AI Absorption Ledger / EQIX

Equinix

EQIX · Q2 2026 · reported 2026-07-29 · revenue $2.63bn

Assessment

Revenues were , against a year earlier, lifted by one-time xScale fees: the company closed megawatts of xScale leases, including Hampton, which contributed about . Recurring revenues were and annualized gross bookings . The CEO says the vast majority of the largest deals were driven by AI workloads, similar to prior quarters, and that a significant proportion of demand is enterprises modernizing on-premises infrastructure, with AI-native workloads the remainder. No statement gives a dollar share, and a share of deals by count separates nothing in dollars, so the AI part of retail revenue stays directional and unsized, with recurring revenues and bookings as its ceiling.

The build accelerated: the 2026 capital expenditure plan rose to between and , against in Q1, and management plans up to a year through 2029, funded from retained cash flow and debt. Management calls AI an accelerant of demand that is mostly enterprise modernization. The same 10-Q adds a warning that a broad slowdown or correction in AI-related investment could reduce demand and affect the growth plans. The xScale fees, which the 10-Q again credits to cloud and AI together, are described and unsized, with the services revenue of as the ceiling. Capital spending, here and in the xScale ventures, is described and unsized for the same reason, traced and left out of flow totals.

What changed: the Distributed AI Hub, launched in Q1, is not mentioned; Fabric Intelligence is listed among Fabric capabilities with no price. On its own AI use the company says more than in Q1: the new CFO expects automating processes with AI, beside functionalization and standardization, to help margins, and the 10-Q adds that its use of AI may become too costly. No saving is measured.

The only sizes are the ledger’s own inference on the company’s AI tools, a saving of and a bill of ; every facilities channel is unsized. The company sells facilities under other companies’ compute, and its hyperscale and neocloud customers may be on this ledger, where the same dollars are spend; the ledger records each company’s own flows without netting them.

Sized channels against the income statement, Q2 2026

2 of 8 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 new6 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.

Paid for AI$61k to $1.3mn sized

new $61k to $1.3mnexpanded not sized

Incremental total $61k to $1.3mnpoint $327k
Cost displaced by AI$351k to $11mn sized

expanded $351k to $11mn

Incremental total $351k to $11mnpoint $2.1mn

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 · $61k to $1.3mn sized · $61k to $1.3mn incremental · 2 not sized

other

Equity put into the xScale joint ventures that build hyperscale data centers off the balance sheet

Not sized

AI named beside another cause: the 10-Q credits hyperscale demand to the largest cloud service providers and to demand ‘driven in part by the adoption of AI’ (c18, c19), and nothing separates AI’s part. The ceiling is the quarter’s purchases of equity investments, , against a carrying value of .

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

Purchases of equity investments were in the quarter, in the first half, and the xScale ventures were carried at . The company holds of each venture other than AMER 3, and an effective of the AMER 3 venture’s assets. The channel is described and unsized, with the equity purchases as the ceiling. Capitalized: traced and left out of flow totals. Funding mixed: the venture partners fund construction in proportion to their stakes (c37); the stored sources name PGIM Real Estate as a partner at the anchor and no other; the company also lends to the AMER 2 venture, a commitment of quoted and not a size (c36).

Evidence: 4 quotes, 6 figures, 2 confounds, 4 from before coverage

The company’s own equity contributions to the xScale joint ventures, which hold hyperscale data centers off its balance sheet (a minority stake in each venture; the 10-Q’s ownership percentages are quoted in the Q2 2026 reading). It is capital, traced and shown and left out of flow totals. The other side of the money is mixed: the venture partners, of which the stored sources name PGIM Real Estate (the first US venture, eqix-anchor-c8) and none of the others, the ventures’ lenders, and the company itself, which also lends to the AMER 2 venture and guarantees part of the EMEA 2 venture’s debt. The 10-Q quotes the loan commitment, future equity commitments and debt guarantees; they are commitments, not sizes. The filings name AI together with cloud as the cause of hyperscale demand and give no AI share, so the contribution is a ceiling and the channel is described and unsized. Facilities layer.

Why this motive

Capital committed to hyperscale capacity for demand the filings credit partly to AI, with no measure of AI’s part (c18, c19); carried from Q1.

Before LLMs: expanded

At the anchor the company already invested in xScale ventures, with purchases of equity investments of in 2024 and an expected global xScale portfolio of more than when built out (eqix-anchor-c7, eqix-anchor-c8). During coverage new capacity was built for hyperscale customers and put into the ventures: the Hampton campus was sold to the AMER 3 venture and the partners fund approved developments (c25, c26), so volume moved and the channel is expanded. No quarter before AI can be traced. The size is the whole of an activity that existed before.

“Hyperscalers require infrastructure to support demanding workload requirements for cloud and AI initiatives.”
Filing, business, 10-K periodic report, 2025-02-12
“In our xScale program, demand remains robust as cloud and AI needs are translating into strong pre-leasing activity.”
CEO, prepared remarks, earnings call, 2024-05-09
“To serve the needs of the growing hyperscale data center market, including the world's largest cloud service providers and increased demand driven in part by the adoption of AI, we have entered into joint venture partnership arrangements across our Americas, EMEA and Asia-Pacific regions to develop and operate xScale data centers.”
Filing, mdna, 10-K periodic report, 2025-02-12
“Additionally, in mid-April, we announced our first US xScale joint venture with PGIM Real Estate for our SV12X asset.”
CEO, prepared remarks, earnings call, 2024-05-09

Figures

  • Purchases of equity investments, first half of 2026 (cash flow statement) · 2026-01-01..2026-06-30
  • Purchases of equity investments (first half less Q1) · 2026-CQ2
  • Carrying value of the xScale joint venture investments · as-of 2026-06-30
  • Maximum commitment of the company’s loan to the AMER 2 venture · as-of 2026-06-30
  • Ownership percentage in the xScale joint ventures other than AMER 3 (investments table; every such venture is at this share, EMEA 2 row shown) · as-of 2026-06-30
  • Effective interest in the AMER 3 joint venture’s assets · as-of 2026-06-30

What else could explain it

  • other: The filings name cloud and AI together as the cause of hyperscale demand and give no share.
  • line composition: Purchases of equity investments may include investments other than the xScale ventures.

Quotes

“To serve the needs of the growing hyperscale data center market, including the world's largest cloud service providers and increased demand driven in part by the adoption of AI, we continue to look at attractive opportunities to grow our market share and selectively improve our footprint and offerings.”
c18 · Filing, mdna, 10-Q periodic report, 2026-07-29
“We believe hyperscale customers will play a large role in the growth of the market for AI.”
c19 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“with the AMER 2 Joint Venture, as a lender, with a maximum commitment of $392 million”
c36 · Filing, notes, 10-Q periodic report, 2026-07-29
“The joint ventures' partners are required to make additional equity contributions proportionately to fund capital necessary to complete the construction of approved developments.”
c37 · Filing, notes, 10-Q periodic report, 2026-07-29

By quarter

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

other

Capital expenditures on retail data center capacity, the part built for AI workloads

Not sized

AI named beside other causes and nothing separates its part: management calls AI an accelerant as customers modernize (c23, c10), and in Q1 the CEO said AI inference needs no different capital, ‘No, we don't see any difference in the capital that will be required’ (c29). The ceiling is purchases of other property, plant and equipment, , with the 2026 plan between and .

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

Purchases of other property, plant and equipment were in the quarter; management’s figure was about , mostly capacity expansion (c22). The 2026 plan rose to between and excluding xScale and real estate, from in Q1, and the plan through 2029 is up to a year (c20, c21), with about megawatts under construction. Management names AI as an accelerant beside enterprise modernization (c23, c10). Management’s plan excludes xScale and real estate; the cash line may include sites built before their transfer to a venture. In Q1 the CEO saw no difference in the capital AI inference needs (c29). The channel is described and unsized, with the cash line as the ceiling. Capitalized: traced and left out of flow totals. Funding mixed: retained cash flow and debt (c24), with equipment pre-purchased from suppliers (c42).

Evidence: 17 quotes, 8 figures, 2 confounds, 2 from before coverage

Cash paid for property, plant and equipment: new IBX data centers and expansions, built to power and cooling densities the 10-Q says AI is driving. Management’s capital expenditure figure and plan exclude xScale and land; the cash line used here excludes real estate acquisitions (a separate line) but may include sites built before their transfer to a venture. Management raised the 2026 plan and set an annual range through 2029 on stronger demand, with AI named as an accelerant beside enterprise modernization, cloud and networking, and in Q1 2026 the CEO said AI inference needs no different capital from the existing build (eqix-2026-cq1-c29). Neither the filing nor the call gives an AI share, so total capital expenditure is a ceiling and the channel is described and unsized. Capitalized: traced and shown and left out of flow totals; the income statement carries it as depreciation, which is not read as a separate channel because no AI share of the fleet is stated. The payees are construction contractors and equipment suppliers; the build is funded from retained cash flow and debt. The planned purchase of atNorth with Canada Pension Plan Investment Board is an acquisition of a data center company, not an AI company, and is a confound with nothing credited to AI. Facilities layer.

Why this motive

Investing ahead of demand the 10-Q calls an AI strategy for customers’ AI workloads, with no AI share of the spending and a new warning on an AI correction (c27, c26); carried from Q1.

Before LLMs: expanded

Purchases of other property, plant and equipment were in 2024, already partly for AI density: the 10-K described a Singapore data center designed for AI workloads with liquid cooling, and the CEO linked larger footprint deals partly to AI (eqix-anchor-c5, eqix-anchor-c3). During coverage new capacity is being built for data-centre customers (c25), so volume moved and the channel is expanded. No quarter before AI can be traced. The size is the whole of an activity that existed before.

“we did see an uptick, I will say, this quarter in larger footprint opportunities. And I think some of that is, is really associated with AI-related, you know, workloads, both service provider and enterprise.”
CEO, qa, earnings call, 2024-05-09
“This new high performance data center will feature a design built to efficiently compute intensive workloads like artificial intelligence ("AI"), supported by capabilities such as advanced liquid cooling.”
Filing, business, 10-K periodic report, 2025-02-12

Figures

  • Purchases of other property, plant and equipment (first half less Q1) · 2026-CQ2
  • Purchases of other property, plant and equipment, first half of 2026 (cash flow statement) · 2026-01-01..2026-06-30
  • Total capital expenditures in the quarter, management figure (about) · 2026-CQ2
  • Capital expenditures planned for 2026, low end of range · FY2026
  • Capital expenditures planned for 2026, high end of range · FY2026
  • Capital expenditures planned per year through 2029, high end of range · FY2027
  • Capacity under construction, megawatts (about) · as-of 2026-07-29
  • Equity the company committed to the planned atNorth purchase with Canada Pension Plan Investment Board (up to) · as-of 2026-02-26

What else could explain it

  • line composition: Capital expenditure builds capacity for all demand; the AI part is not separated. The cash line excludes real estate acquisitions but may include sites built before their transfer to a venture.
  • acquisition: The planned purchase of atNorth with Canada Pension Plan Investment Board, for which the company committed up to of equity, is of a data center company, not an AI company, and nothing of it is credited to AI. It had not closed.

Quotes

“As a result, we now expect 2026 CapEx to range between $5 billion and $6 billion.”
c20 · CEO, prepared remarks, earnings call, 2026-07-29
“To capture the robust demand in front of us, we plan to invest $5 billion-$7 billion in CapEx annually through 2029.”
c21 · CEO, prepared remarks, earnings call, 2026-07-29
“total capital expenditures for the quarter were about $1.6 billion, approximately 90% of which was invested in capacity expansion.”
c22 · CFO, prepared remarks, earnings call, 2026-07-29
“With AI as an accelerant, we expect demand to remain robust as customers modernize their technology architectures and increasingly orchestrate their strategies on our platform.”
c23 · CFO, prepared remarks, earnings call, 2026-07-29
“We would expect to fund the growth through two levers. One, retain cash flow.”
c24 · CFO, qa, earnings call, 2026-07-29
“Today we have 3 GW of land under control. We're building about 700 MW of that right now.”
c25 · CEO, qa, earnings call, 2026-07-29
“A broad slowdown or correction in AI-related investment could reduce or delay customer demand and the information technology spending which could impact our growth plans.”
c26 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“We are currently investing in our AI strategy to serve the large footprint we foresee for customers’ AI workloads.”
c27 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“This increased power consumption, which we expect to accelerate with the adoption of AI, has driven us to build out our new IBX data centers to support power and cooling needs twice that of previous IBX data centers.”
c28 · Filing, mdna, 10-Q periodic report, 2026-07-29
“To serve the needs of the growing hyperscale data center market, including the world's largest cloud service providers and increased demand driven in part by the adoption of AI, we continue to look at attractive opportunities to grow our market share and selectively improve our footprint and offerings.”
c18 · Filing, mdna, 10-Q periodic report, 2026-07-29
“Additionally, the workloads related to new and evolving technologies such as AI are increasing the demand for high density computing power.”
c41 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“We have, as Olivier has already mentioned, the benefit of a very strong balance sheet, which has meant that we have been able, where appropriate, to secure our M&E by pre-purchasing elements of the equipment that we need for our data centers.”
c42 · CEO, qa, earnings call, 2026-07-29
“We added 4,200 net cabinet billings, and our backlog sold but not yet installed is at a record level.”
c48 · CFO, prepared remarks, earnings call, 2026-07-29
“Further, because of the expected growth and opportunity related to AI, we anticipate significant investments in the data center industry by both current competitors and new investors and companies looking to capture this opportunity.”
c49 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“we entered into an equity commitment letter with a subsidiary of Canadian Pension Plan Investment Board to contribute up to $963 million”
c38 · Filing, notes, 10-Q periodic report, 2026-07-29
“In Q2, we saw that the vast majority of our largest deals were driven by AI workloads, similar to what we've seen in the previous quarters.”
c1 · CEO, qa, earnings call, 2026-07-29
“so AI is an accelerant to the ongoing digitization activities of our customers.”
c10 · CEO, qa, earnings call, 2026-07-29

By quarter

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

vendor bill

Third-party AI tools paid for employees’ use

0% to 0.05% of the quarter’s revenue

Incremental total: counts in full.

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

The 10-Q again names third-party AI tools among the risks of employees’ AI use and adds that the company’s use of AI may become too costly (c34, c33). No vendor, seat count or amount is given. The size, , is the ledger’s own, seats priced from the ledger’s coding-tools reference.

Evidence: 3 quotes, 1 confound

What the company pays for the third-party AI tools its employees use, which the 10-Q names as a source of security, privacy and legal risk. No vendor, seat count or amount is given, so the size is a seat-based reference class from the ledger’s coding-tools reference.

Why this motive

Tools in early use with no amount given (c32, c34); carried from Q1.

Before LLMs: new

A bill for third-party AI tools cannot exist without LLMs, so the channel is new whatever the anchor says. The anchor says employees had begun using AI and machine learning capabilities but is silent on any bill for third-party AI tools, and gives no amount.

What else could explain it

  • line composition: Any AI tools cost sits inside general and administrative expense with other software.

Quotes

“We have begun leveraging AI and machine learning capabilities for our employees to use in their day-to-day operations.”
c32 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“As we embark on these initiatives, we may encounter challenges such as a shortage of appropriate data to train internal AI models, a lack of skilled talent to effectively execute our strategy of leveraging AI internally, our desired use of AI becomes too costly, or the possibility that the tools we utilize may not deliver the intended value.”
c33 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“Use of third-party AI tools can also bring information security, data privacy and legal risks.”
c34 · Filing, risk factors, 10-Q periodic report, 2026-07-29

By quarter

  • Q1 2026described · our inference · exploratory · $61k to $1.3mn
  • Q2 2026described · our inference · exploratory · $61k to $1.3mn

Cost displaced by AI1 channel · $351k to $11mn sized · $351k to $11mn incremental

back office · cheap to verify

AI and machine learning tools used by employees in daily operations

0.01% to 0.4% of the quarter’s revenue

Incremental total: counts in full.

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

The new CFO says a functionalized, standardized organization allows the company to automate processes with AI, which he expects to drive margin expansion in go-to-market and support functions (c30, c31). The 10-Q again says employees have begun using AI tools and now adds that their use may become too costly and that employees may deploy AI agents (c32, c33, c35). No saving or cost line is attributed to AI. The size, , is the ledger’s own, a saving rate on part of sales, marketing, general and administrative expense of .

Evidence: 5 quotes, 3 figures, 2 confounds, 1 from before coverage

Employees’ use of AI and machine learning tools in day-to-day operations, which the 10-Q says the company has begun, and which the new CFO in Q2 2026 says it will use to automate processes in go-to-market and support functions, beside standardization from a functionalized organization. No saving, headcount or cost line is attributed to it. The displaced cost would sit in sales and marketing and general and administrative expenses, paid to the company’s own staff and contractors.

Why this motive

A forward statement with no measure and no line moving: the CFO says automating processes with AI will help margins (c30); the 10-Q says the company has only begun.

Before LLMs: expanded

The fiscal 2024 10-K already said employees had begun using AI and machine learning tools (eqix-anchor-c4), across a workforce of . The work existed and the tools are deployed in it, so the channel is expanded with no measured saving. The size is the change AI made, not the whole line.

“We have begun leveraging AI and machine learning capabilities for our employees to use in their day-to-day operations.”
Filing, risk factors, 10-K periodic report, 2025-02-12

Figures

  • Sales and marketing · 2026-CQ2
  • General and administrative · 2026-CQ2
  • Sales and marketing plus general and administrative · 2026-CQ2

What else could explain it

  • transformation program: The CFO credits future margin to functionalization and standardization together with AI; the AI part is not separated (c31).
  • operating leverage: General and administrative expense barely moved year over year, which the 10-Q does not attribute to AI.

Quotes

“Allows us also to automate AI, our processes. We believe that that will drive margin expansion, and this improvement of the various functions will impact go-to-market operations and also all the support functions.”
c30 · CFO, qa, earnings call, 2026-07-29
“We are a functionalized organization. All the elements of the value chain at Equinix are functionalized, and functionalization drive standardization.”
c31 · CFO, qa, earnings call, 2026-07-29
“We have begun leveraging AI and machine learning capabilities for our employees to use in their day-to-day operations.”
c32 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“As we embark on these initiatives, we may encounter challenges such as a shortage of appropriate data to train internal AI models, a lack of skilled talent to effectively execute our strategy of leveraging AI internally, our desired use of AI becomes too costly, or the possibility that the tools we utilize may not deliver the intended value.”
c33 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“Moreover, to the extent our employees develop or deploy AI-enabled tools or "agents" to perform tasks or make decisions with limited oversight”
c35 · Filing, risk factors, 10-Q periodic report, 2026-07-29

By quarter

  • Q1 2026described · our inference · exploratory · $343k to $10mn
  • Q2 2026described · our inference · exploratory · $351k to $11mn

Revenue arriving through AI4 channels · 4 not sized

customer cohort

Colocation and interconnection sold for AI workloads in the retail data centers

Not sized

A share of the largest deals by count separates nothing in dollars: the CEO says the vast majority of the largest deals were driven by AI workloads (c1), as about were in Q1, and gives no AI share of bookings or revenue. The ceiling is quoted in the metrics: recurring revenues of , colocation , interconnection and annualized gross bookings of .

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

Recurring revenues were , against a year earlier; colocation and interconnection , with a record net interconnections added. The CEO says the vast majority of the largest deals were driven by AI workloads, similar to the of Q1 (c1), and that a significant proportion of demand is enterprises modernizing on-premises infrastructure, the remainder AI-native workloads and their providers (c2); neither statement gives a dollar share. Customer examples are AI clouds, a sovereign inference node and AI factories built with Cisco and NVIDIA (c4, c5, c6). Annualized gross bookings were , with about of pre-selling on top. The channel is unsized: the deal share is by count and separates nothing in dollars, and the data-center revenue and bookings are the ceiling on the AI part. The 10-Q adds a warning that a correction in AI-related investment could reduce demand (c26). Counterparty mixed: AI model providers and neoclouds, mostly investor-funded, and hyperscalers and enterprises, funded from operations; the company names none of them, and some may be on this ledger as MSFT, AMZN, GOOGL, META, ORCL and CRWV, where the same dollars are spend.

Evidence: 28 quotes, 8 figures, 3 confounds, 3 from before coverage

Cabinets, power and interconnection in the retail IBX data centers sold for AI workloads: network and inference nodes placed by AI model providers and neoclouds, private and sovereign AI stacks and AI factories run by enterprises, and liquid-cooled high-density deployments. Management gives a share of its largest deals that were AI-related, by count, and no AI share of bookings or revenue; the rest of the same lines is enterprise colocation, cloud on-ramps, network and interconnection demand the company does not attribute to AI. The data-center revenue and the bookings are therefore a ceiling on this channel; a share of deals by count separates nothing in dollars, so the channel is read directional and unsized, with both shares and the ceiling quoted. The payers are mixed: AI model providers and neoclouds, which are mostly funded by investors, and hyperscalers and enterprises, funded from operations; the company names no customer and says its largest customer is a small share of recurring revenue. Hyperscalers on this ledger (MSFT, AMZN, GOOGL, META, ORCL) and neoclouds (CRWV) may pay part of this revenue, where the same dollars are lease or colocation spend; this ledger does not net them. Facilities layer: space, power and interconnection sold to whoever runs the compute.

Why this motive

The one measure is a share of the largest deals by count, ‘the vast majority’ this quarter (c1); it moves no line and is not a dollar share, colocation is not a separately priced AI product, and management calls AI an accelerant of customers’ digitization (c10), so the less durable motive is taken.

Before LLMs: expanded

At the anchor colocation revenue was and interconnection in 2024, inside recurring revenues of ; hyperscalers alone represented more than of annualized retail revenue (eqix-anchor-c9). The CEO already linked part of an uptick in larger footprint deals to AI-related workloads (eqix-anchor-c3), and the 10-K described a Singapore data center designed for AI with liquid cooling (eqix-anchor-c5). Selling space and power for AI workloads predates coverage; during coverage new capacity was built and sold to data-centre customers, with a record backlog of cabinets sold but not yet installed (c13, c48), so volume moved and the channel is expanded. No AI share was given, so the level has no baseline. The size is the whole of an activity that existed before.

“we did see an uptick, I will say, this quarter in larger footprint opportunities. And I think some of that is, is really associated with AI-related, you know, workloads, both service provider and enterprise.”
CEO, qa, earnings call, 2024-05-09
“This new high performance data center will feature a design built to efficiently compute intensive workloads like artificial intelligence ("AI"), supported by capabilities such as advanced liquid cooling.”
Filing, business, 10-K periodic report, 2025-02-12
“We remain an integral and growing part of hyperscaler architectures, with these customers collectively representing more than $1.3 billion of annualized revenue in Q1 in our retail business alone”
CEO, prepared remarks, earnings call, 2024-05-09

Figures

  • Annualized gross bookings in the quarter (the annual value of new monthly recurring revenue booked, not revenue) · 2026-CQ2
  • Pre-selling activity on top of bookings (approximately) · 2026-CQ2
  • Recurring revenues · 2026-CQ2
  • Recurring revenues, prior-year quarter · 2025-CQ2
  • Colocation revenues (segment note) · 2026-CQ2
  • Interconnection revenues (segment note) · 2026-CQ2
  • Net interconnections added in the quarter · 2026-CQ2
  • Share of the largest deals that were AI-related, by count (approximately) · 2026-CQ1

What else could explain it

  • line composition: Colocation and interconnection also carry enterprise modernization, cloud and network demand the company does not attribute to AI; the CEO puts a significant proportion of demand there (c2), and the AI part is not separated.
  • other: The stated share is of the largest deals, in words, not of bookings or revenue dollars.
  • mix shift: Firm pricing and rising density lift revenue per cabinet; the company does not attribute pricing to AI (c43).

Quotes

“In Q2, we saw that the vast majority of our largest deals were driven by AI workloads, similar to what we've seen in the previous quarters.”
c1 · CEO, qa, earnings call, 2026-07-29
“A significant proportion of this demand comes from the world's largest enterprises modernizing their on-prem infrastructure that was never built for today's broad-based distributed workloads. The remainder comes from net new AI native workloads and service providers powering them.”
c2 · CEO, prepared remarks, earnings call, 2026-07-29
“Eight of the top 10 model providers, as well as eight of the top 10 neo clouds, are already running their key networking workloads on Equinix today.”
c3 · CEO, prepared remarks, earnings call, 2026-07-29
“Leading AI cloud infrastructure provider OrionVM selected Equinix to power its fully managed private agentic AI bundle”
c4 · CEO, prepared remarks, earnings call, 2026-07-29
“SCX.ai, Australia's sovereign AI infrastructure provider, partnered with Equinix to build the country's first sovereign AI inferencing node, leveraging our Sydney operations.”
c5 · CEO, prepared remarks, earnings call, 2026-07-29
“Our expanded collaboration with Cisco and NVIDIA tackles this head- on by bringing standardized AI factory blueprints and automation across our global IBX network.”
c6 · CEO, prepared remarks, earnings call, 2026-07-29
“Expanded collaboration with Cisco and NVIDIA to help enterprises accelerate AI deployment through standardized AI factory architectures, secure infrastructure and real-world testing environments across Equinix's global data center footprint.”
c7 · Filing, press release, 8-K earnings release, 2026-07-29
“This is where they're deploying centers of excellence like AI factories for model training, but also for batch inferencing at Equinix. A lot of this is driven by our opportunity to provide liquid cooling in our facilities.”
c8 · CEO, qa, earnings call, 2026-07-29
“The first I'm going to call stack, which is where enterprises are running open models but on private AI infrastructure. They're doing that to cut down their token cost.”
c9 · CEO, qa, earnings call, 2026-07-29
“so AI is an accelerant to the ongoing digitization activities of our customers.”
c10 · CEO, qa, earnings call, 2026-07-29
“Certainly, the density of our deals is moving upwards as customers seek to secure the capacity that they need for their energy and their compute future.”
c11 · CEO, qa, earnings call, 2026-07-29
“we delivered annualized growth bookings of $424 million, up 23% year-over-year, a notable acceleration from Q1”
c12 · CEO, prepared remarks, earnings call, 2026-07-29
“AI requirements are additive to these connectivity budgets.”
c13 · CEO, qa, earnings call, 2026-07-29
“We are focused on the enterprise sector, and we believe in the long term that the enterprise sector will be the beneficiaries of AI technology”
c14 · CEO, qa, earnings call, 2026-07-29
“With AI as an accelerant, we expect demand to remain robust as customers modernize their technology architectures and increasingly orchestrate their strategies on our platform.”
c23 · CFO, prepared remarks, earnings call, 2026-07-29
“A broad slowdown or correction in AI-related investment could reduce or delay customer demand and the information technology spending which could impact our growth plans.”
c26 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“We are currently investing in our AI strategy to serve the large footprint we foresee for customers’ AI workloads.”
c27 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“This increased power consumption, which we expect to accelerate with the adoption of AI, has driven us to build out our new IBX data centers to support power and cooling needs twice that of previous IBX data centers.”
c28 · Filing, mdna, 10-Q periodic report, 2026-07-29
“Equinix is uniquely positioned to serve the networking, cloud and AI infrastructure needs of enterprises around the world.”
c39 · CEO, press release, 8-K earnings release, 2026-07-29
“Further, as a result of the increase in demand for AI infrastructure, we are anticipating chip shortages relative to those experienced in the market in prior years. This shortage could impact our customers and delay or deter customer server deployments within our IBX data centers.”
c40 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“Additionally, the workloads related to new and evolving technologies such as AI are increasing the demand for high density computing power.”
c41 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“Our net pricing actions were strong in Q2, and they continue to trend very favorably.”
c43 · CEO, qa, earnings call, 2026-07-29
“I think we will continue to see this kind of growth in the Americas, given that much of the AI activity and company base exists here in the first instance.”
c44 · CEO, qa, earnings call, 2026-07-29
“This shift is being accelerated by the proliferation of hybrid multi-cloud architectures and the adoption of AI.”
c45 · Filing, mdna, 10-Q periodic report, 2026-07-29
“Cloud providers, neo cloud, enterprise AI model, all of this trend is playing to Equinix strength.”
c46 · CFO, qa, earnings call, 2026-07-29
“We added 4,200 net cabinet billings, and our backlog sold but not yet installed is at a record level.”
c48 · CFO, prepared remarks, earnings call, 2026-07-29
“Further, because of the expected growth and opportunity related to AI, we anticipate significant investments in the data center industry by both current competitors and new investors and companies looking to capture this opportunity.”
c49 · Filing, risk factors, 10-Q periodic report, 2026-07-29
“I mentioned externally the market dynamics are changing to move inference ahead of perhaps where we initially scheduled or intended that it would be at, and that's much stronger than everything we saw a year ago.”
c50 · CEO, qa, earnings call, 2026-07-29

By quarter

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

product revenue

Fees from the xScale joint ventures for hyperscale builds and leases

Not sized

AI named beside another cause: the 10-Q again says hyperscale demand comes from the world’s largest cloud service providers and ‘increased demand driven in part by the adoption of AI’ (c18), and nothing separates AI’s part. The ceiling is the services revenue from the ventures, , which includes about of one-time lease fees.

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

Income from arrangements with the ventures was , of which was interest on the AMER 2 Loan, leaving of services revenue, against in Q1. The company closed megawatts of xScale leases, including Hampton, which contributed about of non-recurring fees (c15), and the 10-Q credits of Americas revenue growth to non-recurring services to the ventures (c17); non-recurring revenues were , against a year earlier. The 10-Q names AI beside cloud as the cause of hyperscale demand (c18). The channel is described and unsized; the services revenue is the ceiling, and its remainder is hyperscale cloud demand. Funding mixed: the ventures pay from partner equity and their own debt, and the hyperscale lessees pay rent to the ventures. The hyperscale lessees, which the sources do not name, may include companies on this ledger (MSFT, AMZN, GOOGL, META, ORCL), where the same dollars would be lease cost; that is a possibility, not something the sources state.

Evidence: 5 quotes, 8 figures, 3 confounds, 3 from before coverage

Revenue the company earns from its unconsolidated xScale joint ventures, which build and lease large single-tenant data centers to hyperscale cloud companies: development, facilities, asset management and sales services, and one-time fees when a venture signs a hyperscale lease (the Hampton campus lease in Q2 2026). The 10-Q shows these as income from equity method investees in the related-party note and as non-recurring services to the joint ventures in the revenue explanation. The filings name AI together with cloud as the cause of hyperscale demand and give no AI share, so the fees are a ceiling and the channel is described and unsized. The payer of record is the venture, funded by its partners’ equity and its own debt; the lessees are hyperscalers the sources do not name, which may include companies on this ledger (MSFT, AMZN, GOOGL, META, ORCL); that is a possibility, not a stated fact. Facilities layer.

Why this motive

AI named beside cloud as the cause of hyperscale demand, with no measure of AI’s part (c18, c19); carried from Q1.

Before LLMs: expanded

At the anchor the ventures already paid the company: related-party income of from the xScale ventures that are VIEs and from the EMEA 1 venture in 2024, and the 10-K tied hyperscale demand to cloud and AI together (eqix-anchor-c1, eqix-anchor-c7, eqix-anchor-c2). During coverage the ventures contracted new hyperscale capacity, closing leases including Hampton in Q2 2026 (c15), so volume moved and the channel is expanded. The fees were AI-linked already at the anchor, so no quarter before AI can be traced. The size is the whole of an activity that existed before.

“Hyperscalers require infrastructure to support demanding workload requirements for cloud and AI initiatives.”
Filing, business, 10-K periodic report, 2025-02-12
“In our xScale program, demand remains robust as cloud and AI needs are translating into strong pre-leasing activity.”
CEO, prepared remarks, earnings call, 2024-05-09
“To serve the needs of the growing hyperscale data center market, including the world's largest cloud service providers and increased demand driven in part by the adoption of AI, we have entered into joint venture partnership arrangements across our Americas, EMEA and Asia-Pacific regions to develop and operate xScale data centers.”
Filing, mdna, 10-K periodic report, 2025-02-12

Figures

  • Income from arrangements with equity method investees (related-party note) · 2026-CQ2
  • Interest income on the AMER 2 Loan inside that income · 2026-CQ2
  • Revenue from services to the equity method investees, mostly the xScale ventures (income less loan interest) · 2026-CQ2
  • Incremental Americas revenues from non-recurring services to the joint ventures, year over year · 2026-CQ2
  • Non-recurring xScale fees from the leases closed, including Hampton (approximately) · 2026-CQ2
  • Megawatts of xScale leases closed in the quarter · 2026-CQ2
  • Non-recurring revenues · 2026-CQ2
  • Non-recurring revenues, prior-year quarter · 2025-CQ2

What else could explain it

  • line composition: The related-party income also carries services to venture data centers whatever their use; the AI part is not separated.
  • other: The filings name cloud and AI together as the cause of hyperscale demand and give no share.
  • one time item: Most of the quarter’s fees are one-time lease-signing fees, including Hampton, which slipped from Q1; the CFO guides non-recurring revenue back to its usual share of revenue.

Quotes

“As expected, we closed 134 MW of xScale leases, including Hampton, which contributed approximately $120 million in non-recurring fees.”
c15 · CFO, prepared remarks, earnings call, 2026-07-29
“This is a result of continued cost discipline, scaling our operating leverage, and our xScale leasing fees.”
c16 · CFO, prepared remarks, earnings call, 2026-07-29
“$124 million of incremental revenues from non-recurring services provided to our joint ventures”
c17 · Filing, mdna, 10-Q periodic report, 2026-07-29
“To serve the needs of the growing hyperscale data center market, including the world's largest cloud service providers and increased demand driven in part by the adoption of AI, we continue to look at attractive opportunities to grow our market share and selectively improve our footprint and offerings.”
c18 · Filing, mdna, 10-Q periodic report, 2026-07-29
“We believe hyperscale customers will play a large role in the growth of the market for AI.”
c19 · Filing, risk factors, 10-Q periodic report, 2026-07-29

By quarter

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

product revenue

Distributed AI Hub: a private on-ramp to AI model companies and GPU clouds

Not sized

The quarter’s sources do not mention the Distributed AI Hub; it has no price or revenue.

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

No source this quarter mentions the Distributed AI Hub. Interconnection products named on the call are Fabric Geo Zones, a sovereignty offering, Fabric Intelligence and Cloud Router.

Evidence: 0 quotes, 2 from before coverage

An interconnection offering, launched in Q1 2026, that gives enterprises one private, low-latency connection to AI model companies, GPU clouds, data platforms and security services. No price, revenue or customer count is given; its sales, if any, sit inside interconnection revenue, which the colocation and interconnection channel already covers. The anchor shows the same activity, private on-ramps through Equinix Fabric to clouds and service providers, so the tie-break reads the new name as relabelled. Facilities layer, as interconnection sold to AI users.

Why this motive

Carried from Q1: a launched offering with no price or measure; the quarter is silent.

Before LLMs: relabelled

At the anchor Equinix Fabric already connected businesses to networking, storage, compute and application providers, and the company sold private cloud on-ramps across its metros (eqix-anchor-c6, eqix-anchor-c10). No line, price or volume is shown to move because of the AI Hub.

“Equinix Fabric enables businesses to connect globally to their choice of thousands of networking, storage, compute and application service providers in the industry’s largest infrastructure ecosystem.”
Filing, business, 10-K periodic report, 2025-02-12
“In the quarter, we added one new native cloud on-ramp in Madrid, bringing us to 220 native cloud on-ramps across our portfolio, spanning 47 metros.”
CEO, prepared remarks, earnings call, 2024-05-09

By quarter

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

product revenue

Fabric Intelligence: AI that monitors and reconfigures the interconnection network

Not sized

No price or general availability is given: in Q1 the feature was in preview and this quarter adds no launch or price, so it stays described and unsized, never sized at zero.

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

The CEO lists Fabric Intelligence among the Fabric capabilities, for observability and management (c29). No price, revenue or customer count is given.

Evidence: 1 quote, 1 from before coverage

A feature of the Equinix Fabric interconnection platform that, in the release’s words, embeds AI in the network to interpret telemetry and act on it without human intervention. It was introduced in Q1 2026 and is in preview with customers and partners; no price or revenue is given. It is an AI feature sold to businesses on an existing product, so end-use layer, and an increment to Fabric revenue once priced.

Why this motive

A feature listed among Fabric capabilities with no price or launch date (c29); carried from Q1.

Before LLMs: expanded

At the anchor Equinix Fabric was sold as software-defined interconnection (eqix-anchor-c6), with no automated monitoring or reconfiguration described; the feature adds to an existing line. The size is the change AI made, not the whole line.

“Equinix Fabric enables businesses to connect globally to their choice of thousands of networking, storage, compute and application service providers in the industry’s largest infrastructure ecosystem.”
Filing, business, 10-K periodic report, 2025-02-12

Quotes

“Fabric intelligence providing additional capabilities of observability and management for our customers.”
c29 · CEO, qa, earnings call, 2026-07-29

By quarter

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

Reported lines, year-over-year growth

Revenue +16.4%

Q2 2026. Growing slower than revenue: cost of revenues (+13.5%), sales and marketing (+8.1%), general and administrative (+2.4%), total costs and operating expenses (+11.2%). 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 revenuesSales and marketingGeneral and administrativeTotal costs and operating expensesRevenue
-10%-5%0%5%10%15%20%Q1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026RevenueCost of revenuesTotal costs and operating expensesSales and marketingGeneral and administrative
Reported values and filings
LineQ1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026
Revenues$2.23bn$2.26bn$2.32bn$2.42bn$2.44bn$2.63bn
Cost of revenues$1.08bn$1.08bn$1.14bn$1.20bn$1.19bn$1.23bn
Sales and marketing$229mn$221mn$219mn$234mn$241mn$239mn
General and administrative$438mn$451mn$470mn$481mn$444mn$462mn
Total costs and operating expenses$1.77bn$1.76bn$1.84bn$2.00bn$1.87bn$1.96bn