Equinix

Refresh delta (since 2026-06-21)

What changed

1. Q2 2026 printed a 16% revenue quarter — and roughly a third of that growth was a single non-recurring fee. Revenue $2,625M vs $2,256M (+16% reported, +15% cc), but recurring revenue was $2,377M, +11% reported / +9% cc, while non-recurring revenue went $113M → $248M (+119%). The Americas MD&A names it: "$124 million of incremental revenues from non-recurring services provided to our joint ventures." The call names it more precisely — the Hampton xScale transaction "contributed $120 million in fees."

2. The "record 53% adjusted EBITDA margin" is that same fee. Adjusted EBITDA $1,396M (53.2%) vs $1,129M (50.0%). Americas adjusted EBITDA rose $175M on $247M of revenue — a 71% incremental margin against a 46.4% segment baseline. Strip the JV fee at its implied ~95% flow-through and group margin lands at ~51.1%, i.e. flat on Q1's 50.9% and exactly the full-year guide.

3. Guidance was raised — including the long-term outlook, hard. FY2026 revenue to $10.205-10.285B (11-12%), AFFO/share to $42.69-43.29 (from $42.31-43.11), adjusted EBITDA $5.210-5.270B, capexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs. $5.0-6.0B. And 2027-2029 went to 10-13% revenue growth (from 7-10%) and 9-12% AFFO/share growth (from 5-9%), with 53%+ margin by 2029 and $5-7B of capex a year. The CEO called it "the largest single guidance raise in the history of our company."

4. The Q3 guide undoes the Q2 headline, and the market read it correctly. Q3 revenue guided $2.525-2.575B — a midpoint below both consensus ($2.58B) and Q2's own actual ($2.625B) — with margin back to ~51%. The stock fell 3% on the print.

5. Capex has roughly doubled and free cash flow has gone deeply negative. H1 capex $3,058M vs $1,838M (+66%); H1 operating cash flow $1,784M vs $1,753M (+1.8%). H1 FCFFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. = -$1,274M against -$85M a year ago, with $1,029M of dividends paid on top. Funded with debt: ~$5.4B of notes issued YTD (Mar $1.5B at 4.40%/4.70%; May C$1.25B ≈ $919M; Aug 6 $3.0B at 5.000%/5.250%/5.500%/5.800%) against $1.38B of repayments.

6. Both open governance items from the prior dossier closed — and the prior dossier had them wrong. The new CFO is Olivier Leonetti (ex-Eaton, ex-Johnson Controls), appointed 2026-03-07, effective 2026-03-16 — announced three months before the prior dossier listed "new CFO appointment" as a pending catalyst. And the Delaware derivative suit was dismissed: motion to dismiss granted 2026-05-27, case dismissed — the last live legacy item from the Hindenburg episode.

What held

The structural thesis stands, and one leg of it got stronger. Recurring revenue is still 91-92% of the total and still compounding at 9-11%; churn 1.8%; 9,700 net interconnection adds — a company record; bookings $424M (+23%); presales lifted ~25% → ~30% of 2026 retail capacity sold; 194 stabilised assets yielding 27% cash-on-cash; 8 of the top 10 AI model providers and 8 of the top 10 neoclouds on-platform. The prior dossier's central claim — that inference and distributed AI want interconnection density, and that this is EQIX's shape of demand — is what management now says out loud, naming four AI use cases (stack, sovereign, batch, latency-sensitive) and shipping product against them (Cisco/NVIDIA Secure AI Factory blueprints, Fabric Geo Zones sovereignty in preview with 80 enterprises).

Two things did not hold as cleanly as the prior dossier implied. (a) Interconnection revenue growth decelerated — $453M, +9% YoY in Q2, down from Q1's $446M/+13.5%, despite record volume adds. Record units, slowing revenue, is price/mix, and it is the moat line. (b) The prior dossier's framing of "recurring revenue re-accelerating to +12%" does not survive the Q2 cut: on the filing's own constant-currency basis recurring is +9%, and the reported-basis acceleration was FX and fees.

Phase A — Understand the business

Lens 1 (Company Overview) — unchanged, see the previous dossier. Two refinements: 52 major projects now span 33 markets (was 35 metros), and management pulled 7,000+ cabinets forward from 2027 into 2026, expecting to double H2 cabinet deliveries.

Lens 2 (Supply Chain) — unchanged, see the previous dossier. Commitment scale moved with the build: unaccrued capex contractual commitments $6.3B → $6.1B, other non-capital purchase commitments (power) steady at ~$2.1B, total lease payments ~$5.0B.

Lens 3 (Moats) — unchanged, see the previous dossier, with one honest amendment: the prior dossier called interconnection "the highest-margin line growing faster than colocation." At Q2 that reversed — interconnection revenue +9% vs total recurring +11% reported. Record net adds with slower revenue means average revenue per interconnection is falling, consistent with mix shifting toward virtual/Fabric connections (Cloud Router bookings +170%) that are cheaper per unit than physical cross-connects. That is a real, watchable erosion in the moat's pricing, even as its density sets records.

Segments — REFRESHED (the mix moved, and it moved the wrong way for quality)

Q2 2026 by geography ($M) — all ``:

Recurring vs non-recurring is the disclosure that matters this quarter:

The recurring share of revenue fell four points in a single year (95% → 91%; 95% → 92% on the six-month view). For a business whose entire valuation premium rests on recurring, contracted, sticky revenue, that is the single most important line in the filing — and it is not in the press-release headline.

Product-line detail (colocation / interconnection / managed infrastructure) is not disclosed in the 10-Q — the disaggregation is recurring/non-recurring by region only. Interconnection revenue $453M, +9% comes from the call. Colocation and managed-infrastructure splits for Q2: n/a this pass.

---

Phase B — Measure performance

Earnings Result (Q2 2026, reported 2026-07-29) — REFRESHED

The print, as reported. All `` unless noted.

Vs consensus: EPS $4.83 vs $4.67 consensus (+$0.16); revenue $2.62B vs $2.58-2.59B. A beat on both lines.

The beat, decomposed. This is the finding of the refresh.

The 10-Q attributes $124M of the Americas' $247M revenue increase to "non-recurring services provided to our joint ventures"; the call attributes ~$120M to the Hampton xScale transaction specifically, inside 134MW of xScale leases closed. The related-party note corroborates the size independently: income from equity-method investees was $161M in Q2 2026 vs $36M in Q2 2025 (of which $9M/$7M is AMER 2 Loan interest), i.e. ~+$123M of related-party service revenue.

Working the arithmetic through, all `` with the steps shown:

• Revenue ex the incremental fee: $2,625M − $124M = $2,501M vs $2,256M = +10.9% reported. The headline +16% becomes ~+11%.
• Margin ex the fee: Americas adjusted EBITDA rose $175M on $247M of revenue = 70.9% flow-through, against a Q2 2025 Americas margin of 46.4% ($466M/$1,004M). If the non-JV part of the Americas increase ($247M − $124M = $123M) earned that 46.4% baseline, it contributed ~$57M — leaving ~$118M of EBITDA on $124M of fee revenue (~95% flow-through). Group adjusted EBITDA ex-fee ≈ $1,396M − $118M = $1,278M on $2,501M = 51.1% — versus Q1's 50.9% and versus the company's own ~51% full-year guide. The 320bp margin expansion is the fee.
• Adjusted EBITDA growth ex the fee: $1,278M / $1,129M = +13.2% (vs +24% reported).
• AFFO ex the fee: the fee flows to AFFO essentially untouched (the JV-related AFFO adjustment line is $2M). $1,168M − $118M = ~$1,050M vs $972M = +8.0% (vs +20% reported); AFFO/share ≈ $10.59 vs $11.78 reported.

None of this is hidden — management disclosed it in the MD&A and named it on the call. What is worth saying plainly is that the press release headlines ("16% revenue growth", "record adjusted EBITDA margin of 53%", "AFFO grew 20%") carry no such caveat, and the Q3 guide silently removes it: revenue $2.525-2.575B (midpoint −2.9% sequentially from Q2's actual) at ~51% margin. Management is telling you the 53% does not repeat.

Guidance — raised on every horizon:

Balance-sheet flags — three, and they are the honest ones.

1. Free cash flow is deeply negative and the gap is debt-funded. H1 OCF $1,784M (+1.8% YoY) against H1 capex $3,058M (+66% YoY) = -$1,274M, plus $1,029M of dividends. OCF grew 1.8% while net income grew 26%, because working capital absorbed it: accounts receivable used $258M (vs $169M) and other assets/liabilities used $155M (vs providing $152M). Consolidated AR went $1,001M → $1,256M.
2. A large slice of that receivable build is the JV fee itself. Related-party balances: accounts receivable $145M vs $35M at 31 Dec 2025; related-party contract assets and other receivables in other current assets $197M vs $58M; other assets $437M vs $350M. So ~$249M of incremental related-party receivables/contract assets against $208M of six-month related-party income. The fee revenue that made the quarter has largely not been collected in cash.
3. Capitalised interest is doing more work. Interest expense $151M (+12%) but interest capitalised $38M vs $14M, so interest charges incurred were $189M vs $149M (+27%). The build lets EQIX capitalise a rising share of a rising interest bill; reported interest expense — which is what hits AFFO — understates the true cost of the debt stack by a widening margin.

Leverage itself is fine. Total debt $21,989M (senior notes $19,689M + finance leases $2,280M + mortgage/loans $20M); cash $979M + short-term investments $1,245M → net debt ~$19,765M. Against the FY2026 adjusted-EBITDA guide midpoint of $5,240M that is ~3.8x — below the ~4.0x in the prior dossier and below the 5.1x sector average. Liquidity $7.7B including an upsized revolver. Investment-grade and comfortable; the question is trajectory, not the level.

Market reaction: sold. Shares fell ~3% on 2026-07-29 as the soft Q3 midpoint overshadowed the raise. Price ~$1,053 on 2026-08-06 — essentially unchanged from the ~$1,050 in the prior dossier seven weeks earlier, through the largest guidance raise in company history. 52-week range $720.62-$1,128.68.

Conflict surfaced, not resolved. The 10-Q reports recurring revenue +9% constant currency; the company and the call report MRR +11% "normalised and constant currency" — a company-defined basis that adjusts for items (power pass-throughs and similar) the filing does not strip. Both are stated as given. Separately, two web sources disagree on 2026 YTD return — barchart implies +41%, TIKR +26% — neither is reconcilable against the other from what was sourced this pass; the durable fact is that the price is flat versus the prior dossier.

Earnings Calls — REFRESHED (first transcript now on the shelf)

transcripts/2026-q2.md is newly ingested (Motley Fool transcript via The Globe and Mail, 2026-07-29). It is partial — prepared remarks plus the Fool's structured takeaways; the Q&A is truncated by the syndication source. Sentiment comparison to prior quarters therefore remains part ``.

The arc, extended: defensive (2024) → offensive (Q1 2026) → expansionary (Q2 2026). The Q1 call was "the AI demand is real and we are winning it." The Q2 call is a capital-allocation call: the news is not the quarter, it is the decision to spend $5-7B a year through 2029.

New and changed language:

• The multi-year commitment, stated flatly. "To capture the robust demand in front of us, we plan to invest $5 billion to $7 billion in CapEx annually through 2029... we fully expect the new capital we are deploying to deliver the mid-20 percent yield you have grown accustomed to." — CEO Fox-Martin. That yield claim is the load-bearing sentence of the whole outlook, and it is checkable against the disclosed 27% cash-on-cash on 194 stabilised assets.
• AI segmented, not asserted. Four named use cases — stack, sovereign, batch, latency-sensitive. This is a step up in specificity from Q1's "60% of largest deals are AI-related," and it is the inference-not-training framing the prior dossier argued the market was mispricing.
• The fee was disclosed, not buried. Total revenue described as "growing 16% year over year driven by underlying performance of Equinix and nonrecurring fees from the xScale business." Credit where due.
• Sovereignty is now a product line, not a talking point — Fabric Geo Zones in preview with 80 enterprises.
• New voice on the call: CFO Olivier Leonetti, first full quarter. His framing is disciplined and capital-focused — "Most of the deployment of the capital, 80% of it, will be on our top 25 metros — markets that we understand well" — and he pushed the growth acceleration out to the back of the planning period rather than promising it now.
• Stopped saying: nothing defensive remains. The accounting/litigation vocabulary of 2024-25 is fully gone, which is consistent with the derivative suit's dismissal (Lens 10).

Operating tells from the call: bookings $424M (+23%, second-highest on record); total sales activity incl. presales +30%; net cabinet billings +4,200; churn 1.8%; presales $110M with ~30% of 2026 retail expansion pre-sold; 52 projects across 33 markets; dividends ~$2.0B for 2026; a new Chief Product Officer, Chris Audi, announced the day of the call.

Comps — REFRESHED (lightly; the peer that matters moved)

The relative read has an edge to it now. EQIX still carries a ~40% P/FFO premium to DLR — unchanged from the prior dossier. But DLR is up ~27% in 2026 while EQIX is roughly flat since June, and DLR also raised FFO guidance on data-centre demand. In other words, the market has spent 2026 paying up for the wholesale leg of the trade while the retail-interconnection leg re-rated sideways. That is the prior dossier's bear leg #1 — "AI's centre of gravity is wholesale" — showing up in relative price rather than in EQIX's fundamentals. Sell-side has not followed the tape: consensus PTConsensus price targetThe average of what published analysts think the share price should be. An opinion poll, not a forecast. ~$1,199-1,226, 26 buys / 0 sells, Truist to $1,220 on 2026-08-02. Peer EV/Sales, EV/EBIT and 5-yr ROE: n/a.

Stock-Price Catalysts — REFRESHED (new entries since 2026-06-21)

1. 2026-06-16 — Cisco + NVIDIA "Secure AI Factory" collaboration (UP, modest). Standardised AI-factory blueprints deployable across the IBX footprint, plus a Presidio P.A.T.H. Lab for pre-rollout validation inside Equinix facilities. Strategically this is the enterprise-AI on-ramp thesis made concrete.
2. 2026-07-29 — Q2 print + the largest guidance raise in company history (DOWN ~3%). The instructive one. A beat on revenue and EPS, a raise on every horizon, a record margin — and the stock fell, because the Q3 midpoint came in under consensus and under the quarter just printed. The market priced the repeatability of the quarter, not the quarter.
3. 2026-07-30 → 2026-08-06 — $3.0B senior-notes sale (neutral-to-positive; a passed test). Bloomberg framed the offering as "selling into a recently weakened environment for debt related to the artificial intelligence boom" — an explicit test of whether credit still funds AI infrastructure. It cleared on 2026-08-06 at 5.000% (2029), 5.250% (2031, ~3.95% after cross-currency swap to EUR), 5.500% (2033) and 5.800% (2036). Coupons well above the legacy stack (which carries 0.250%, 1.000%, 3.250% paper) — access is intact, price is not.
4. 2026-05-27 — Delaware derivative suit dismissed (UP, quiet). The last Hindenburg-legacy item, closed with no market event.
5. 2026-08-05 — All Nippon Airways cloud network hub; Fabric Geo Zones sovereignty launch (neutral). Product/logo news, not price-moving.

No ±5% single-day move surfaced between 2026-06-21 and 2026-08-10. The largest identified move is the ~3% earnings-day decline. What the tape now says, updated: the prior dossier concluded the market reacts to governance credibility and the AFFO/share guide. Governance is resolved and the AFFO guide went up — and the stock did nothing. The marginal buyer has moved on to a third question: can the growth be funded, and is it repeatable?

---

Phase C — Judge people & books

Management — REFRESHED (the transition completed; the prior dossier was behind)

• CFO succession is done, and the prior dossier missed it. Olivier Leonetti was approved as CFO on 2026-03-07, effective 2026-03-16, succeeding Keith Taylor (who stays on as Special Advisor). Background: CFO of Eaton (Feb 2024 - Mar 2026), CFO of Johnson Controls (Sep 2020 - Jan 2024) — a large-cap industrial finance operator, not a REIT specialist. Package: $700k base, 100% target bonus, $10M annual equity (33% RSU / 47% financial-metric PSU / 20% TSR PSU), $200k sign-on cash, $5M sign-on RSUs vesting over three years. The prior dossier listed "new CFO appointment" as a pending 2026 catalyst three months after it had happened — a shelf-coverage gap, not a company event. The prior dossier's stated risk ("losing the 27-year CFO one year after a new CEO") is now a fact to judge on execution rather than a risk to price.
• A new Chief Product Officer, Chris Audi, joined to lead the infrastructure + AI solution portfolio, announced on the Q2 call day.
• Board and auditor ratified without incident. All 10 directors re-elected at the 2026-05-13 annual meeting; PwC ratified; say-on-pay approved (77.6M for / 7.9M against). A shareholder proposal to lower the special-meeting ownership threshold failed (25.8M for / 59.3M against). Note the say-on-pay dissent (~9%) and the Nanci Caldwell vote (6.2M against, the highest of any director) as mild but non-trivial governance signals.
• An executive severance plan was adopted in Feb 2026 covering executive officers, with a CEO amended-and-restated agreement on similar terms: 100% of salary + target bonus on a no-cause termination, 200% and full equity acceleration inside a change-of-control window. Standard for the cap size; worth noting it was put in place during a leadership handover.
• Capital allocation — the decision of the quarter, and it is a big one. Committing to $5-7B of annual capex through 2029 against ~$5.2B of FY2026 adjusted EBITDA is a decision to run negative free cash flow for years and fund the difference externally, while paying ~$2.0B of dividends. Management's defence is the yield: 27% cash-on-cash on 194 stabilised assets, and an explicit promise that new capital delivers "the mid-20 percent yield you have grown accustomed to." That promise is the single most falsifiable claim in the outlook, and it is the one to hold them to. Dividend $5.16/quarter ($20.64/yr, +10% YoY) = a 48% payout on the FY26 AFFO/share guide midpoint — still comfortable.
• Red flags: none new of the governance kind. The watch-item has changed shape: it is no longer "will the accounting hold up," it is "does the new CFO's first big communicated quarter set a precedent of headlining a number that his own next-quarter guide retracts."

Forensic Red Flags — REFRESHED (one item closed, three new ones opened)

Closed — completely.

• Delaware derivative suit dismissed. Filed 2025-08-06 (Section 14(a), fiduciary breach, unjust enrichment, waste, plus an insider-trading allegation, covering statements 2019-05-03 to 2024-03-24); EQIX moved to dismiss 2025-10-20; granted 2026-05-27, case dismissed. With the SEC closing its investigation (Nov 2025), the DOJ declining, and the class action settled by insurance (Dec 2025), the entire Hindenburg forensic episode is now closed with no restatement, no enforcement and no surviving claim. The 10-Q adds: "there are no other pending claims for which the outcome is expected to result in a material adverse effect."
• Regulatory sweep re-run to 2026-08-10: zero SEC Litigation Releases and zero AAERs naming Equinix in the five-year window. Non-SEC agencies: nothing material surfaced.

New — three, and they belong to the build rather than the past.

1. Related-party revenue is now material, high-margin, and largely uncollected. Income from equity-method investees: Q2 $161M vs $36M; H1 $208M vs $104M. Against that, related-party receivables and contract assets rose ~$249M in six months (AR $35M→$145M; contract assets/other receivables inside other current assets $58M→$197M). The counterparties are unconsolidated JVs in which EQIX holds 20% (most xScale vehicles) to 25% (AMER 3 effective interest) — entities the 10-Q itself describes as VIEs "because they do not have sufficient funds from operations to be self-sustaining." Total maximum exposure to loss from unconsolidated VIEs: $1,620M (equity $631M, AR $140M, other receivables $133M, contract assets $158M, loan commitment $392M, future equity contributions $118M, debt guarantees $48M). The honest read: this is real work genuinely performed (development management, facilities and asset management, procurement, sales and marketing) and it is a legitimate capital-light way to participate in hyperscale. It is also non-recurring, related-party, cash-lagging revenue that drove the entire headline of the quarter, in the exact metric — AFFO — that was the subject of the 2024 short thesis. Both things are true. It deserves a direct question, not an accusation (Lens 14).
2. "Recurring capital expenditures" — the AFFO deduction — is flat while the asset base compounds. Recurring capex: Q2 $49M vs $55M; H1 $81M vs $81M, i.e. dead flat, while gross capex ran +66% and net PP&E went $23,584M → $25,222M in six months and D&A ran $1,101M in the half. Recurring capex is now ~2.6% of H1 gross capex. This is precisely the line Hindenburg attacked, and the maintenance/growth boundary is genuinely judgement-laden — but a flat maintenance number against a 7% half-year growth in the depreciating asset base is a trend to watch, not to wave through. It has cleared an Audit Committee review, PwC's ICFR opinion and an SEC investigation; it has not been re-tested since the capex ramp began.
3. Capitalised interest is rising fast and flatters AFFO. Interest charges incurred $189M vs $149M (+27%) in Q2, of which $38M capitalised vs $14M — so reported interest expense grew only 12%. Wholly normal for construction in progress, and wholly real as a cash cost. As capex goes to $5-7B/yr, the wedge between interest incurred and interest expensed widens, and AFFO is struck after the smaller number.

Unchanged, still true. Cash flow exceeds earnings as a D&A-heavy REIT should; impairments modest ($17M in Q2, $19M H1 — versus $233M in 2024); SBC of $273M in the half is still added back to AFFO, still the perennial legitimate quibble, and now ~12% of H1 AFFO. No new material weakness; no restatement; PwC re-ratified by shareholders.

---

Phase D — Project & stress-test

Forward Projection (FY2026 → FY2028, AFFO/share) — REFRESHED

Model integrity — read this before any number below. model.xlsx was rebuilt on 2026-08-10 from the newly populated our figures. It contains no computed values — our model reports "51 formulas and no cached results", value-per-share, balance check and every output blank, share price and CAPM inputs unset, and only 2 quarters of history (so its revenue growth seeded at a default 8% and LTM figures are partial). Nothing in this lens is cited from the workbook. Every figure below is arithmetic off the filings, the transcript and the guidance, shown inline. The workbook is a scaffold pending an Excel evaluation pass and ≥8 quarters of backfill.

Anchors:

• H1 2026 actual: revenue $5,069M; adjusted EBITDA $2,641M (52.1%); AFFO $2,233M; AFFO/share $22.57 — 52.5% of the FY guide midpoint already banked.
• FY2026 guide: revenue $10.205-10.285B; adjusted EBITDA $5.210-5.270B; AFFO/share $42.69-43.29 (mid $42.99); capex $5.0-6.0B.
• 2027-2029 outlook: revenue +10-13%/yr; AFFO/share +9-12%/yr; 53%+ margin by 2029; capex $5-7B/yr.

Driver assumptions (each labelled):

• Revenue: the company's raised 10-13% band. Treat the top of it sceptically — H1's reported 13% includes ~$110M of incremental non-recurring JV fees; the underlying recurring line is +9% cc, and the Q3 guide is +9-11%.
• Margin: +~50-100bps/yr toward 53% by 2029. Q2's 53.2% is not the base — ~51% is.
• Interest: the real headwind, and it is growing. Interest incurred +27% YoY; ~$5.4B issued in 2026 at 3.95-5.80% against a legacy stack averaging far lower; $19.9B of senior-note principal plus $4.5B of contracted interest. Maturities are back-loaded ($607M remaining 2026, $1,076M 2027, $1,426M 2028, $2,185M 2029, $3,194M 2030, $11,385M thereafter) — refinancing risk is a 2029-30 problem, not a 2027 one.
• Dilution: ~+1%/yr (98.05M → 99.14M diluted WA YoY). Note the ATM was not used in H1 2026 ($0 vs $99M in H1 2025) — the build is being funded with debt, not equity, which protects per-share AFFO now and raises the interest burden later.
• Funding gap: FY26 adjusted EBITDA guide mid $5.24B − cash interest ~$0.74B (H1 incurred $369M ×2) − cash taxes ~$0.2B ≈ $4.3B available, against capex ~$5.5B + dividends ~$2.0B = $7.5B → a ~$3.2B/yr external funding requirement. That is the number that governs the next three years.

Three-year AFFO/share path ``:

At $1,053: FY2026 base = 24.5x AFFO; FY2027 base ≈ 22.2x; FY2028 base ≈ 20.1x. Materially the same picture as the prior dossier (24.6x / 22.3x / 20.4x) — the raise and the price move offset almost exactly, which is itself the observation: the market has declined to pay for the raised outlook.

Per unattended-run rules, no our model create was logged. Suggested Brier seed for a later human pass, sharpened by this refresh: "EQIX Q3 2026 adjusted EBITDA margin ≥51.0% AND recurring revenue cc growth ≥9%, with no xScale non-recurring fee above $25M — p≈0.65, resolves 2026-10-31."

Bull vs Bear — REFRESHED

Bull case (strengthened where it counts). The demand signal is no longer a story, it is a capital commitment: bookings +23%, a record 9,700 net interconnections, presales at ~30% of 2026 capacity, 7,000+ cabinets pulled forward from 2027, and a long-term outlook raised from 7-10% to 10-13% revenue growth with AFFO/share from 5-9% to 9-12%. Management is putting $5-7B a year behind it and pointing at 27% cash-on-cash yields on 194 stabilised assets as the warrant. The Hindenburg overhang is now fully, finally dead — SEC closed, class action settled, derivative suit dismissed May 2026 — and the CFO succession that worried the last dossier is complete and seated. Product is catching up to the thesis: Cisco/NVIDIA AI-factory blueprints, Fabric Geo Zones sovereignty, Cloud Router bookings +170%. And you are paying 24.5x forward AFFO for it — the same multiple as seven weeks ago, on a guide that has since gone up. The market is charging you nothing for the raise.

Bear case (permanent-impairment risks, updated).

1. The quality of the print. The quarter that justified the raise was ~a third fee income: non-recurring revenue +119%, recurring share of revenue down 95% → 91%, and essentially the entire 320bp margin expansion attributable to a one-off xScale fee that management's own Q3 guide removes. If you underwrite 53% margins and 16% growth you have underwritten an artefact.
2. The funding arithmetic. ~$3.2B/yr of external capital needed for a multi-year build, into a credit market Bloomberg described as "a recently weakened environment for debt related to the artificial intelligence boom." New paper costs 5.0-5.8% against a legacy stack carrying 0.25-3.65%. Interest incurred is already +27% YoY. If AI credit spreads widen, the growth algorithm gets funded at a worse price or gets smaller.
3. The moat's pricing, not its density. Interconnection revenue +9% on record volume adds is the first real evidence that the highest-margin line is being repriced downward by mix (virtual/Fabric connections displacing physical cross-connects). Record units at falling revenue-per-unit is how a network-effect moat quietly becomes a commodity.
4. Relative positioning. DLR up ~27% in 2026 on wholesale AI demand while EQIX is flat. The market keeps voting for the wholesale leg.
5. Rates and the multiple. Unchanged from the prior dossier and still live: 24.5x AFFO and a ~2% yield on a now-$22B (soon ~$25B) debt stack.

Pre-mortem (18 months out, thesis broke). Q3 and Q4 2026 print ~51% margins and 9-10% growth — perfectly respectable, and 500bp below what Q2 appeared to promise. The 2027 guide comes in at the bottom of the 10-13% band. Meanwhile capex runs at $6B, free cash flow stays negative for a third straight year, and a 2027 refinancing prices 150bp wide of the 2026 deals. The stock de-rates from 24.5x to 19x on FY2027 AFFO of ~$46 → ~$875, and the post-mortem reads: the record quarter was a fee, the raise was extrapolated from it, and the build was funded into a tightening AI credit window. Note carefully — that is a de-rating, not an impairment. Nothing in this refresh damages the underlying annuity.

Are multiples too high? Full, not bubbly — same conclusion as the prior dossier, reached differently. 24.5x forward AFFO for a company-guided 9-12% AFFO/share CAGR is a PEG-equivalent of ~2.3x. The ~40% P/FFO premium to DLR is intact and still defensible on mix and moat. What has changed is the margin of error: the number the multiple is struck against (FY26 AFFO/share of $42.99) now contains a one-off fee that will not be in FY27's base.

Contrarian view (what the market refuses to see) — updated. The prior dossier's contrarian call was that inference, not training, is the durable workload and EQIX is mis-bucketed. The Q2 call is the first quarter where management said that in the company's own words — naming batch inference, metro/latency-sensitive inference and sovereign deployments as distinct demand vectors, and shipping product (Geo Zones, AI-factory blueprints) against each. The market, meanwhile, sold the stock on a quarterly guide midpoint and has paid nothing for a long-term outlook that was raised by 300bp of annual growth on both lines. The mispricing the prior dossier identified has, if anything, widened — but the entry case is now cleaner if you wait one quarter for a fee-free print. That is not a call to act; it is where the asymmetry sits.

Devil's Advocate (short-seller) — REFRESHED

The accounting-fraud leg is dead and I am not resurrecting it. Here is the durable short, sharpened by this filing.

• They headlined a number they knew would not repeat, and the sell-side wrote it down as a trend. "Record 53% adjusted EBITDA margin." "Revenue +16%." "AFFO +20%." Every one of those is true and every one of those is ~$120M of Hampton fee income. Ex-fee: revenue +10.9%, margin 51.1%, AFFO +8.0%. The company's own Q3 guide — 51% margin, revenue below the quarter just printed — is the confession. Then, on that quarter, they raised the three-year outlook by 300bp of annual growth on both the top and bottom line. You are being asked to extrapolate a lease closing.
• The revenue that made the quarter is a receivable from entities they own a fifth of. Related-party income $161M vs $36M; related-party receivables and contract assets up ~$249M in six months; the JVs are VIEs that by the company's own description "do not have sufficient funds from operations to be self-sustaining"; and EQIX's share of their results is a $24M loss in the quarter. So: book the fee, carry the receivable, absorb the equity loss, and let $1.6B of maximum exposure sit off balance sheet. Nothing about that is illegal. Everything about it makes "revenue growth" a weaker word than it was a year ago.
• The maintenance-capex line never moved. Gross capex +66%. Net PP&E +$1.6B in six months. Recurring capex — the only capex that reduces AFFO — $81M vs $81M, exactly flat. The company took an SEC investigation over precisely this boundary and came out clean; it has not been re-tested against a doubled build. Every incremental dollar of a $5-7B/yr programme classified as growth rather than maintenance is a dollar of AFFO, and AFFO is the metric in the incentive plan.
• The build eats the dividend's cover from underneath. H1 free cash flow -$1.27B; dividends $1.03B. FY26 external funding need ~$3.2B. The dividend is paid out of the debt market, not out of free cash flow, and will be for years. A 48% AFFO payout looks safe; a 48% payout on a metric struck before growth capex, funded by a bond market that Bloomberg says is going off AI paper, is a different sentence.
• The moat line is decelerating. +9% interconnection revenue on a record 9,700 adds. The bulls will tell you volume is the leading indicator. It is also possible that Fabric and Cloud Router are cannibalising $/cross-connect and the 500,000-interconnection trophy is being won at a lower price per unit each year.
• Most dangerous competitor bulls still underestimate: unchanged — hyperscaler self-build and neoclouds for the largest AI tenants, DLR for contiguous wholesale blocks. Add one: EQIX's own JVs, which take the hyperscale economics 75-80% off Equinix's own P&L while leaving Equinix with the development risk, the guarantees, and the receivable.
• What must hold for $1,053: 10%+ AFFO/share growth ex-fees through 2029, mid-20s yields on $5-7B/yr of new capital, a credit market that keeps funding a $3B/yr gap at reasonable spreads, and interconnection pricing that stops decelerating. Knock the AFFO/share CAGR to 6% and re-rate to 19x → ~$820-875 — right at the top of the 52-week low zone.
• Single scenario that permanently impairs: still very few. Interconnection remains a genuine annuity and the 91% recurring base is real. The asymmetry the prior dossier identified is unchanged and is the whole argument for patience over chase: hard to break, easy to de-rate — and this quarter handed the de-rating case its best evidence in two years.

Lens 14 (Management Questions) — carried, see the previous dossier. Questions 3 (CFO transition), 5 (derivative suit) and 4 (cooling ceiling) are now answered or moot. Three replacements, in priority order, arising from this filing:

• (new 1) Of the ~$120M of Hampton xScale fees in Q2 — what is the cash-collection schedule, what is the margin, how much of it was eliminated against your 20-25% ownership, and what is the expected cadence of comparable JV fees across 2027-2029? Should we model them as recurring at all?
• (new 2) Gross capex is up 66% and net PP&E up $1.6B in six months, while recurring capital expenditures were flat at $81M. What objective test allocates a dollar between recurring and growth, has it changed since the SEC review, and at what level of the $5-7B/yr programme would you expect recurring capex to scale?
• (new 3) You need roughly $3B a year of external capital through 2029 and just priced 5.0-5.8% coupons into a softening AI-credit market. At what spread does the $5-7B capex plan get cut rather than funded — and does the ATM come back before the plan does?

---