Energy / Electric & gas utilities · Deep dive

Duke Energy

A regulated Carolinas-Florida-Midwest electric monopoly (~8.6M electric customers, ~$33B revenue) betting an industry-record $103B five-year capex plan on the data-center demand wave — turning 7.8 GW of signed hyperscaler ESAs and a 15.4 GW pipeline into rate base at a 9.8% NC-settled ROE, while behind-the-meter bypass, a $10B equity overhang, and unhealed coal-ash scars test whether the regulated compounding machine keeps working.

well positioned

A regulated monopoly footprint over three of the top-five US data-center growth markets — the Carolinas, Florida and Indiana — earning a 9.8% NC-settled ROE (with sharing to 10.3%) means each of the 7.8 GW of signed hyperscaler ESAs (and the 15.4 GW late-stage pipeline) converts almost mechanically into rate base at a return investors will pay ~20x forward earnings for.

My take ★★ Very interesting ★ Interesting − Maybe ✗ Not interesting

Snapshot

Duke Energy is a Charlotte-based investor-owned electric and gas utility serving ~8.6M electric customers across the Carolinas, Florida, Indiana, Ohio and Kentucky — roughly .2B in TTM revenue (July 2026) and a ~.3B market cap. It is one of the largest US utilities by rate base and customer count. The story now is the data-center demand wave: in Q2 2026 Duke reaffirmed a B five-year capex plan through 2030 (an industry record), disclosed 7.8 GW of signed hyperscaler electric-service agreements (ESAs) with a 15.4 GW late-stage pipeline targeted for conversion by mid-2027, and beat Q2 with .43 in adjusted EPS versus .25 a year earlier. New CEO Harry Sideris — who succeeded Lynn Good on April 1, 2025 — inherits a regulated growth machine, unresolved coal-ash costs, a B equity issuance ahead, and the question of whether hyperscalers eventually stop needing regulated utilities at all.

Founding story

In the late 1890s Dr. W. Gill Wylie, a New York surgeon originally from South Carolina, decided the Piedmont’s addiction to cotton was starving it of industry, and the answer was hydro-electric power on the Catawba River. Wylie roped in civil engineer William States Lee and tobacco magnate James Buchanan Duke to fund it. The Catawba Hydro Station lit up Victoria Cotton Mills on April 30, 1904 — Duke’s founding date. Southern Power Company was incorporated in 1905; in 1924 the holdings took the name Duke Power. J.B. Duke’s parallel bequest that year created the endowment that renamed Trinity College as Duke University; the utility and the school share a family, not an ownership.

The modern Duke Energy is the product of three deals: the 1997 combination with Houston gas-pipeline company PanEnergy (creating the “Duke Energy” name); the April 2006 B acquisition of Cinergy (~1.5M electric + ~500k gas customers in OH, IN, KY); and the July 2012 merger with Progress Energy — B including debt, folding in 3.1M customers in Florida and the Carolinas and making Duke the largest US utility by customer count. Lynn Good, the ex-Cinergy CFO promoted to Duke CEO in July 2013, spent a decade cleaning up what those mergers created — coal ash, an aborted CEO handoff, an activist campaign — before passing the baton to 29-year Duke veteran Harry Sideris on April 1, 2025.

How it works

Duke is a regulated utility, which in America is a very specific machine. In each state a public utility commission (NCUC, Florida PSC, Indiana IURC, PUCO, Kentucky PSC, PSCSC) grants Duke an exclusive right to serve customers inside a defined territory. In exchange, Duke serves everyone at rates the commission approves. Duke invests capital — poles, wires, transformers, gas turbines, nuclear refuelling, IT, coal-ash closure — and that spend, once in service, becomes “rate base.” The commission then lets Duke earn an approved return on equity on that rate base and recover costs and depreciation through customer bills.

Every dollar of qualifying capex becomes near-guaranteed regulated earnings. The July 2026 partial settlement in Duke Energy Carolinas’ latest NC rate case set the allowed ROE at 9.8% with a 53% equity structure, an earnings-sharing mechanism of ±50 basis points (Duke can earn up to 10.3% before sharing kicks in), a M revenue requirement, ~.7B retail rate base, and residential rate hikes of 5.9% in 2027 and 3.6% in 2028. The data-center wave supercharges this: hyperscaler ESAs commit specific customers to specific new capacity, so the corresponding gas, transmission and battery capex flows into rate base with a pre-committed off-taker rather than merchant risk.

Product and business overview

Three segments. Electric Utilities and Infrastructure (~85%+ of earnings) — the regulated electric business across six states. Gas Utilities and Infrastructure — Piedmont Natural Gas serving NC, SC and TN (Duke announced the sale of Piedmont’s Tennessee operations in 2025). Commercial Renewables — largely divested in 2023 to focus on the regulated core. Generation mix is a mid-transition tapestry: nuclear (Duke is the second-largest US nuclear operator, with Oconee, Catawba, McGuire, Robinson, Brunswick, Shearon Harris), coal (full retirement targeted by 2035, but Belews Creek, Cliffside, Marshall and Indiana’s Gibson have all been extended), gas (adding ~7.5 GW of new combined-cycle including the ~1.4 GW Anderson County, SC plant approved March 2026, in-service ~2031), renewables and storage (~4.5 GW of batteries in the plan), and — from the mid-2030s — SMRs, with a January 2026 early-site-permit filing to the NRC for ~600 MW at Belews Creek.

Business model and pricing

Rates are set by regulators, not markets — there is no price list. The formula: allowed ROE times equity portion of rate base, plus recovery of debt costs, opex and depreciation, collected through tariffs the commission signs off on. What analysts price is the trajectory. Q2 2026 adjusted EPS was .43 (vs. .25 Q2 2025); FY2026 guidance reaffirmed .55-.80; the long-term algorithm is 5-7% adjusted EPS growth through 2030, top half of range from 2028 as data-center load ramps. Management flagged an additional -10B of capital upside above the B plan as ESAs convert, concentrated in Indiana and Florida. Every incremental billion of capex, at ~53% equity and ~9.8% allowed ROE, translates to roughly M+ of pre-tax regulated earnings — the market currently prices DUK at ~19-20x forward.

Traction over time

Market analysis

For twenty years US electricity demand was flat; AI data centers ended that. US data centers consumed ~183 TWh (~4.4% of US electricity) in 2023-2024; EPRI, DOE/LBNL and the IEA project 6.7-12% of US electricity by 2028-2030, with global data-center demand up ~130% by 2030. Roughly 30% of announced US data-center capacity — ~56 GW across 46 sites — is being planned with behind-the-meter generation, per work surfaced by measuredai and S&P Global. Duke’s territories sit on the front lines: the Carolinas (top-three US data-center growth market, with Google, Amazon, Microsoft, Meta and QTS building near Charlotte and the Triad), central Florida (Google, Meta, Microsoft), and central Indiana (a surprising top-three growth market thanks to Meta, Google and AWS).

Competitive intel

Regulated utilities do not compete for customers — the point of the monopoly grant is that they do not have to. They compete for capital (which investor gives you dollars for rate-base growth) and regulatory goodwill (which commission lets you earn 9.8% versus 9.4%). Dominion Energy is Duke’s sharpest peer: it owns Virginia’s Data Center Alley (roughly a quarter of global data-center capacity by traffic), and every AI-thesis dollar going to D does not go to DUK. Southern Company has the head start Duke does not: Vogtle 3 & 4 are the only newly commissioned US nuclear in a generation, feeding Georgia’s hyperscaler corridor with 24/7 clean firm power. NextEra Energy is both a Florida rival (via FPL) and the exemplar the market values above Duke on multiple, largely because NEER’s ~300 GW pipeline lets it serve hyperscalers beside anyone’s regulated territory. Xcel plays the same regulated-growth song for Colorado and Minnesota load; Entergy showed a mid-cap regulated utility can win a marquee AI deal by locking in Meta’s ~2 GW Louisiana campus.

The genuinely disruptive competition is behind-the-meter. Talen sold Cumulus Data to AWS in March 2024 for ~M and signed a June 2025 retail PPA delivering up to 1,920 MW off Susquehanna nuclear through 2042; Bloom Energy’s fuel cells sit on data-center pads today; Base Power and other entrants build distributed capacity outside the rate base. FERC’s November 2024 rejection of the original Talen/AWS deal stopped one instance of bypass, but Sen. Cotton’s January 2026 DATA Act — exempting off-grid data centers from FERC oversight altogether — points the political weather the other way. Every gigawatt hyperscalers self-supply is a gigawatt that never enters Duke’s rate base.

History and evolution

• 1904-04 — Catawba Hydro Station lights Victoria Cotton Mills.
• 1924 — Renamed Duke Power; J.B. Duke’s endowment renames Trinity College as Duke University.
• 1997-06 — Duke Power + PanEnergy merger creates Duke Energy.
• 2006-04 — Cinergy merger (~B); Duke enters OH, IN, KY.
• 2012-07 — Progress Energy merger (~B incl. debt) — largest US utility by customer count. The Progress CEO is pushed out within hours of closing; NC regulators later fine Duke.
• 2013-07 — Lynn Good becomes CEO.
• 2014-02 — Dan River spill: corroded stormwater pipe under an ash pond collapses; ~39,000 tons of ash into the river.
• 2015-05 — Duke subsidiaries plead guilty to nine Clean Water Act violations; M fines/restitution.
• 2020 — NC settlement shifts .1B in coal-ash costs from ratepayers to shareholders; NC/SC ash closure ~-9B total.
• 2021-05 — Elliott urges break-up into three regional listed cos; Duke rejects; Elliott later stands down.
• 2022-02 — Duke pledges full coal exit by 2035; coal to <5% of generation by 2030.
• 2023 — Divests commercial renewables to Brookfield (~.8B).
• 2025-04 — Sideris becomes CEO; Lynn Good retires.
• 2026-01 — NRC early site permit filed for ~600 MW SMRs at Belews Creek (first unit ~2036).
• 2026-03 — PSCSC approves ~1.4 GW Anderson County gas plant (in-service ~2031).
• 2026-04 — B 2026-2030 capex plan announced (industry record).
• 2026-07 — NC rate settlement: 9.8% ROE, 53% equity, sharing to 10.3%.
• 2026-08-04 — Q2 2026: .43 adjusted EPS; ESAs to 7.8 GW; 15.4 GW pipeline; B equity plan 2027-30.

What people say

The case for. Sell-side underwrites the pitch: a B plan aligned with contracted data-center load is a lower-risk way to earn 5-7% EPS growth than pure-play IPPs, and the NC 9.8% ROE with sharing to 10.3% is close enough to Duke’s request to signal constructive regulation. Seeking Alpha’s July 2026 note called the contracted load a step-change in credibility. Employees, on Glassdoor’s 3.9/5 across 2,077 reviews (69% recommend, 71% positive outlook), praise stable pay, benefits, and — surprisingly often — a stated commitment to “doing the right thing” and reliable work-life balance for a hurricane-state utility.

The complaints. The bear file is thick. Ratepayer groups and the NC Attorney General spent a decade fighting Duke over coal-ash cost recovery; the 2020 settlement pushed .1B onto shareholders, but the ~-9B NC/SC total ash-closure cost still casts a long shadow, and the Dan River spill remains a public-trust wound — 39,000 tons of ash into a drinking-water source, a M federal Clean Water Act plea, and a widely-cited pattern of captured monitoring. Duke Energy Carolinas’ original 2025 filing sought hikes above 13%, halved to ~5.9% in the July 2026 settlement — but the earnings-sharing mechanism will bite if actual ROE runs hot. Employees on Glassdoor recount the familiar utility complaints: “Upper management is completely out of touch,” “They will bury you in work,” time-off approval games. Elliott’s 2021 letter — accusing Duke of a “bigger is better” strategy without effective independent board oversight — was rejected but more sober than its reception acknowledged, and much of its operational critique quietly informed Sideris’s simpler regulated-core story.

Outlook: well positioned or at risk?

Well-positioned — and specifically because the moat Duke owns is the exact one AI power demand is now willing to pay for. A regulated monopoly footprint over three of the top-five US data-center growth markets (Carolinas, Florida, Indiana), earning a 9.8% NC-settled ROE with sharing to 10.3%, means every gigawatt of signed hyperscaler load converts almost mechanically into rate base at a return the market pays roughly 20x forward earnings for. The 7.8 GW of signed ESAs and 15.4 GW late-stage pipeline give the B capex plan a contracted demand-side counterparty that pure-play IPPs and merchant renewables cannot match; the Belews Creek SMR filing positions Duke — a decade late — for the firm-clean-power auction Southern and Constellation are already winning.

Three counter-arguments should be taken seriously. Behind-the-meter bypass: if Talen/AWS-style deals become the default and the DATA Act neuters FERC on off-grid data centers, some share of the 15.4 GW pipeline becomes power to somebody else. Duke’s answer — hyperscalers still want redundancy and interconnection from a regulated utility — is genuine but insufficient if the marginal gigawatt bypasses the meter. Financing cost: a B equity issuance across 2027-2030 into a still-elevated rate environment dilutes per-share earnings and compresses the multiple; the “top half of 5-7%” 2028+ guide already implicitly funds that dilution. PUC and political risk: the 9.8% NC ROE is the settled version of a request that would have hit residential bills 13%+ in 2027; ratepayer politics can turn fast, an unfriendly Indiana IURC could freeze rate cases, and the coal-ash file is one bad headline from re-opening. On a five-year view the position compounds — but the buyer of DUK at 20x is paying a growth multiple for a business whose growth depends on hyperscalers not bypassing the meter, regulators staying constructive, and equity markets absorbing the dilution. The moat is real; it is not automatic.

How a challenger would attack it

The wedge is the meter itself. You cannot out-regulate a regulated monopoly, so a challenger doesn’t file for a franchise — it sells hyperscalers what Duke structurally cannot: speed and price outside the rate base. Duke’s Anderson gas plant approved in March 2026 won’t serve load until ~2031; an SMR at Belews Creek arrives ~2036. A challenger stacking Bloom fuel cells, gas turbines and batteries behind the meter can energize a campus in 18-24 months, and the Talen/AWS playbook — up to 1,920 MW of Susquehanna nuclear through 2042 — proves hyperscalers will sign decades-long deals to skip the queue. The DATA Act gives the attack political cover by exempting off-grid data centers from FERC entirely. The second exploitable seam is ratepayer politics: Duke’s original 2025 NC filing sought 13%+ residential hikes to fund hyperscaler-driven capex, and every headline pairing AI data centers with household bill increases erodes the regulatory goodwill the 9.8% ROE depends on. A challenger that lets hyperscalers self-supply removes them from the rate base and hands consumer advocates the argument. Duke’s $10B equity overhang means it cannot cut price to respond — its earnings are the price.

Same playbook, new buyer

Duke’s actual playbook — contracted large-load ESAs converted into guaranteed-return infrastructure — travels to buyers Duke will never serve. The most promising shift is doing ESA-style power contracting for the mid-market: colocation operators, industrial reshorers, and 50-200 MW AI training clusters too small for Duke’s hyperscaler-grade pipeline attention but desperate for firm power on a date certain. Entergy proved a mid-cap can win Meta’s ~2 GW Louisiana campus; the next tier down is wide open in states outside Duke’s six-state footprint — Texas, Georgia’s periphery, the Southwest — where a developer offering energy-as-a-service with utility-grade reliability but merchant-speed delivery has no incumbent equivalent. Duke won’t follow because it structurally can’t: its model requires a commission-granted territory, capex must clear a rate case, and anything outside NCUC/PSC jurisdiction is the commercial business it deliberately divested to Brookfield in 2023. The regulated compounding machine is the moat and the cage — Duke chose the cage, and Sideris’s simpler-regulated-core mandate locks it.

Sources and further reading

• Duke names Harry Sideris President and CEO — Duke Energy, January 2025.
• Duke Energy Q2 2026 slides: EPS beats amid data center boom — Investing.com, Aug 4, 2026.
• Utility giant Duke plans industry record B — Fortune, April 2026.
• Duke targets -10B capital upside, reaffirms 5-7% EPS growth — Seeking Alpha, August 2026.
• Duke Energy Carolinas halves rate hike in new settlement — WFAE, July 2026. 9.8% ROE, 53% equity, sharing to 10.3%.
• Duke to issue B equity for gas generation growth — Utility Dive, 2026.
• Duke Energy, Progress Energy Complete Merger — Duke Energy, July 2012. B deal.
• Duke Energy history and Duke Power — NC History — 1904 Catawba, J.B. Duke, Wylie, Lee.
• Duke fined M in coal ash spill and NC DEQ Dan River fine.
• .1B coal-ash settlement shifts costs from ratepayers — 2020.
• Elliott letter to Duke’s Board — May 2021.
• Talen behind-the-meter AWS fight; Talen-Amazon supply agreement.
• Grid Bypass: onsite power at US data centers (~56 GW / 46 sites).
• Duke Belews Creek SMR early site permit — Jan 2026.
• Anderson County SC gas plant approved — March 2026.
• Duke Energy Reviews (2,077) — Glassdoor, 2026.
• Duke Energy market cap — Aug 2026. ~.3B.