US Energy Market Intel

The axis above $300/MWh is compressed (see the break line) so the everyday $0–$70 band stays readable while spikes still show their true shape and are labeled with their peak value. Massachusetts saw a January cold-snap spike to $643/MWh (Jan 27) and a July heat-wave spike to $302/MWh (Jul 2) — both typical of New England's weather-driven price volatility. California stayed in a much narrower $0–$70/MWh band over the same period.

Bars show the mean settlement price per region; the number in parentheses is the median, a more robust read since a handful of scarcity-priced trades can pull the mean up. Sample sizes range from ~1.8M to ~50M individual settlements per region in the quarter.

Multi-decade contracts often file one row per delivery year with an escalating rate; the table above shows the nearest active year per contract. Buyer list is extensible (Meta, Oracle, and others found no priced hits in this sample quarter). *One contract was filed as a capacity payment ($/kW-month) rather than an energy rate ($/MWh); it's converted here assuming a 40% capacity factor (typical for SPP-region wind) so all rows are comparable — treat it as an estimate, not a filed rate.

Copper only reports monthly, so its line is a step function between month-end readings, carried forward to align with the daily power/oil grid. Axis compressed above index 350 (see the break line) — Massachusetts' January cold-snap briefly pushed its index well above that, and now still shows as a labeled peak rather than being flattened off.

Method: linear regression of month-over-month % change in wholesale power price against month-over-month % change in WTI oil price (Jan–Aug 2026, n=7 monthly observations). The fitted slope (β) is applied as a one-time shift to the price at the chosen horizon under the chosen oil scenario; a 10,000-path Monte Carlo, bootstrapping from the regression's own historical residuals (in log-return space, demeaned so the no-shock baseline has no artificial drift), builds the surrounding P10–P90 uncertainty band for both the shocked and no-shock (baseline) cases. With only 7 monthly data points, treat β and R² as illustrative, not statistically reliable — this is a transparent, small-sample fit, not a validated causal model. The bands are wide because wholesale power genuinely is this volatile in the underlying data (both markets have already shown 2–6x single-month swings in 2026); that width is a real feature of this asset class, not a modeling error.