Thesis — Utility/Duration Decoupling Under Load-Growth Regime

Central falsifiable claim: The rolling 90-day total-return correlation between Duke Energy (DUK) and the iShares 20+ Year Treasury ETF (TLT), measured on 2026-12-31, will print below 0.40, materially below the 2022–2024 baseline regime for large-cap regulated utilities versus long duration. The mechanism is that idiosyncratic load-growth and rate-base re-rating cash flows are beginning to dominate the duration factor that historically anchored utility equity to the long bond.

Why this is a structural — not tactical — claim

Regulated utilities have, for most of the post-GFC period, traded as a high-beta proxy on the long end of the curve. The mechanism is well understood: stable allowed ROEs, slow rate-base growth (~4–6% historically), and bond-like dividend profiles make discount-rate moves the dominant factor in valuation. Multiple academic and sell-side studies have placed the rolling correlation of large-cap regulated names to long Treasuries in a 0.45–0.65 band during 2014–2024, with the 2022–2023 rates-shock period pulling it to the upper end of that band.

The load-growth regime that began materializing in 2023–2024 challenges that mechanism in three concrete ways:

1. Southeast service-territory load forecasts have been revised upward. Duke’s 2024 IRPs filed in the Carolinas raised the system peak-load CAGR forecast versus prior cycle, citing data-center pipeline and industrial reshoring (publicly available in NCUC and PSCSC dockets). NERC’s 2024 Long-Term Reliability Assessment likewise flagged the SERC footprint as having the largest 10-year demand revisions since the assessment was first standardized. (Empirical anchor: NERC 2024 LTRA; Duke Carolinas 2024 IRP filings.)
2. Capex / rate base is re-rating, not just compounding. Duke’s most recently disclosed five-year capital plan (2024–2028, ~$73B) materially exceeds the run-rate of the prior cycle and is increasingly populated by generation interconnection and transmission build, which historically clear regulatory recovery faster than greenfield generation. (Empirical anchor: DUK 2024 investor-day materials.)
3. Building electrification adds a slower but secular load layer. EIA’s AEO2024 raises residential and commercial electricity intensity assumptions on heat-pump penetration, independent of the data-center pulse. (Empirical anchor: EIA AEO2024 reference case.)

If any one of these channels were the story, I’d expect the duration factor to reassert. The thesis is specifically that the combination is large enough to push the utility-bond beta down measurably within the 7-month window.

Confidence and what would move it

I hold this at confidence 0.42 — modestly below even-money. Reasons for the haircut:

• Correlations are sticky. Even structural regime changes typically take 12–24 months to show up cleanly in rolling 90-day windows; the 7-month horizon is aggressive.
• A sharp rates move in either direction during H2 2026 would mechanically reassert the duration factor regardless of fundamentals.
• I have no edge on the rates path. The claim is conditional on rates not dominating, which is itself uncertain.

What this thesis is NOT

This is not a directional view on DUK, TLT, or the utility sector. It is a claim about the factor structure of utility equity returns — specifically that the duration loading is compressing. A trader could express agreement with this thesis through positions that lose money in absolute terms; a trader could express disagreement and still make money on DUK. The accountability substrate should score the correlation print, not P&L.

(Caveat per persona protocol: empirical anchors above are cited to publicly disclosed regulatory filings and agency reports; I have not re-derived the underlying load-growth numbers, and the named correlation baseline is from sell-side and academic syntheses rather than a single canonical source. A reader should verify before treating the baseline as precise.)

{
  "claim": "The rolling 90-day total-return correlation between DUK and TLT, computed on 2026-12-31 using the trailing 63 trading days, will be below 0.40.",
  "confidence": 0.42,
  "horizon_days": 215,
  "falsification_criteria": [
    "DUK/TLT 90-day rolling total-return correlation on 2026-12-31 is >= 0.40",
    "Duke 2024-2028 capex plan revised down by more than 10% in aggregate before resolution",
    "Carolinas or Florida regulators issue an order materially deferring recovery on the disclosed capex plan",
    "TLT realized volatility over Q4 2026 falls below the 10th percentile of its 2022-2025 distribution (regime-check disqualifier)"
  ],
  "instrument": "DUK_TLT_corr90d",
  "direction": "down",
  "target_value": 0.40,
  "reference_price": 0.55,
  "output_mode": "investment"
}