AI Absorption Ledger / DUK

Duke Energy

DUK · Q2 2026 · reported 2026-08-04 · revenue $7.59bn

Assessment

The quarter's AI passages are, on the call, the CEO on AI tools that monitor construction deadlines and progress, and the risk-factor line in the 10-Q and the release. No Q2 source names AI as a cause of data-centre demand; the 10-Q calls that demand a significant contributor to projected load growth. Every channel holds where it was in Q1: described, unsized, and none has moved.

Signed ESAs with data center customers rose to GW, and the start dates are unchanged: load begins as early as the second half of 2027. The CFO puts the upside to the current capital plan at to if the pipeline converts, which has not started. Capital expenditures were in the quarter against a year earlier; that line, capital and out of totals, is the ceiling on any AI part.

Operation, maintenance and other fell by on the year, which the 10-Q puts down to lower storm amortization in Florida (claim c17), and the CEO credits cost efficiency to a culture of continuous improvement (claim c22); no operating line is attributed to AI.

Paid for AI1 channel · 1 not sized

other

Capital spending on generation and grid for data-centre load, inside the regulated capital plan

Not sized

AI named beside another cause: no Q2 source names AI as a cause of the growth the plan serves, the earlier statements name it only beside other causes, and no data-centre or AI share of capital spending is stated, so no ballpark is built on a judgment share; the ceiling (the quarter’s capital expenditures, the plan and its stated upside) is in the metrics.

described, no sizedisclosure: described· motive: exploratory· before LLMs: expanded· layer: facilities

Capital expenditures were in the quarter, the first half less Q1, against a year earlier, a change of ; the CEO says the company is deploying more than a month (claim c5). The CFO puts the upside to the current plan at to if further contracts are signed, mainly in Indiana and Florida, triggered only when the ESAs are signed (claims c4 and c19); that upside has not started and is not a size. Turbines under the GE Vernova framework rose to and the build now adds GW by 2031 (claims c16 and c20). Capital spending is left out of totals.

Evidence: 10 quotes, 11 figures, 2 confounds, 4 from before coverage

The part of the utilities’ capital spending on generation, transmission and distribution that serves contracted data-centre load. The CEO ties the capital plan to regional growth driven by innovation in AI technologies and advanced manufacturing, the 10-Qs say data-centre demand supports the long-term regulated capital plan, and the CFO says further capital upside would be triggered as more ESAs are signed. The same plan serves population growth, other economic development, electrification, reliability, storm hardening, nuclear licence extensions and the replacement of retiring coal units, and no source states a data-centre share of it, let alone an AI share. Under the joint-cause rule the channel is described and unsized, with no ballpark built on a judgment AI share; the ceiling is quoted instead (the quarter’s capital expenditures and the stated capital plan). Capitalized, so it is traced and shown and left out of flow totals; it reaches the income statement as depreciation and as the return regulators allow on rate base, which no covered quarter separates for data-centre assets. The payees are turbine makers (GE Vernova under a framework agreement), EPC contractors (Zachry in the Carolinas) and other suppliers. Funding is mixed: operating cash flow, debt including convertible notes, equity priced under the at-the-market program, proceeds from the sale of the Tennessee gas business and the minority investment in Duke Energy Florida, and, for data-centre projects, the customers’ refundable capital advances.

Why this motive

Carried from Q1: capital committed ahead of the load it serves, which begins as early as the second half of 2027 (claim c3); the further upside is triggered only when more ESAs are signed (claim c19).

Before LLMs: expanded

At the anchor the CEO described a multi-year capital plan of , and the annual report projected capital and investment expenditures of for 2025 and for 2026, planned in part for new generation to meet demand that it traced to data centres and the expected artificial intelligence revolution beside other causes (claims duk-anchor-c1, duk-anchor-c3, duk-anchor-c5 and duk-anchor-c6). Capital spending on generation and the grid predates LLMs; no quarter of the part serving data centres can be traced, so the level has no baseline. The size is the whole of an activity that existed before.

“we are now in the early stages of the approval and planned construction of significant new generation investments and anticipate growing energy demands in the coming decades from continued migration into our attractive service territories, onshoring of domestic industries, electrification, and data centers and other investments from the expected artificial intelligence revolution.”
Filing, mdna, 10-K periodic report, 2025-02-27
“In the face of unprecedented demand from AI, data centers, chip manufacturers, and other economic development, natural gas remains an essential tool to provide reliable and affordable energy for customers and complements our substantial investments in renewables and energy storage.”
CEO, prepared remarks, earnings call, 2024-05-07
“Our strategy will drive continued growth underpinned by our 5-year $73 billion capital plan, efficient recovery mechanisms, and track record of constructive regulatory outcomes.”
CEO, prepared remarks, earnings call, 2024-05-07
“Duke Energy’s projected capital and investment expenditures, including AFUDC debt and capitalized interest, for the next three fiscal years are included in the table below.”
Filing, mdna, 10-K periodic report, 2025-02-27

Figures

  • Capital expenditures, the quarter (six months less Q1) · 2026-CQ2
  • Capital expenditures, prior-year quarter (six months less Q1) · 2025-CQ2
  • Capital expenditures for the quarter, year-over-year change · 2026-CQ2
  • Capital expenditures (cash flow statement), six months · 2026-01-01..2026-06-30
  • Capital expenditures (cash flow statement), six months of the prior year · 2025-01-01..2025-06-30
  • Capital upside to the current capital plan if further contracts are signed, low end · as-of 2026-08-04
  • Capital upside to the current capital plan if further contracts are signed, high end · as-of 2026-08-04
  • Capital deployed per month under the regulated capital plan, more than · as-of 2026-08-04
  • Gas turbines available under the GE Vernova framework agreement · as-of 2026-08-04
  • Generation capacity to be added by 2031, gigawatts · as-of 2026-08-04
  • Regulated capital plan, total as stated on the call (the quote gives no years) · as-of 2026-05-05

What else could explain it

  • other: The quarter names AI only for construction monitoring (claim c9); no Q2 source names AI as a cause of data-centre demand, so the AI cause rests on the Q1 call (c1) and the anchor (duk-anchor-c1, duk-anchor-c3), each naming AI beside other causes. The CFO names life sciences and advanced manufacturing beside data centres (claim c15). Nothing separates AI’s part.
  • line composition: The capital line also holds grid hardening and reliability work, nuclear uprates and licence extensions, storm restoration, gas utility spending and the replacement of retiring coal units; no data-centre share is stated.

Quotes

“As you can see on the right side of the slide, customers are also making strong progress building their facilities, with several moving to vertical construction. We continue to expect these customers to begin taking energy as early as the second half of 2027 and into 2028 and ramp into their full contracted load through the early 2030s.”
c3 · CFO, prepared remarks, earnings call, 2026-08-04
“As additional contracts are signed, there is $5 billion-$10 billion of upside to our current five-year capital plan to support additional generation and transmission needs, particularly in Indiana and Florida.”
c4 · CFO, prepared remarks, earnings call, 2026-08-04
“To meet this record demand and to continue long-term value for our customers, communities, and shareholders, we're executing on the industry's largest regulated capital plan, deploying more than $1 billion per month.”
c5 · CEO, prepared remarks, earnings call, 2026-08-04
“Demand associated with data center development remains a significant contributor to projected load growth.”
c11 · Filing, mdna, 10-Q periodic report, 2026-08-04
“We continue to expand our portfolio of data center electric service agreements, increasing contracted capacity while maintaining a disciplined approach to infrastructure investment.”
c12 · Filing, mdna, 10-Q periodic report, 2026-08-04
“These trends continue to support Duke Energy’s long‑term regulated capital plan while supporting reliable service and customer affordability.”
c13 · Filing, mdna, 10-Q periodic report, 2026-08-04
“The variance is primarily driven by proceeds received from the sale of Piedmont's Tennessee business, partially offset by higher capital expenditures within the EU&I segment.”
c14 · Filing, mdna, 10-Q periodic report, 2026-08-04
“Lastly, we're executing on the construction of new dispatchable capacity, including increasing the number of gas turbines available under our framework agreement with GE Vernova to 26 to align with the next phase of build in the IRPs.”
c16 · CEO, prepared remarks, earnings call, 2026-08-04
“That'll be triggered when the ESAs are signed and the requisite generation and transmission is modeled for those contracts.”
c19 · CFO, qa, earnings call, 2026-08-04
“Slide seven shows our continued progress on our record generation build, now on track to add 15 GW of capacity by 2031, which reflects additions from our latest 10-year site plan in Florida.”
c20 · CEO, prepared remarks, earnings call, 2026-08-04

By quarter

  • Q1 2026described · described, no size · exploratory
  • Q2 2026described · described, no size · exploratory

Cost displaced by AI1 channel · 1 not sized

operations · cheap to verify

AI tools that track construction milestones and progress on the generation build

Not sized

Management describes the tools and gives no magnitude, and the saving would fall on capitalized construction work: the oversight it displaces sits inside construction cost, and a savings channel cannot carry the capital flag, so any ballpark would land in the operating savings total as if it were an operating saving. Under the methodology's exception for savings on capitalized work, no ballpark is built.

described, no sizedisclosure: described· motive: exploratory· before LLMs: relabelled

Asked about bottlenecks in the gas build, the CEO cites the oversight the company is putting in using AI tools to monitor construction deadlines and progress (claim c9), beside the programmatic EPC approach (claim c18). No result is given.

Evidence: 3 quotes, 2 confounds, 3 from before coverage

The project management and construction team uses AI tools, with its equipment suppliers and EPC providers, to track construction milestones and progress, down to the cubic yard of dirt excavated and concrete poured (Q1 2026 call), and to monitor construction deadlines and progress (Q2 2026 call). The work displaced is project oversight on capital projects, done by the company’s own staff, so the counterparty is internal; a progress reading can be checked on site, so verification is cheap. No saving, rate, count or line is given. The oversight cost sits inside construction cost, which is capitalized, so any saving would lower capital spending and later depreciation rather than an operating line in the quarter. The CEO also credits the programmatic EPC approach (a programmatic contractor in the Carolinas, identical sites, the same materials) with cost and schedule synergies, which is a confound here. The sources do not say whether the tools are LLMs or older analytics, construction oversight was already part of the work at the anchor, and no line, price or rate moves, so under the tie-break the activity is read as renamed.

Why this motive

The CEO again names AI tools in construction oversight with no measure and no line moving (claim c9): the exploratory tell, carried from Q1.

Before LLMs: relabelled

At the anchor the annual report said that Duke Energy’s long-term strategy requires the construction of new projects, which involve a number of risks, including construction delays, nonperformance by equipment and other third-party suppliers, and increases in equipment and labor costs (claim duk-anchor-c7): overseeing schedules, suppliers and construction cost was already part of the work. The anchor also says the company was leveraging new technology, digital tools and data analytics across the business, and names the cost of keeping pace with technologies including artificial intelligence (claims duk-anchor-c8 and duk-anchor-c9); these are context, not evidence of the same tool under an older name. The sources never say the milestone tools are LLMs, and no line, price or rate is shown to move because of them, so under the tie-break the activity is read as renamed.

“Duke Energy’s long-term strategy requires the construction of new projects, either wholly owned or partially owned, which involve a number of risks, including construction delays, delays in or failure to receive required regulatory approvals and/or sitting or environmental permits, nonperformance by equipment and other third-party suppliers, and increases in equipment and labor costs.”
Filing, risk factors, 10-K periodic report, 2025-02-27
“We are leveraging new technology, digital tools and data analytics across the business in response to a transforming landscape and our grid improvement programs continue to be a key component of our growth strategy.”
Filing, mdna, 10-K periodic report, 2025-02-27
“This involves significant development and implementation costs to keep pace with changing technologies, including artificial intelligence, and customer demand.”
Filing, risk factors, 10-K periodic report, 2025-02-27

What else could explain it

  • other: The oversight work sits inside construction cost, which is capitalized; any saving would lower capital expenditures ( in the quarter) and later depreciation, not an operating line in the quarter.
  • transformation program: The CEO credits the programmatic EPC approach, identical sites and the same materials with cost and schedule synergies (claim c18), and describes milestone monitoring without AI in the prepared remarks (claim c10).

Quotes

“We feel really good about our plan and our ability to execute as well as the oversight that we're putting in using AI tools to monitor construction deadlines and progress.”
c9 · CEO, qa, earnings call, 2026-08-04
“We've contracted with EPC partners and we're closely monitoring construction milestones, enabling us to check and adjust in real time.”
c10 · CEO, prepared remarks, earnings call, 2026-08-04
“We're buying the same materials, the sites are identical, which gives us opportunities for synergies in how we build these things, both on the cost side, also on making sure that the schedule is met and the quality is met.”
c18 · CEO, qa, earnings call, 2026-08-04

By quarter

  • Q1 2026described · described, no size · exploratory
  • Q2 2026described · described, no size · exploratory

Revenue arriving through AI1 channel · 1 not sized

customer cohort

Electricity sold to data-centre customers, chiefly under electric service agreements

Not sized

Not yet revenue, and AI named beside another cause. Load under the ESAs starts as early as the second half of 2027 (claim c3); existing data-centre usage sits inside retail revenue with no amount, and AI is named only beside other causes, so no ballpark is built on a judgment share; the ceiling (EU&I and commercial revenue) is in the metrics.

described, no sizedisclosure: described· motive: exploratory· before LLMs: expanded· layer: facilities

Signed ESAs with data center customers reached GW, GW more than at the Q1 call, and the company expects the rest of the GW pipeline to convert by the first half of 2027 (claims c1 and c2). Customers are building, several in vertical construction, and the start dates are unchanged (claim c3). The CEO says the company serves many hyperscale customers (claim c7) and the CFO says minimum take provisions are the basis for revenue growth projections (claim c8). No data-centre revenue is stated, and the quarter does not name AI as a cause of the demand.

Evidence: 10 quotes, 6 figures, 3 confounds, 5 from before coverage

Electricity, capacity and grid service the regulated utilities supply to data-centre customers, chiefly under electric service agreements (ESAs) that carry minimum demand provisions, credit support, refundable capital advances and termination charges. The company names AI as a cause only beside other causes: on the Q1 2026 call the CEO said growth in its regions is driven by innovation in AI technologies and advanced manufacturing, and the fiscal 2024 annual report named data centres and other investments from the expected artificial intelligence revolution beside migration, onshoring and electrification. No source says what share of data-centre load is AI work, so any size would be a ceiling on the AI part; the remainder is the enterprise and general cloud work the same data centres run, and the non-AI growth (advanced manufacturing, life sciences, population) is named beside it. Load under the ESAs begins as early as the second half of 2027 and ramps over the following years, so none of it is revenue in the covered quarters. Data-centre usage the company already served at the anchor sits inside retail sales and is never given as an amount. The counterparty is mixed: hyperscale customers the CEO says the company serves, and other data-centre customers the sources do not name. The layer is facilities (power sold to whoever runs the data centres). Hyperscalers on this ledger (MSFT, AMZN, GOOGL, META) would carry the same money as data-centre cost; the ledger records each company’s own flows and does not net them. Duke Energy is registered as a buyer for its own AI use; this channel is the facilities-layer revenue a regulated utility earns from data-centre load, read as Constellation’s and Chevron’s power channels are read.

Why this motive

Carried from Q1 and still pre-revenue: data-centre customers are still expected to begin taking energy as early as the second half of 2027 (claim c3).

Before LLMs: expanded

At the anchor data-centre usage was already part of commercial retail sales and a driver of their growth, and the annual report named data centres and the expected artificial intelligence revolution among the sources of future demand (claims duk-anchor-c1, duk-anchor-c2 and duk-anchor-c4). Data-centre usage was an unstated part of the whole company's revenue, of total operating revenues in the first quarter of 2024 across every customer class; that figure is context, not the channel's level. No data-centre amount was given then or since, so the level has no baseline. In coverage new capacity is contracted to data-centre customers under ESAs, so volume moved and the tag is expanded (methodology, facilities for data centres). The size is the whole of an activity that existed before.

“we are now in the early stages of the approval and planned construction of significant new generation investments and anticipate growing energy demands in the coming decades from continued migration into our attractive service territories, onshoring of domestic industries, electrification, and data centers and other investments from the expected artificial intelligence revolution.”
Filing, mdna, 10-K periodic report, 2025-02-27
“Weather-normal sales volumes have shown growth in 2024 compared to 2023 due primarily to strong residential customer growth and strength in the commercial sector including data center usage.”
Filing, business, 10-K periodic report, 2025-02-27
“In the face of unprecedented demand from AI, data centers, chip manufacturers, and other economic development, natural gas remains an essential tool to provide reliable and affordable energy for customers and complements our substantial investments in renewables and energy storage.”
CEO, prepared remarks, earnings call, 2024-05-07
“Data center growth was a key driver in that in the quarter, and we expect that to continue throughout the year.”
CFO, qa, earnings call, 2024-05-07
“We successfully worked with our state partners to win 78 economic development projects in 2024 alone, representing approximately $26 billion in new capital investment and over 16,000 new jobs within our service territories. These projects include transformational life sciences, automotive, and semiconductors facilities as well as data centers.”
Filing, mdna, 10-K periodic report, 2025-02-27

Figures

  • Electric service agreements signed with data center customers, gigawatts (total to date) · as-of 2026-08-04
  • Electric service agreements added since the Q1 call, gigawatts · 2026-05-05..2026-08-04
  • Late-stage large-load pipeline including signed ESAs, gigawatts; the unsigned remainder is expected to convert by the first half of 2027 · as-of 2026-08-04
  • Electric Utilities and Infrastructure revenue from contracts with customers, three months · 2026-CQ2
  • Commercial retail electric revenue, three months (the class the anchor places data-centre usage in) · 2026-CQ2
  • EU&I revenue increase from weather-normal retail sales volumes, year over year · 2026-CQ2

What else could explain it

  • other: The quarter names AI only for construction monitoring (claim c9); no Q2 source names AI as a cause of data-centre demand, so the AI cause rests on the Q1 call (c1) and the anchor (duk-anchor-c1, duk-anchor-c3), each naming AI beside other causes. The CFO names life sciences and advanced manufacturing beside data centres (claim c15). Nothing separates AI’s part.
  • line composition: Data centres also run enterprise and general cloud work, so any data-centre amount would be a ceiling on the AI part. The retail lines that would hold existing data-centre usage, commercial revenue of inside EU&I revenue of , hold every other customer too.
  • other: Weather-normal retail volumes added to EU&I revenue (claim c21); the 10-Q does not say which customers, so no part of it is credited to data centres.

Quotes

“We have now secured 7.8 GW of electric service agreements with data center customers.”
c1 · CFO, prepared remarks, earnings call, 2026-08-04
“Our teams are working with prospective customers to advance large load projects, and we continue to expect the remainder of the 15.4-GW pipeline to be converted to ESAs by the first half of 2027.”
c2 · CFO, prepared remarks, earnings call, 2026-08-04
“As you can see on the right side of the slide, customers are also making strong progress building their facilities, with several moving to vertical construction. We continue to expect these customers to begin taking energy as early as the second half of 2027 and into 2028 and ramp into their full contracted load through the early 2030s.”
c3 · CFO, prepared remarks, earnings call, 2026-08-04
“Our contracts ensure large users of energy pay the costs of serving their facilities, and these projects are expected to deliver billions of dollars in benefits for existing customers over time.”
c6 · CEO, prepared remarks, earnings call, 2026-08-04
“It aligns with the goal of the Ratepayer Protection Pledge, which we signed in late July, joining many of the hyperscale customers we serve.”
c7 · CEO, prepared remarks, earnings call, 2026-08-04
“As a reminder, our contracts contain minimum take provisions, which serve as the basis for revenue growth projections.”
c8 · CFO, prepared remarks, earnings call, 2026-08-04
“Demand associated with data center development remains a significant contributor to projected load growth.”
c11 · Filing, mdna, 10-Q periodic report, 2026-08-04
“We continue to expand our portfolio of data center electric service agreements, increasing contracted capacity while maintaining a disciplined approach to infrastructure investment.”
c12 · Filing, mdna, 10-Q periodic report, 2026-08-04
“Beyond data center activity, we continue to see strong interest from a diverse set of commercial and industrial sectors, including life sciences and advanced manufacturing.”
c15 · CFO, prepared remarks, earnings call, 2026-08-04
“•an $84 million increase in weather-normal retail sales volumes;”
c21 · Filing, mdna, 10-Q periodic report, 2026-08-04

By quarter

  • Q1 2026described · described, no size · exploratory
  • Q2 2026described · described, no size · exploratory

Reported lines, year-over-year growth

Revenue +1.1%

Q2 2026. Growing slower than revenue: operation, maintenance and other (−16.3%), total operating expenses (−2.4%). A displaced cost shows up as a line that stays under the dashed revenue line. These are the audited lines, as first reported; nothing here is attributed to AI by the filing.

Operation, maintenance and otherTotal operating expensesRevenue
-20%0%20%40%Q1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026RevenueTotal operating expensesOperation, maintenance and other
Reported values and filings
LineQ1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026
Total operating revenues$8.25bn$7.51bn$8.54bn$7.94bn$9.18bn$7.59bn
Operation, maintenance and other$1.50bn$1.66bn$1.76bn$1.78bn$1.75bn$1.39bn
Total operating expenses$5.91bn$5.69bn$6.22bn$5.83bn$6.84bn$5.55bn