed for decades. Firm megawatts are necessary but interchangeable; the harder thing to secure is delivery that can absorb several hardware generations without a major retrofit.

Reference

Regulatory tracking

Rule changes as they landed, oldest first, so you can see how the Texas large-load framework actually moved.

Questions, corrections or disagreement on any of this are welcome: kris@tafelpower.com