AI Absorption Ledger / VZ

Verizon

VZ · Q2 2026 · reported 2026-07-24 · revenue $34.25bn

Assessment

The Q2 call moves the AI story from operations to revenue. The CEO returns to AI Connect, first launched under that name on the 2024-CQ4 call: an agreement with Google valued at over for dark fiber between its data centers, other deals expected by year end, and central offices being retrofitted as edge data centers for inference. Management says the new deals' revenue begins next year, on top of AI Connect revenue booked since 2024-CQ4 and called small then. The release names AI infrastructure revenue for the first time; the 10-Q mentions AI only in its list of network technologies.

On the cost side the call is thinner than Q1. AI models fixing network issues in minutes, investment in the AI tech stack, and a general tie between becoming AI-centric and the operating leverage are the only AI statements; care voice agents, the coding tool, energy savings, micro-segmentation and the vendors are not mentioned, and the general tie is not read as a care saving. Every cost movement the 10-Q explains is credited to workforce reduction initiatives (severance charges of ), Frontier, asset rationalization or one-time charges, and the transformation program targets at least of savings with no AI share.

Steps from Q1: fiber for AI infrastructure moved from described to bounded; network energy moved from bounded to not mentioned; care, the coding tool, micro-segmentation and the vendor bill moved from described to not mentioned; edge inference space and AI infrastructure capital spending opened. Every size remains the ledger's own: network operations at , now on a stated rate of the work, the stack build at , and AI Connect revenue at . Edge inference space is left unsized because its revenue has not started, and AI infrastructure capital spending because the fiber build names AI beside other growth with nothing to separate its part; both sit on the facilities layer.

Sized channels against the income statement, Q2 2026

3 of 10 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, Q2 2026

1 new7 expanded2 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.

Paid for AI$4.4mn to $94mn sized

new not sizedexpanded $4.4mn to $94mn

Incremental total $0 to $94mnpoint $0$22mn in 1 channel has no traced baseline
Cost displaced by AI$1.8mn to $54mn sized

expanded $1.8mn to $54mnrelabelled not sized

Incremental total $1.8mn to $54mnpoint $14mn
Revenue arriving through AI$716k to $36mn sized

expanded $716k to $36mn

Incremental total $0 to $36mnpoint $0$7.2mn in 1 channel has no traced baseline

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 AI3 channels · $4.4mn to $94mn sized · $0 to $94mn incremental · 2 not sized

vendor bill

AI model, agent and tooling bill

Not sized

The Q2 sources name no AI vendor and no bill; the CEO's mention of investing in the AI tech stack is read on the build channel.

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

No AI vendor is named on the Q2 call or in the filings.

Evidence: 0 quotes

What the company pays outside AI vendors: model access from Google and Anthropic (including the Mythos model used in cybersecurity under Glasswing) and voice-agent platforms (Sierra, ElevenLabs, Google). The CEO says the AI tech stack is built from best-of-breed vendors rather than the company's own. The AI coding tool's vendor is not stated: it appears in the stored transcript as Quadcode, possibly a transcription of a third-party product name, and is left out of the claims on this channel. No source gives an amount; the bill would sit in cost of services and selling, general and administrative expense.

Why this motive

Exploratory, carried from Q1; silence is not evidence about motive.

Before LLMs: new

A bill for model access, tokens and agent platforms could not exist without large language models. The anchor 10-K and call name no AI vendor and no such bill.

By quarter

  • Q1 2026described · our inference · exploratory · $3.4mn to $52mn
  • Q2 2026not mentioned · inscrutable · exploratory

engineering

Internal investment in the AI tech stack

0.01% to 0.27% of the quarter’s revenue

Incremental total: counts at zero at the point and in full at the high end, with no traced baseline.

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

The CEO says the company is investing in its AI tech stack, with no amount or progress measure; the Q1 call had said it would be substantially complete by July. The size shown is the ledger's own estimate, .

Evidence: 1 quote, 1 figure, 1 confound, 2 from before coverage

What the company spends building its AI tech stack (a data and intelligence layer, an agent development factory, runtime engines and a control plane, per the CEO on the Q1 2026 call) and on the AI specialists recruited for it. Separate from the outside vendor bill. The staff and systems sit across cost of services and selling, general and administrative expense; no amount is given.

Why this motive

The CEO lists investing in the AI tech stack among the uses of cash (claim c9): the investing-for-later tell.

Before LLMs: expanded

At the anchor the Q1 2024 call said several generative AI projects were going live and the 10-K said AI was already used in network and support work; no amount was given then or in coverage. The size is the whole of an activity that existed before.

“Finally, we are using AI in our network deployment and maintenance as well as our customer and employee support services.”
Filing, risk factors, 10-K periodic report, 2025-02-12
“Enabling AI at scale for improved customer service is a key. We're also aggressively driving AI transformative potential with our businesses, something our network was built to support. We're already at several generative AI projects going live.”
CEO, prepared remarks, earnings call, 2024-04-22

Figures

  • Cost of services plus selling, general and administrative expense, excluding severance charges · 2026-CQ2

What else could explain it

  • line composition: Technology staff and systems sit across both operating lines and capitalized software; no amount is reported.

Quotes

“We're investing in our AI tech stack, we are growing our dividend, and we are paying down our debt.”
c9 · CEO, prepared remarks, earnings call, 2026-07-24

By quarter

  • Q1 2026described · our inference · exploratory · $4.4mn to $93mn
  • Q2 2026described · our inference · exploratory · $4.4mn to $94mn

engineering

Fiber capital spending for AI infrastructure customers

Not sized

The only statement names AI infrastructure builds among the fiber deployed at scale to capture growth opportunities (claim c10) and calls the capital success-based (claim c15); nothing separates AI's part of the capital (the methodology rule for capital named together with AI). The quarter's capital expenditures, , are quoted as the ceiling.

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

Opened this quarter. The CFO says fiber deployment includes AI infrastructure builds across the country. Capital expenditures were for the quarter, the ceiling the AI part sits inside; nothing separates that part, so the channel is left unsized. It is capital and would be left out of the spend total in any case.

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

Capital spending on fiber routes and central-office retrofits built for AI infrastructure customers, which the CFO says the company deploys at scale, including AI infrastructure builds, and the CEO calls success-based capital. Capitalized, so it is traced and shown and left out of flow totals (methodology, capital spending); it reaches the income statement later as depreciation. No AI share of capital expenditures is given.

Why this motive

The CEO calls the spending success-based capital tied to signed demand (claim c15), ahead of revenue: the exploratory tell.

Before LLMs: expanded

At the anchor capital expenditures, including capitalized software, were for FY2024, and the 10-K already described building fiber connectivity to data centers. On the 2024-CQ4 reference call the Business Group CEO said capital for AI Connect deals (extending fiber to data centers and retrofitting technical space) was already inside the 2025 capital envelope, with no amount given. The size is the whole of an activity that existed before.

“We continue to build local fiber networks by adding fiber connectivity to premises, venues, cell tower locations and data centers.”
Filing, business, 10-K periodic report, 2025-02-12
“And we do have the CapEx in there because, listen, some of these deals, what we'll have to do is spend some capital to edge out our fiber a little bit to go meet people at the right data centers and/or locations of business and then retrofitting technical space. So we have money in for that, but we're ready to go with that. It's all included in the envelope.”
Executive, qa, earnings call, 2025-01-24

Figures

  • Capital expenditures (including capitalized software), Q2 2026 (six months less Q1) · 2026-CQ2
  • Capital expenditures (including capitalized software), six months · 2026-01-01..2026-06-30

What else could explain it

  • line composition: Capital expenditures hold the wireless, fiber-to-the-home and IT programs; no AI share is given.
  • other: The fiber build serves other growth opportunities beside AI infrastructure (claim c10); the AI builds are not split from it.

Quotes

“We also continue to deploy fiber at scale to capture growth opportunities, including AI infrastructure builds across the country. As Dan said, we expect the amount of revenue associated with these deals to be in the billions over the next several years.”
c10 · CFO, prepared remarks, earnings call, 2026-07-24
“All of this is success-based kind of capital that we are putting in, and we see this accelerating business plus our Verizon AI Connect infrastructure providing us with a very different revenue future than we've had in a long, long time.”
c15 · CEO, qa, earnings call, 2026-07-24

Cost displaced by AI5 channels · $1.8mn to $54mn sized · $1.8mn to $54mn incremental · 4 not sized

customer support · cheap to verify

Customer care cost displaced by AI voice agents and agent tools

Not sized

The Q2 call, release and 10-Q say nothing about AI in customer care; no filing reports a care line, and the only related statement is company-level.

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

Q2 is silent on AI in care. The CEO says the transformation initiatives show up as lower costs to serve and faster customer journeys and that the operating leverage reflects becoming AI-centric (claim c3), a company-level statement about the cost program that names no function, so it is not read as a care saving. The 10-Q credits severance charges of and Business personnel decreases of to workforce reduction initiatives (claims c18 and c20). The counterparty is mixed: the company’s own care staff and outsourced care.

Evidence: 0 quotes, 2 figures, 8 from before coverage

Customer service and care cost (care staff, outsourced care and the systems behind them) that is lower than it would have been because AI voice agents and AI tools take part of the work: on the Q1 2026 call the CEO describes putting voice agents from Sierra, ElevenLabs and Google into some customer service operations, still being tested and fine-tuned, and says AI and automation in customer interactions are already lowering costs. The filings split aggregate customer service costs between cost of services and selling, general and administrative expense and report no care line. Verizon's care is done by its own staff and by outsourcers (cost of services holds costs to support outsourcing contracts), and no source splits the two, so the counterparty is unknown, as at Expedia and Klarna. This is the buyer side of the door the ledger reads at Concentrix and TTEC.

Why this motive

Exploratory, carried from Q1; silence is not evidence about motive.

Before LLMs: expanded

At the anchor the 10-K said the company already used AI in customer and employee support services, the Q1 2024 call named enabling AI at scale for customer service as a priority and customer care changes and outsourcing done the year before, and the 2024-CQ4 reference call described FastPass, deployed in the mobility call centers to pair customers with the best available care representative. Care costs sat inside cost of services () and selling, general and administrative expense () for FY2024, with no care line reported. The size is the change AI made, not the whole line.

“Finally, we are using AI in our network deployment and maintenance as well as our customer and employee support services.”
Filing, risk factors, 10-K periodic report, 2025-02-12
“We are focused on achieving profitable growth as we continue to deliver strong revenues and undertake initiatives to reduce costs and improve efficiencies, including through AI-driven technologies.”
Filing, mdna, 10-K periodic report, 2025-02-12
“Enabling AI at scale for improved customer service is a key. We're also aggressively driving AI transformative potential with our businesses, something our network was built to support. We're already at several generative AI projects going live.”
CEO, prepared remarks, earnings call, 2024-04-22
“Secondly, many of the larger sort of transactions and platform transactions we started already last year, the outsourcing, the HCL, the customer care changes we did, which are large transactions. Without any interruptions for our customers, we have done those.”
CEO, qa, earnings call, 2024-04-22
“Selling, general and administrative expense includes salaries and wages and benefits not directly attributable to a service or product, the provision for credit losses, taxes other than income taxes, advertising and sales commission costs, call center and information technology costs, regulatory fees, professional service fees and rent and utilities for administrative space.”
Filing, mdna, 10-K periodic report, 2025-02-12
“Aggregate customer service costs, which include billing and service provisioning, are allocated between Cost of services and Selling, general and administrative expense.”
Filing, mdna, 10-K periodic report, 2025-02-12
“A good example would be FastPass, which I've talked about before, which is deployed in our mobility call centers to intelligently pair customers with the best available care representative.”
CEO, prepared remarks, earnings call, 2025-01-24
“And then Hans mentioned the focus on cost transformation and the work we continue to do to make the business more efficient, whether it's customer care and the work that Sampath is doing, including AI, and you heard that upfront, as well as the managed services work.”
Executive, qa, earnings call, 2025-01-24

Figures

  • Severance charges · 2026-CQ2
  • Decrease in Business segment personnel costs from workforce reduction initiatives · 2026-CQ2

By quarter

  • Q1 2026described · described, no size · exploratory
  • Q2 2026not mentioned · inscrutable · exploratory

engineering · cheap to verify

Software delivery and vendor support cost displaced by AI coding tools

Not sized

The Q2 call, release and 10-Q say nothing about AI in software development or vendor support, and no filing reports a software or vendor-support line.

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

The Q1 call described deploying an AI coding tool across the software lifecycle with a vendor-support cut as a target (c17); Q2 is silent on it.

Evidence: 0 quotes, 3 from before coverage

The cost of software development and of outside vendor support for it that is lower because the company deploys an AI coding tool (named on the Q1 2026 call as Quadcode) across the software development lifecycle. The CEO says the company spends heavily on vendor support there and sees a way to cut that cost by more than a stated share with AI, a target rather than a measured saving. Information technology costs sit in selling, general and administrative expense and are not split out. The bill for the tool itself is in the AI vendor bill.

Why this motive

Exploratory, carried from Q1; silence is not evidence about motive.

Before LLMs: expanded

At the anchor the Q1 2024 call named large outsourcing and platform transactions done the year before, the HCL agreement among them, and the 10-K placed information technology costs and professional service fees inside selling, general and administrative expense ( for FY2024), with no software or vendor-support line. The size is the change AI made, not the whole line.

“We are focused on achieving profitable growth as we continue to deliver strong revenues and undertake initiatives to reduce costs and improve efficiencies, including through AI-driven technologies.”
Filing, mdna, 10-K periodic report, 2025-02-12
“Secondly, many of the larger sort of transactions and platform transactions we started already last year, the outsourcing, the HCL, the customer care changes we did, which are large transactions. Without any interruptions for our customers, we have done those.”
CEO, qa, earnings call, 2024-04-22
“Selling, general and administrative expense includes salaries and wages and benefits not directly attributable to a service or product, the provision for credit losses, taxes other than income taxes, advertising and sales commission costs, call center and information technology costs, regulatory fees, professional service fees and rent and utilities for administrative space.”
Filing, mdna, 10-K periodic report, 2025-02-12

By quarter

  • Q1 2026described · our inference · exploratory · $0 to $87mn
  • Q2 2026not mentioned · inscrutable · exploratory

operations · expensive to verify

Network operations work done by AI models

0.01% to 0.16% of the quarter’s revenue

Incremental total: counts in full.

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

The CEO says AI models now fix network issues autonomously in minutes rather than hours, with network metrics improving each month since the January outage. Repair time cut from hours to minutes is a measured change management attributes to AI, which with the Q1 kit reduction makes the channel expanded. Cost of services rose to from . The size shown is the ledger's own estimate, .

Evidence: 6 quotes, 3 figures, 3 confounds, 4 from before coverage

Network operations cost (fault resolution, deployment planning, field dispatch and the equipment kits behind them) that is lower because AI models resolve network issues autonomously and simplify deployment. The CEO says most network issues are now resolved autonomously and, on the Q2 2026 call, that AI models fix issues in minutes rather than hours. Network operating cost sits in cost of services. An error here is an outage, so the work is read as expensive to verify. Network energy savings are a separate channel.

Why this motive

The CEO describes AI models fixing network issues in minutes (claim c1), framed as reliability after the January outage rather than cost. No cost line shrinks and the 10-Q explains cost of services without AI, which does not contradict the CEO; AI named with no cost line moving reads exploratory, as in Q1.

Before LLMs: expanded

At the anchor the 10-K said the company already used AI in network deployment and maintenance (anchor claim vz-anchor-c1), and the Q1 2024 call described AI used for network performance and capacity deployment (anchor claim vz-anchor-c9). Network operating cost sat in cost of services ( for FY2024). In coverage management gives measured changes it attributes to AI: fault repair from hours to minutes and the network bill of materials cut to a few kits, so the activity expanded rather than only being renamed. The size is taken as an increment against the prior-year rate of the work, because the estimates measure cost removed from network operations, not the whole function. The size is the change AI made, not the whole line.

“Finally, we are using AI in our network deployment and maintenance as well as our customer and employee support services.”
Filing, risk factors, 10-K periodic report, 2025-02-12
“We are focused on achieving profitable growth as we continue to deliver strong revenues and undertake initiatives to reduce costs and improve efficiencies, including through AI-driven technologies.”
Filing, mdna, 10-K periodic report, 2025-02-12
“Our AI strategy focus on three priorities. First, optimizing internal processes and operation through machine learning, such as creating efficiencies in fuel consumption. AI is already central to our cost transformation program and will become even more important over time.”
CEO, prepared remarks, earnings call, 2024-04-22
“We're already using, for example, personalization in my Plan with AI, and we do using it in our in our network when it comes to performance or the capacity deployment as well as power consumption. So we are using AI and generative AI already now commercially.”
CEO, qa, earnings call, 2024-04-22

Figures

  • Cost of services · 2026-CQ2
  • Cost of services, prior year (same row, 2025 column) · 2025-CQ2
  • Cost of services excluding asset rationalization charges · 2026-CQ2

What else could explain it

  • transformation program: Asset rationalization of network and real estate assets under the transformation initiatives (, claim c21) and lower third-party access costs move network cost without AI.
  • acquisition: Frontier adds network staff and facilities; building and facility costs rose (claim c22).
  • relabel: The anchor 10-K already said AI was used in network deployment and maintenance (anchor claim vz-anchor-c1); part of what is described may be the existing capability under the AI name.

Quotes

“Following the actions we took after the January event, our network performance metrics have improved sequentially every month as we embed sophisticated AI models that allow us to autonomously fix network issues in minutes as opposed to hours.”
c1 · CEO, prepared remarks, earnings call, 2026-07-24
“Fourth, the transformation workstreams I described last quarter are producing tangible results, and we are fully on track to deliver at least the $9 billion of OpEx and CapEx savings we said we would.”
c2 · CEO, prepared remarks, earnings call, 2026-07-24
“They are showing up as lower costs to serve, faster customer journeys, and a step change in productivity across the organization. We are rapidly becoming an AI-centric company in how we operate, and the operating leverage we are experiencing in our results reflect that.”
c3 · CEO, prepared remarks, earnings call, 2026-07-24
“Our networks leverage advanced technologies, including 5G wireless, fiber-based transport, cloud infrastructures, artificial intelligence (AI) and automation, private networks and IP routing solutions.”
c17 · Filing, mdna, 10-Q periodic report, 2026-07-31
“During both the three and six months ended June 30, 2026, we recorded pre-tax asset rationalization charges of $258 million predominately related to the decision to cease use of certain real estate and network assets as part of our transformation initiatives.”
c21 · Filing, mdna, 10-Q periodic report, 2026-07-31
“an increase of $79 million in building and facility costs primarily due to maintaining additional buildings and facilities associated with the acquisition of Frontier in 2026 along with higher utility rates”
c22 · Filing, mdna, 10-Q periodic report, 2026-07-31

By quarter

  • Q1 2026direction only · described, no size · exploratory
  • Q2 2026described · our inference · exploratory · $1.8mn to $54mn

operations · cheap to verify

Network energy cost saved by AI optimization

Not sized

The Q2 sources give no energy saving from AI; the only figure, on the Q1 call, had no period, and energy is not reported on its own.

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

Q2 is silent on AI in network energy. The 10-Q again names higher utility rates and Frontier facilities behind a rise in building and facility costs of , a confound on any energy saving (claim c22).

Evidence: 0 quotes, 1 figure, 4 from before coverage

Electricity and fuel cost for the network that is lower because AI optimizes energy use. On the Q1 2026 call the CEO gives a stated dollar amount of energy savings from deploying AI in the network, with no period. The cost sits in building and facility costs inside cost of services, which the 10-Qs say rose on higher utility rates and Frontier facilities.

Why this motive

Exploratory, carried from the corrected Q1 reading; silence is not evidence about motive.

Before LLMs: relabelled

At the anchor the Q1 2024 call named machine learning that created efficiencies in fuel consumption as the first AI priority and AI used for network power consumption, and the 10-K described energy-efficiency initiatives for networks, facilities and fleet. The covered quarters give a saving under the AI name for an activity the anchor already showed as machine learning; no filing line is shown to move because of it.

“We are focused on achieving profitable growth as we continue to deliver strong revenues and undertake initiatives to reduce costs and improve efficiencies, including through AI-driven technologies.”
Filing, mdna, 10-K periodic report, 2025-02-12
“Our AI strategy focus on three priorities. First, optimizing internal processes and operation through machine learning, such as creating efficiencies in fuel consumption. AI is already central to our cost transformation program and will become even more important over time.”
CEO, prepared remarks, earnings call, 2024-04-22
“We're already using, for example, personalization in my Plan with AI, and we do using it in our in our network when it comes to performance or the capacity deployment as well as power consumption. So we are using AI and generative AI already now commercially.”
CEO, qa, earnings call, 2024-04-22
“We are balancing increased energy needs from ongoing expansion and densification of our networks with initiatives to effectively manage energy consumption, including using more energy-efficient equipment, discontinuing or migrating legacy services and pursuing opportunities to use renewable energy sources.”
Filing, business, 10-K periodic report, 2025-02-12

Figures

  • Increase in building and facility costs in cost of services (Frontier facilities and utility rates) · 2026-CQ2

By quarter

  • Q1 2026bounded · our inference · exploratory · $17mn to $75mn
  • Q2 2026not mentioned · inscrutable · exploratory

marketing · cheap to verify

Acquisition and retention spend reduced by AI micro-segmentation

Not sized

The Q2 call credits lower acquisition and retention costs to the new value proposition, lower upgrades and better segmentation, without naming AI; no line carries acquisition and retention spend.

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

Management again reports lower costs of acquisition and retention but names segmentation, the new plans and lower upgrade volumes as causes; AI is not mentioned.

Evidence: 0 quotes, 4 from before coverage

Promotional, device-subsidy and retention spend (the cost of acquisition and the cost of retention) that is lower because AI models and the data behind them tailor offers to each customer (the internal initiative the CEO calls Every Customer Has a Name). The CEO lists the beginning of AI-enabled processes, with channel mix, less friction, better tools and modeling, and fiscally responsible offers, as the causes of a lower cost of acquisition and retention; AI is one of several named causes and gets no share.

Why this motive

Exploratory, carried from Q1; silence is not evidence about motive.

Before LLMs: relabelled

At the anchor the Q1 2024 call described AI-based personalized plan recommendations in myPlan and AI tools used in churn management to direct retention spend, and the 2024-CQ4 reference call described Segment of Me, a tool that uses AI to personalize offers and products, the same capability as the covered quarters' Every Customer Has a Name. No acquisition or retention cost line was reported then; the covered quarters describe the same capability under a wider AI programme, with no AI share of any measured movement.

“Secondly, enhancing product experiences with AI capabilities like the personalized plan recommendation on myPlan, which is producing good early results.”
CEO, prepared remarks, earnings call, 2024-04-22
“And thirdly, we're laser-focused to churn management. I mean, the team with Sampath, the AI tools we have, so, so we actually spend on the right customers when we see they have a churn.”
CEO, qa, earnings call, 2024-04-22
“We're already using, for example, personalization in my Plan with AI, and we do using it in our in our network when it comes to performance or the capacity deployment as well as power consumption. So we are using AI and generative AI already now commercially.”
CEO, qa, earnings call, 2024-04-22
“A great product here would be Segment of Me, which is a tool that uses AI to personalize customer experience with unique offers and products.”
CEO, prepared remarks, earnings call, 2025-01-24

By quarter

  • Q1 2026described · described, no size · exploratory
  • Q2 2026not mentioned · inscrutable · exploratory

Revenue arriving through AI2 channels · $716k to $36mn sized · $0 to $36mn incremental · 1 not sized

product revenue

Fiber transport sold for AI infrastructure (AI Connect)

0% to 0.1% of the quarter’s revenue

Incremental total: counts at zero at the point and in full at the high end, with no traced baseline.

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

The CEO says the company signed an agreement with Google valued at over for dark fiber between its data centers and expects other deals worth multiple billions over several years; revenue from the new deals begins next year at margins at or above existing ones. In the Q&A the CEO puts the Google agreement at well in excess of the same amount. The release quotes the CEO on the emergence of AI infrastructure revenue. AI Connect revenue has been booked since 2024-CQ4 (anchor claims vz-anchor-c22 and vz-anchor-c25); the quarter's amount is the ledger's own estimate, . Selling capacity to the AI buildout is read as the buildout's money moving. The counterparty is mixed: hyperscalers buying capacity for AI (Google and Meta), alternative cloud providers, and large enterprises.

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

Revenue from dark and lit long-haul and metro fiber sold to hyperscalers, alternative cloud providers and large enterprises to connect data centers for AI training and inference, which the company calls AI Connect. In Q2 2026 the company signed an agreement with Google, valued above a stated amount, for dark fiber routes between its data centers, and says revenue from the new deals begins next year, on top of AI Connect revenue booked since 2024-CQ4. The revenue would sit in the Business segment, which reports no AI Connect line. Selling capacity to the buildout is read as the buildout's money moving (the capacity-seller rule in the methodology).

Why this motive

A priced agreement is signed (claim c4), the offensive tell, while the new deals' revenue is deferred to 2027 (claims c8 and c14), the exploratory tell. With the tells in conflict the less durable motive is kept.

Before LLMs: expanded

AI Connect was launched under that name on the 2024-CQ4 call with a funnel of over , largely lit, wavelength and dark fiber for hyperscalers; Google and Meta were already buying capacity for AI workloads, and the first AI Connect revenue was booked in 2024-CQ4 and called small. The FY2024 10-K said the company was adding fiber connectivity to data centers, and Enterprise and Public Sector revenue was for FY2024 and Wholesale revenue . No data-center connectivity line was reported, so no quarter of the activity before AI can be traced; the funnel is not a quarter's revenue and small is not converted. The size is the whole of an activity that existed before.

“And thirdly, establishing an AI-based revenue stream by commercializing our network's unique low latency, high bandwidth, and robust Mobile Edge Compute capabilities. Generative AI workloads represent a great long-term opportunity for us.”
CEO, prepared remarks, earnings call, 2024-04-22
“We continue to build local fiber networks by adding fiber connectivity to premises, venues, cell tower locations and data centers.”
Filing, business, 10-K periodic report, 2025-02-12
“This is creating an opportunity for us and has already created an opportunity as we had revenue and EBITDA impact in the fourth quarter. We're now looking into how we can use our assets and our capability to serve this market when it comes to the next step of generative AI. I'm very proud and excited to introduce Verizon AI Connect today.”
CEO, prepared remarks, earnings call, 2025-01-24
“You know, Verizon AI Connect is the name of our strategy and suite of offerings that are intended to meet the growing demand for AI applications from both our ecosystem partners and end user customers. It's a vision that allows us to utilize existing assets in new ways to service this technology revolution.”
Executive, prepared remarks, earnings call, 2025-01-24
“We already have a funnel of over $1 billion simply leveraging our existing infrastructure. Major players such as Google and Meta have purchased capacity on our network with the intent of using it for their AI workloads.”
Executive, prepared remarks, earnings call, 2025-01-24
“Some of these deals are reflected in our fourth quarter results and are contributing to the margin improvements you saw in the quarter.”
Executive, prepared remarks, earnings call, 2025-01-24
“It's basically lit services, Wave, and dark fiber, but that's why we're happy today to announce that we're going to add the power space and cooling to the portfolio as well.”
Executive, qa, earnings call, 2025-01-24
“and in terms of RevRec, we already booked revenue in this regard in the fourth quarter, and that was certainly helpful for me in our EBITDA trajectory. So we intend to drive more of that to the bottom line going through 2025, and we look forward to driving more business here.”
Executive, qa, earnings call, 2025-01-24
“And to Hans' point, so most of the things that we're going to be using to get into this market, things we've already done. So it's the assets we already have, tweaked a little bit.”
Executive, qa, earnings call, 2025-01-24
“I think the revenue we booked that helped with EBITDA as far as AI Connect is small. That was our first quarter that we booked revenue there, so that did have a little impact,”
Executive, qa, earnings call, 2025-01-24

Figures

  • Value of the dark fiber agreement with Google for data center connectivity (a floor; contract value over its term) · as-of 2026-07-24
  • Business segment total operating revenues · 2026-CQ2

What else could explain it

  • other: Data-center connectivity was already sold before AI (anchor claim vz-anchor-c11); the Google agreement is a contract value over an unstated term, not revenue in the quarter.
  • relabel: The CEO says "this is the first time we're talking about it publicly" (claim c12) and goes on that the first focus was the core business and that the AI infrastructure opportunity emerged behind the scenes (claim c23): the remark appears to refer to the current management's renewed initiative. AI Connect itself was launched by name on the 2024-CQ4 call with a funnel, Google and Meta as buyers and revenue booked (anchor claims vz-anchor-c15, vz-anchor-c19 and vz-anchor-c22), and the Q1 2026 call discussed it without the name (c21 and c22). Management describes it as existing assets used in new ways (anchor claims vz-anchor-c16 and vz-anchor-c23).

Quotes

“We recently signed an agreement with Google, valued at over $1 billion, to use Verizon dark fiber to connect their data centers. We have other deals that we expect to announce by year-end that, taken together, are expected to be worth multiple billions of dollars in revenue over the next several years.”
c4 · CEO, prepared remarks, earnings call, 2026-07-24
“The build-out of AI infrastructure across the United States is one of the largest capital cycles of our lifetime, Verizon is uniquely positioned to participate in it. We own one of the most extensive long-haul and metro fiber footprints in North America.”
c5 · CEO, prepared remarks, earnings call, 2026-07-24
“We expect this initiative to noticeably contribute to our revenue growth starting next year, to grow substantially from there.”
c7 · CEO, prepared remarks, earnings call, 2026-07-24
“That acceleration is independent of the incremental AI infrastructure revenue that begins to layer into our results starting in 2027.”
c8 · CEO, prepared remarks, earnings call, 2026-07-24
“We also continue to deploy fiber at scale to capture growth opportunities, including AI infrastructure builds across the country. As Dan said, we expect the amount of revenue associated with these deals to be in the billions over the next several years.”
c10 · CFO, prepared remarks, earnings call, 2026-07-24
“We have, as we mentioned in the remarks, that we have some early wins on AI Connect that Dan will get into in a second. That's an opportunity for us above the core business, and we expect it to become meaningful in 2027.”
c11 · CFO, qa, earnings call, 2026-07-24
“Mike, let me go into what we call AI Connect, our initiative inside the company. Look, this is the first time we're talking about it publicly.”
c12 · CEO, qa, earnings call, 2026-07-24
“They've got margins that are equal to or greater than our existing margin structures, they will begin to impact our revenues and margins beginning next year and grow substantially over the next 5- 10 years.”
c14 · CEO, qa, earnings call, 2026-07-24
“All of this is success-based kind of capital that we are putting in, and we see this accelerating business plus our Verizon AI Connect infrastructure providing us with a very different revenue future than we've had in a long, long time.”
c15 · CEO, qa, earnings call, 2026-07-24
“Our core connectivity business is gaining momentum, and with the emergence of AI infrastructure revenue, we are fundamentally reshaping Verizon’s growth trajectory.”
c16 · CEO, press release, 8-K earnings release, 2026-07-24
“Behind the scenes, it was clear that there was a kind of once in a generation opportunity for Verizon to participate in the massive AI infrastructure build-out. As we talked to hyperscalers, alternative cloud providers, enterprises, there is a need for ever-increasing compute power.”
c23 · CEO, qa, earnings call, 2026-07-24

By quarter

  • Q1 2026described · our inference · exploratory · $742k to $37mn
  • Q2 2026bounded · our inference · exploratory · $716k to $36mn

product revenue

Central offices retrofitted as edge data centers for AI inference

Not sized

The money has not started: the retrofits are in the early stages with conversations underway with partners (claim c6), the new AI infrastructure revenue begins to layer into results starting in 2027 (claim c8), and the only use so far is a small trial (claim c13), with no revenue stated. A project before its first delivery is unsized (the methodology rule for money not started).

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

Opened this quarter. The CEO says central offices freed of copper are being retrofitted as power-ready edge data centers for AI inference and that a small trial sold out within a day. No revenue or capacity is given, the retrofits are in early stages and AI infrastructure revenue begins in 2027, so the channel is left unsized.

Evidence: 3 quotes, 1 figure, 1 confound, 6 from before coverage

Revenue from power-ready, permitted space in central offices, retrofitted as remote data centers for AI inference at the edge, which the CEO says sold out in a small trial. A different product from fiber transport: space and power rather than connectivity. No line or amount is given.

Why this motive

Early-stage retrofits and a trial (claims c6 and c13): the exploratory tell.

Before LLMs: expanded

At the anchor the Q1 2024 call said the company already had mobile edge compute in many of its sites, built for compute and storage at the edge. The 2024-CQ4 reference call offered spare power, space and cooling in thousands of telco facilities under AI Connect, said power, space and cooling would be added to the portfolio, and announced a GPU-as-a-service provider deploying in a Verizon data center. No edge compute or facilities revenue line was reported then or in coverage. The size is the whole of an activity that existed before.

“And thirdly, establishing an AI-based revenue stream by commercializing our network's unique low latency, high bandwidth, and robust Mobile Edge Compute capabilities. Generative AI workloads represent a great long-term opportunity for us.”
CEO, prepared remarks, earnings call, 2024-04-22
“And we already have mobile edge compute in many of our sites across the country in order to be able to, to meet that compute and storage.”
CEO, qa, earnings call, 2024-04-22
“You know, Verizon AI Connect is the name of our strategy and suite of offerings that are intended to meet the growing demand for AI applications from both our ecosystem partners and end user customers. It's a vision that allows us to utilize existing assets in new ways to service this technology revolution.”
Executive, prepared remarks, earnings call, 2025-01-24
“We have thousands of distributed telco facilities, many of which already have power space and cooling available for this compute at the edge.”
Executive, prepared remarks, earnings call, 2025-01-24
“We have facilities across the United States that either have spare power, space, and cooling, or can be retrofitted. As we sit here today, we have two to 10+ MW of usable power across many of our sites.”
Executive, prepared remarks, earnings call, 2025-01-24
“It's basically lit services, Wave, and dark fiber, but that's why we're happy today to announce that we're going to add the power space and cooling to the portfolio as well.”
Executive, qa, earnings call, 2025-01-24

Figures

  • Business segment total operating revenues · 2026-CQ2

What else could explain it

  • relabel: The anchor call already described mobile edge compute in many sites, built for edge compute and storage (anchor claim vz-anchor-c10).

Quotes

“We are also in the early stages of retrofitting many of our central offices into data centers for inference edge computing, with multiple conversations underway with partners who are eager to utilize these power-ready and permitted locations.”
c6 · CEO, prepared remarks, earnings call, 2026-07-24
“That acceleration is independent of the incremental AI infrastructure revenue that begins to layer into our results starting in 2027.”
c8 · CEO, prepared remarks, earnings call, 2026-07-24
“We have thousands of central offices, many of which we're taking copper out of, and we are retrofitting them to be remote data centers that are power ready, permitted, fully redundant infrastructure. We did a small trial on that and sold out capability in 24 hours, so we're seeing large demand for that as well.”
c13 · CEO, qa, earnings call, 2026-07-24

Reported lines, year-over-year growth

Revenue −0.7%

Q2 2026. No tracked cost line grew slower than revenue. 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. Not drawn: severance charges, which one-time items move by more than 60% in a quarter; the values are in the table below.

Selling, general and administrative expenseTotal operating expensesRevenue
-30%-20%-10%0%10%20%30%Q1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026Selling, general and administrative expenseTotal operating expensesRevenue
Reported values and filings
LineQ1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026
Total operating revenues$33.48bn$34.50bn$33.82bn$36.38bn$34.44bn$34.25bn
Selling, general and administrative expense$7.87bn$7.81bn$7.75bn$10.38bn$7.63bn$8.98bn
Severance chargesn/a$0$0$1.72bnn/a$397mn
Total operating expenses$25.51bn$26.33bn$25.72bn$31.38bn$26.20bn$27.07bn