RuncheyResearch

AI Absorption Ledger / HD

Home Depot

HD · Q3 2026 · reported 2026-08-18 · revenue $47.86bn

Assessment

Magic Apron returns to the call and moves into the stores. Associates and customers now use it in the aisles in a version that knows the store, and the head of interconnected retail gives a monthly question count that is still growing; no sales, session share or labor measure comes with it, and the call does not call it AI. The ledger keeps its first-quarter assumptions for sales caused, , and raises the share of associate hours for the in-aisle version: associate labor saved .

The head of Pro says Pros purchase more with the AI-powered material list builder, a direction without a number, so the Pro tools move from described to directional; the ledger's estimate is . Search features, the tooling bill and the competitive risk go unmentioned; the release and the 10-Q have no AI passage.

The quarter's results are explained without AI: net sales grew with GMS and Mingledorff's, gross margin turned on tariff refunds, and the SG&A rate rose to on operating costs relative to comparable sales. An organizational realignment was announced and described without AI or a cost target.

Sized channels against the income statement, Q3 2026

3 of 6 channels sized

Each blue mark is one channel's dollars for the quarter; a bar is the range of an estimate. Grey marks are the company's reported lines. The distance between them is the point: how large the AI channel is next to the line it sits in.

New money and old money, Q3 2026

2 new3 expanded1 relabelled

Each channel is tagged once for whether its money existed before language models, from the company's annual report and call at the start of the period. A bar splits one flow's sized dollars by that tag. The incremental total is the part that would not be there without the models: a new channel counts in full, an expanded one only for what AI added, a relabelled one at zero.

Cost displaced by AI$76k to $17mn sized

expanded $76k to $17mn

Incremental total $76k to $17mnpoint $1.9mn
Revenue arriving through AI$1.1mn to $64mn sized

expanded $1.1mn to $64mnrelabelled not sized

Incremental total $1.1mn to $64mnpoint $11mn

The sized total counts every channel the company credits to AI, including relabelled money that existed before language models and the ledger's own estimates for it. The incremental total counts relabelled channels at zero. A flow is split by layer where its dollars sit at more than one: end use, compute sold to builders, and hardware. The same dollar can be a buyer's spend, a cloud's revenue and a chipmaker's revenue, so the layers are never added together.

Paid for AI1 channel · 1 not sized

vendor bill

AI tooling, model and partner bill

Not sized

Not mentioned this quarter.

inscrutabledisclosure: not mentioned· motive: exploratory· before LLMs: new

No source in the quarter mentions AI spending, a model partner or AI investments. The channel is carried with no reading of its own.

Evidence: 0 quotes, 2 from before coverage

What the company spends testing, developing and deploying AI tools, including generative AI, part of it with strategic partners and other third-party service providers: model and cloud bills behind Magic Apron and the Pro list tools, AI software and partner services. The 10-K names the activity among its technology investments and gives no amount; technology expense is not a reported line and sits inside selling, general and administrative expenses.

Why this motive

Carried from the prior quarter; silence is not evidence about motive.

Before LLMs: new

At the anchor the fiscal 2023 10-K already listed testing the use and incorporation of artificial intelligence, including generative artificial intelligence, among its substantial information technology investments, made in part with strategic partners and other third-party service providers, and named implementing new technologies such as artificial intelligence among the risks of its initiatives; it gave no amount and named no deployed AI tool. The fiscal 2025 10-K says testing, developing and deploying AI tools, including generative AI, and names the deployed tools. A bill for LLM models and tokens could not exist before LLMs, so the money is read as new. Technology spend sat inside selling, general and administrative expenses ( for fiscal 2023).

“including those associated with managing third-party service providers, employing new online tools and services, implementing new technologies such as artificial intelligence, implementing and restructuring support systems and processes”
Filing, risk factors, 10-K periodic report, 2024-03-13
“outsourcing certain technology and business processes to third-party service providers; making changes to existing systems, including the migration of applications to the cloud; maintaining or enhancing legacy systems that are not currently being replaced; designing or cost-effectively acquiring new systems with new functionality; or testing the use and incorporation of artificial”
Filing, risk factors, 10-K periodic report, 2024-03-13

By quarter

  • Q1 2026described · our inference · exploratory · $2.3mn to $33mn
  • Q2 2026not mentioned · inscrutable · exploratory
  • Q3 2026not mentioned · inscrutable · exploratory

Cost displaced by AI1 channel · $76k to $17mn sized · $76k to $17mn incremental

operations · cheap to verify

Store labor saved by AI tools for associates

0% to 0.03% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: described· motive: exploratory· before LLMs: expanded

Associates now use Magic Apron in the aisles to find products and answer questions, and the head of stores calls it an enabler of less friction for associates and customers; no hours, headcount or cost are given. SG&A was of net sales against a year earlier. The size shown is the ledger's own estimate, , with a higher share of associate hours than the first quarter's for the in-aisle version and the same share of saved hours leaving payroll.

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

Associate hours that AI tools may save, chiefly Magic Apron, which the 10-K says supports associates with product and project inquiries and which associates use in the aisles from 2026-CQ3. Labor sits in selling, general and administrative expenses. Sidekick and Computer Vision (machine-learning tasking and shelf visibility, described without AI), labor model tools and the transition of store tasking to the merchandising execution team are not AI by the company's own words and are read as confounds.

Why this motive

AI tool deployed in associates' work with no measured operating result: associates use Magic Apron in the aisles (claim c2) and the head of stores speaks of less friction (claim c5); the efficiency tell needs a measure, and none is given.

Before LLMs: expanded

At the anchor associates used hdPhones with applications that support customer service, and Sidekick, powered by machine learning, directed their tasks; the company employed associates and selling, general and administrative expenses were for fiscal 2023. An LLM assistant in associates' work changes output per hour, not the existence of the line. The size is the change AI made, not the whole line.

“Our associates assist these customers both in our stores and through online resources and other media designed to provide product and project knowledge.”
Filing, business, 10-K periodic report, 2024-03-13
“Our current generation of these digital “hdPhone” devices offers enhanced functionality that allows associates to readily query inventory, access applications that support customer service, and assist with locating products.”
Filing, business, 10-K periodic report, 2024-03-13
“Powered by our machine learning logic, Sidekick directs associates to key bays where on-shelf availability is low or outs exist.”
Executive, prepared remarks, earnings call, 2023-05-16

Figures

  • Selling, general and administrative expenses · 2026-CQ3
  • Selling, general and administrative expenses as a share of net sales · 2026-CQ3
  • Selling, general and administrative expenses as a share of net sales, prior year · 2025-CQ3

What else could explain it

  • transformation program: An organizational realignment was announced in the quarter, described without AI or a cost target.
  • other: Technology-enabled tools for associates are credited with engagement, satisfaction and sales without naming AI or any labor measure.
  • line composition: The 10-Q explains the SG&A rate by operating costs relative to comparable sales; marketing for the football tournament was a named expense in the quarter.

Quotes

“This, coupled with all of our investments into our associate experience through technology-enabled tools, makes it easier than ever for associates to serve customers.”
c1 · Executive, prepared remarks, earnings call, 2026-08-18
“A great example of how we're building on the momentum that we're seeing is through the evolution of Magic Apron. In addition to a website, now associates and customers can utilize this application in our aisles to navigate our stores more efficiently, find products within seconds, and ask questions about products and projects to feel more confident in their ability to complete a home improvement project.”
c2 · Executive, prepared remarks, earnings call, 2026-08-18
“You called out Michael Magic Apron, and Jordan is in the room, and I think that is a really important enabler when we think about taking friction out of the experience, both for the associate and customer.”
c5 · Executive, qa, earnings call, 2026-08-18
“As a percentage of net sales, SG&A was 17.6% for the second quarter of fiscal 2026 compared to 17.1% for the second quarter of fiscal 2025, primarily reflecting higher operating costs relative to comparable sales performance.”
c7 · Filing, mdna, 10-Q periodic report, 2026-08-25
“Yeah. Look, we're always evolving to better align our organization around our strategy, and these changes are part of what we do to evolve. They'll allow us to work smarter, move faster, and innovate more quickly.”
c8 · CFO, qa, earnings call, 2026-08-18

By quarter

  • Q1 2026described · our inference · exploratory · $35k to $10mn
  • Q2 2026not mentioned · inscrutable · exploratory
  • Q3 2026described · our inference · exploratory · $76k to $17mn

Revenue arriving through AI3 channels · $1.1mn to $64mn sized · $1.1mn to $64mn incremental · 1 not sized

search discovery · cheap to verify

Sales gained through Magic Apron, the AI assistant for customers

0% to 0.08% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: direction only· motive: exploratory· before LLMs: expanded

Magic Apron now runs in the aisles as well as online, in a localized version that uses the store's context and inventory. The head of interconnected retail gives a monthly question count and says use keeps growing; a count of questions is not a sales measure. Online sales were of net sales, up . The size shown is the ledger's own estimate, , on the first quarter's assumptions. Magic Apron is not called AI on this call; the 10-K's description of it as AI-powered is what makes it a channel.

Evidence: 5 quotes, 3 figures, 2 confounds, 2 from before coverage

Sales gained because customers ask Magic Apron, the company's AI-powered assistant on its website and app and, from 2026-CQ3, in the aisles, product and project questions. The assistant carries no price; the money is the increment in online and store sales it causes. The payers are households (DIY and DIFM customers) and Pros. Associates' own use of the tool is read under associate-ai-tools.

Why this motive

A count of use and a growth direction (claim c3), with no sales attributed; the traffic and conversion gain is not credited to the assistant (claim c9). AI product with no measure of money reads exploratory.

Before LLMs: expanded

At the anchor associates already answered customers' product and project questions in stores and through online resources, and the company credited search, site functionality and product content with traffic and conversion; online sales were of net sales of in fiscal 2023, with no assistant. An LLM assistant changes how existing questions are answered and orders are built, so the AI part is the increment it causes, not the line. The size is the change AI made, not the whole line.

“We believe our focus on improving search capabilities, site functionality, category presentation, product content, speed to checkout, and fulfillment options has yielded higher traffic, better conversion and continued sales growth.”
Filing, business, 10-K periodic report, 2024-03-13
“Our associates assist these customers both in our stores and through online resources and other media designed to provide product and project knowledge.”
Filing, business, 10-K periodic report, 2024-03-13

Figures

  • Online sales as a share of net sales · 2026-CQ3
  • Online sales growth, year over year · 2026-CQ3
  • Net sales · 2026-CQ3

What else could explain it

  • bundling: Magic Apron is part of the website, the app and now the in-store experience at no price.
  • line composition: Online growth is told through traffic, conversion, the app and delivery speed (Express Delivery, shorter lead times); none of it is credited to Magic Apron.

Quotes

“A great example of how we're building on the momentum that we're seeing is through the evolution of Magic Apron. In addition to a website, now associates and customers can utilize this application in our aisles to navigate our stores more efficiently, find products within seconds, and ask questions about products and projects to feel more confident in their ability to complete a home improvement project.”
c2 · Executive, prepared remarks, earnings call, 2026-08-18
“Then on Magic Apron, we're getting millions of questions per month now on Magic Apron. It's continued to grow. It's been a great tool for our customers. We've had great feedback.”
c3 · Executive, qa, earnings call, 2026-08-18
“Ann's references this morning were really to a more localized version of Magic Apron that understands the store context and where you are in the store because the nature of your questions and the nature of the help that we can give you is better if we understand, hey, you're in this store.”
c4 · Executive, qa, earnings call, 2026-08-18
“You called out Michael Magic Apron, and Jordan is in the room, and I think that is a really important enabler when we think about taking friction out of the experience, both for the associate and customer.”
c5 · Executive, qa, earnings call, 2026-08-18
“I mean, Michael, you referenced online, and to Billy's comments, online grew 11% in the quarter, and we had an increase in traffic and an increase in conversion.”
c9 · Executive, qa, earnings call, 2026-08-18

By quarter

  • Q1 2026described · our inference · exploratory · $607k to $29mn
  • Q2 2026not mentioned · inscrutable · exploratory
  • Q3 2026direction only · our inference · exploratory · $794k to $38mn

product ranking · cheap to verify

AI in search, recommendations, chat and review summaries on the digital platforms

Not sized

Not mentioned this quarter.

inscrutabledisclosure: not mentioned· motive: exploratory· before LLMs: relabelled

The head of Pro says the company continues to invest in search, without naming AI; the channel is carried with no reading of its own.

Evidence: 0 quotes, 1 from before coverage

Online sales gained because the company uses AI tools to improve search, recommendations and chat features and to summarize customer product reviews on its website and app. Distinct from Magic Apron, the named assistant. The payers are households and Pros.

Why this motive

Carried from the prior quarter; silence is not evidence about motive.

Before LLMs: relabelled

At the anchor the fiscal 2023 10-K already credited improving search capabilities, site functionality, category presentation and product content with higher traffic and conversion, and online sales were of net sales. The fiscal 2025 10-K repeats the same list with product review summaries added and AI named as the means, and measures nothing AI moves, so the activity is read as renamed.

“We believe our focus on improving search capabilities, site functionality, category presentation, product content, speed to checkout, and fulfillment options has yielded higher traffic, better conversion and continued sales growth.”
Filing, business, 10-K periodic report, 2024-03-13

By quarter

  • Q1 2026described · our inference · exploratory · $2.4mn to $55mn
  • Q2 2026not mentioned · inscrutable · exploratory
  • Q3 2026not mentioned · inscrutable · exploratory

search discovery · cheap to verify

Pro sales through the AI material list builder and Blueprint Takeoffs

0% to 0.05% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: direction only· motive: exploratory· before LLMs: expanded

The head of Pro says Pros build material lists and purchase more with the AI-powered material list builder, with no measure. The size shown is the ledger's own estimate, , on the first quarter's assumptions.

Evidence: 1 quote, 2 figures, 2 confounds, 1 from before coverage

Sales to Pros (contractors, remodelers, builders) gained because AI-powered tools generate project-specific materials lists and quotes: the Blueprint Takeoffs tool, which reads a blueprint, and the material list builder, which pre-populates a list from a typed project description. Both sit in the Pro app and digital workspace at no price; the money is the increment in Pro sales they cause. The project planning tool, which management does not call AI, is a confound.

Why this motive

Pros purchase more with the tool, a direction with no level or comparison base (claim c6); AI named with no measure of money reads exploratory.

Before LLMs: expanded

At the anchor the 10-K said Pros look for project planning tools for complex purchases, and online sales were of net sales; materials lists and quotes were built by Pros and sales associates without an AI tool. The tools change how existing Pro orders are built, so the AI part is the increment they cause. The size is the change AI made, not the whole line.

“Our Pros also look for dedicated sales support, competitive credit and pricing options, project planning tools, and product depth and job lot quantities, particularly for their complex purchase needs.”
Filing, business, 10-K periodic report, 2024-03-13

Figures

  • Online sales as a share of net sales · 2026-CQ3
  • Online sales growth, year over year · 2026-CQ3

What else could explain it

  • mix shift: Pros who adopt a list tool are the more engaged ones; purchasing more with it may be selection.
  • other: Outsized online growth with Pros is credited to the B2B experience and the project planning tool, which is not called AI.

Quotes

“We are also seeing Pros build material lists and purchase more with the use of our AI-powered material list builder.”
c6 · Executive, qa, earnings call, 2026-08-18

By quarter

  • Q1 2026described · our inference · exploratory · $243k to $19mn
  • Q2 2026described · our inference · exploratory · $276k to $22mn
  • Q3 2026direction only · our inference · exploratory · $318k to $25mn

Cost imposed, or revenue lost, by others’ AI1 channel · 1 not sized

distribution

Sales and margin lost to AI shopping tools

Not sized

Not mentioned this quarter; lost demand has no line.

inscrutabledisclosure: not mentioned· motive: imposed· before LLMs: new

The 10-Q and the call do not raise AI shopping tools as a competitive risk. The channel is carried with no reading of its own.

Evidence: 0 quotes, 1 from before coverage

Sales and margin that may be lost as AI tools ease competitive entry, price transparency and comparison shopping, and as generative and agentic AI tools change how customers find, compare and purchase products. The 10-K names the risk; there is no line and no figure.

Why this motive

Carried: toll channel, the company did not choose the risk.

Before LLMs: new

At the anchor the 10-K said the internet facilitates competitive entry, price transparency and comparison shopping, and named no AI tool; demand lost to AI shopping tools had no line or figure then or since. The covered 10-K's comparison-shopping sentence is the anchor's sentence with AI tools added beside online capabilities; the new part is the risk that generative and agentic AI tools change how customers shop and make purchases.

“The internet facilitates competitive entry, price transparency, and comparison shopping, increasing the level of competition we face.”
Filing, business, 10-K periodic report, 2024-03-13

By quarter

  • Q1 2026described · inscrutable · imposed
  • Q2 2026not mentioned · inscrutable · imposed
  • Q3 2026not mentioned · inscrutable · imposed

Reported lines, year-over-year growth

Revenue +5.7%

Q3 2026. Growing slower than revenue: cost of sales (+5.3%). 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.

Cost of salesSelling, general and administrativeDepreciation and amortizationTotal operating expensesRevenue
-5%0%5%10%15%20%Q1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026Q3 2026Selling, general and administrativeTotal operating expensesDepreciation and amortizationRevenueCost of sales
Reported values and filings
LineQ2 2025Q3 2025Q4 2025Q1 2026Q2 2026Q3 2026
Net sales$39.86bn$45.28bn$41.35bn$38.20bn$41.77bn$47.86bn
Cost of sales$26.40bn$30.15bn$27.54bn$25.73bn$27.98bn$31.75bn
Selling, general and administrative$7.53bn$7.76bn$7.64bn$7.77bn$7.96bn$8.42bn
Depreciation and amortization$796mn$806mn$826mn$845mn$841mn$852mn
Total operating expenses$8.33bn$8.57bn$8.46bn$8.62bn$8.80bn$9.28bn