Equinix

Phase A — Understand the business

Lens 1 · Company Overview

Equinix is the world's largest retail colocation and interconnection REIT — a global, vendor-neutral data-center platform of 280 data centers across 77 markets (incl. 23 xScale facilities), housing customers' IT equipment and, crucially, letting them physically cross-connect to each other. The business is not "renting space" — it is selling proximity and connectivity. As more networks, clouds and enterprises colocate in an Equinix IBX, every existing tenant's value rises (a true adjacency network effect), which the 10-K names explicitly as the core flywheel.

Revenue model — recurring, sticky, diversified:

• Recurring revenue (colocation + interconnection + managed infrastructure) has been >90% of total for three straight years; >90% of new MRR bookings come from existing customers. FY2025: 94% recurring ($8,739M of $9,217M).
• Contracts run 1–5 years, auto-renewing in 1-year increments.
• Customer concentration is strikingly low for an AI-era infrastructure name: largest customer ≈ 3% of recurring revenue; top 50 ≈ 36%; no single customer ≥10% of revenue or receivables. This is the opposite of the hyperscale wholesale model (where one tenant can be 20%+).

Customers: global enterprises, network/cloud service providers, and — increasingly — AI model builders. Per the Q1 2026 call, 8 of the top 10 AI model providers and 4 of the top 5 neoclouds are expanding inside Equinix.

Key operating tells (FY2025): surpassed 500,000 interconnections ("the most in the industry"); $1.6B annualized gross bookings, +27% YoY; cabinet utilization 77%; 16 new data centers opened; 52 active development projects across 35 metros (55,000+ retail cabinets + 100+MW xScale through 2028).

Gaps vs. KB: the kb/datacenters/wiki/* commercial layer (positioning, supply-chain, bottlenecks) is missing on disk — Lenses 1-4 are grounded in the filing + web rather than a compiled competitive matrix. Flag for kb build-out.

Lens 2 · Supply Chain

Equinix sits in the middle of the digital-infrastructure stack — it is both a buyer of physical build inputs and the landlord/connector for the compute economy.

Upstream (what EQIX buys):

• Power — the binding input. Cost of revenue is dominated by electricity; the 10-K flags power availability as the #1 growth constraint and notes ~$2.1B of non-capital purchase commitments "such as commitments to purchase power in select locations".
• Land & real estate — closed strategic land in Amsterdam, Chicago, London, Milan, Mumbai, Toronto for ~1 GW of future capacity.
• Data-center equipment & liquid cooling — generators, cooling, network gear; the 10-K explicitly warns of "global supply chain challenges" forcing advance purchase commitments ($6.3B unaccrued capex contractual commitments).
• Capital — the true upstream input for a REIT: $4.4B raised in 2025 ($4.3B senior notes + ATM equity).

The company: designs, builds and operates IBX data centers; runs the interconnection fabric (Equinix Fabric / cross-connects).

Downstream (who buys from EQIX) — named:

• Cloud / hyperscalers as tenants and ecosystem anchors (AWS, Azure, Google Cloud, Oracle — the cloud on-ramps live inside IBX).
• AI model providers — 8 of top 10 (the OpenAI/Anthropic/Google/Meta tier) + 4 of top 5 neoclouds (CoreWeave-type).
• Networks / carriers / ISPs — the original tenant base; the reason interconnection density compounds.
• Enterprises — the highest-margin, stickiest cohort.
• xScale JV partners — for hyperscale builds, EQIX uses joint ventures (e.g. the new AMER 3 JV that bought the Hampton/Atlanta campus for ~$470M, Jan 2026) so the capital sits partly off-balance-sheet.

Chokepoints / single-source dependencies: (1) grid power & interconnect queues — the hard ceiling on AI-grade supply; (2) liquid-cooling retrofit pace — management says cooling retrofits, not new builds, are the near-term ceiling (deployments +50% QoQ in Q1 2026); (3) capital-market access for the debt-funded build. None is a vendor monopoly — the constraints are power, cooling and capital, not a sole supplier.

Lens 3 · Competitive Advantages (moats)

The moat is interconnection density, and it is the realest in the sector. Four reinforcing layers:

1. Network effect (the durable one). 500,000+ cross-connects is the largest neutral interconnection fabric on earth. Each network/cloud/enterprise that joins makes the platform more valuable to the next — a flywheel a wholesale landlord (raw space + power) cannot replicate. This is why EQIX earns a colocation+interconnection blend, not a commodity wholesale lease.
2. Switching costs. Once an enterprise has wired its production network, clouds and SaaS partners into an IBX, migrating means re-architecting connectivity — hence >90% of bookings from existing customers and 77% utilization.
3. Scale & geographic span. 280 IBX / 77 markets / 35 metros under construction. For a customer needing the same provider in Frankfurt, São Paulo, Singapore and Virginia, the candidate list is short (effectively EQIX or DLR).
4. Brand / vendor-neutrality. EQIX deliberately does not compete with its tenants (unlike a hyperscaler-owned facility), which is why clouds and rivals coexist inside its halls.

Bargaining power: over customers — high for interconnection (no substitute), moderate for raw colocation (DLR/wholesale compete). Over suppliers — moderate; EQIX is a large buyer of power and gear but is a price-taker on electricity and is exposed to grid timelines.

The honest moat caveat (sets up Lens 13): the network-effect moat protects the interconnection + enterprise business. It does not extend to AI training, which wants contiguous multi-MW blocks (wholesale/hyperscale), where EQIX participates only via capital-light xScale JVs. The moat is a fortress around the highest-margin 18% of revenue and a much shallower trench around the AI-training land-grab.

Lens 4 · Segments

Equinix reports three geographic segments (Americas / EMEA / Asia-Pacific) and discloses revenue by product line. All figures ``.

By geography — FY2025 revenue / Adjusted EBITDA ($M):

US revenue alone was $3.6B (≈39% of total). APAC carries the highest EBITDA margin (54.6%) — a quiet positive often missed.

By product line — FY2025 ($M), and the trend that matters:

Interconnection is growing faster than colocation and is the highest-margin line — and it accelerated in Q1 2026 to $446M, +13.5% YoY (from $393M). Managed infrastructure is flat/declining partly because Equinix Metal is being wound down ($29M Americas revenue headwind + $233M impairment taken in 2024) — a deliberate exit from a sub-scale bare-metal business.

Trend read: the mix is shifting toward the moat (interconnection up, low-margin managed/Metal pruned). Acceleration is real and is being driven by AI-adjacent demand, not financial engineering.

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print: Q1 2026, reported ~Apr 2026)

The quarter beat and the year was raised — recurring revenue is re-accelerating. All `` unless noted.

• Drivers: all three segments grew Adj EBITDA +16-18% with management crediting "disciplined operating expense management" — i.e. genuine operating leverage, not just revenue. Interconnection +13.5%.
• Bookings: Q1 annualized gross bookings $378M, a Q1 record (+9% YoY); ~60% of the largest deals were AI-related; ~25% of 2026 retail capacity expansion already sold.
• Guidance — RAISED: FY2026 revenue $10.12-10.22B (+9-10% normalized cc); Adj EBITDA $5.14-5.22B (51% margin, +~200bps); AFFO/share $42.31-43.11 (+9-11% cc).
• Balance-sheet flags: clean. Operating cash flow strong; the only "flag" is the scale of the build (investing outflow — see Lens 10/11).
• Market reaction: stock rose on the print; +28% over the prior three months on AI bookings + 2026 guidance.
• Vs. own history: this extends an 87+ quarter streak of consecutive revenue growth — one of the longest in US large-cap. Recurring growth accelerating from FY2025's +7% to Q1's +12% (reported) is the signal.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts/ on the shelf (transcripts=0) — grounded ``.

Tone arc, 2024 → 2026: from defensive to confidently offensive.

• 2024 (post-Hindenburg): the narrative was defense — "no restatement," "87th consecutive quarter of growth," audit-committee review, pulled bond deal. Management spent calls rebutting the AFFO/maintenance-capex accusation.
• Late 2025: pivot to AI demand and bookings records (Q4 2025 bookings $474M, +42%).
• Q1 2026: unambiguously offensive — "8 of top 10 AI model providers expanding," "4 of top 5 neoclouds," "~60% of largest deals AI-related," liquid cooling "+50% QoQ," guidance raised.

Recurring phrases (now): "AI inference," "interconnection density," "distributed AI," "liquid cooling," "disciplined opex." Stopped saying: the defensive accounting language of 2024 is gone — consistent with the SEC closing its probe (Lens 10). CEO Adaire Fox-Martin (since 2024) has reframed EQIX from "is the AI story real?" to "we are the inference + interconnection layer of AI".

Lens 7 · Comps

Data-center REIT peers. Multiples are `` with source/date; where unsourced, n/a.

Sources:; EQIX fwd AFFO is ``: $1,050 price ÷ $42.71 guidance-midpoint AFFO/sh = 24.6x. Sector net debt/EBITDA ≈ 5.1x; sector forward FFO ≈ 28.2x.

Read: EQIX trades at a ~40% P/FFO premium to DLR and a sub-2% yield — the market pays up for (a) the interconnection moat, (b) higher recurring %, (c) lower customer concentration. The premium is deserved but not cheap. On 5-yr avg ROE: n/a (REIT ROE is distorted by depreciation; AFFO/share growth of ~10-12% is the more honest return metric). EV/Sales and EV/EBIT for peers: n/a — not individually sourced this pass.

Lens 8 · Stock-Price Catalysts (the >5% movers, ~5yr)

Mostly ``; the pattern is unusually clean for this name.

1. Mar 20-25, 2024 — Hindenburg short report (DOWN, hard). Alleged AFFO inflated ≥22% via "maintenance capex booked as growth capex"; stock pummeled, EQIX pulled a bond offering. This is the single biggest catalyst in the name's recent history.
2. 2024-2025 — the slow exoneration (UP, grind). Audit Committee: no restatement; record bookings each quarter; SEC closes probe (Nov 2025); class action dismissed/settled by insurance (Dec 2025). Stock round-tripped the crater — 52-wk low $710.52 → ~$1,050 (+48%) into 52-wk high $1,128.68.
3. AI-demand re-rating (UP). Each bookings beat tied to AI (Q4 2025 +42%, Q1 2026 record) drove +28% in three months.
4. Guidance resets (UP/DOWN). As a recurring-revenue REIT, the AFFO/share guide is the swing factor — the Feb 2026 (FY guide) and Apr 2026 (raise) prints both moved it.

What the tape reveals: the market reacts to (a) accounting/governance credibility (the Hindenburg episode) and (b) the AFFO-per-share guide / bookings trajectory — not to single-customer news (there is no whale) or macro rates as sharply as a pure bond-proxy REIT (the AI growth optionality partly decouples it).

---

Phase C — Judge people & books

Lens 9 · Management

A platform mid-leadership-transition — the one genuine governance watch-item.

• CEO — Adaire Fox-Martin (President & CEO since 2024). Background: enterprise infrastructure / go-to-market (ex-Google Cloud, ex-SAP exec). Track record at EQIX so far: navigated the company out of the Hindenburg crisis and re-anchored the story on AI inference + interconnection; raised 2026 guidance. Still relatively early — judge on execution, not yet a long EQIX record.
• CFO — Keith Taylor, retiring in 2026 after 27 years (succession announced Dec 3, 2025; will advise ~1 year post-handover). Taylor was the architect of the REIT conversion and the AFFO framework — the institutional-memory holder. Losing the 27-year CFO one year after a new CEO, in the same window the AFFO accounting was under federal scrutiny, is a real (if benign-looking) succession risk. The handover is being staged deliberately, which mitigates it.
• Skin in the game / insider ownership: insider-transactions.csv absent; specific insider % n/a this pass. Professional-manager (not founder) archetype — appropriate for a $100B+ regulated REIT.
• Capital allocation: disciplined and improving. Pruned sub-scale Equinix Metal (impairment taken, exited); uses xScale JVs to fund hyperscale builds capital-light; recycling capital (sold Hampton campus to AMER 3 JV for ~$470M, Jan 2026); raised the dividend +10% to $5.16/qtr ($20.64/yr) at a ~53% AFFO payout — room to keep growing it. ROE/ROIC trend distorted by REIT depreciation; AFFO/share +10-12% is the cleaner read.
• Red flags: the Hindenburg allegation was specifically that metrics were manipulated to inflate executive comp — that is now legally closed with no restatement (Lens 10), but it is the reason management credibility is the swing factor for this stock. No related-party scandal, no promotional behavior beyond normal IR.

Lens 10 · Forensic Red Flags

Forensic equity-analyst lens. The headline here is that the bear's forensic case was tested by the SEC and DOJ and did not survive.

Accounting-risk surface (the Hindenburg thesis, examined):

• Maintenance vs. growth capex — the core allegation. EQIX discloses "recurring capital expenditures" as an explicit AFFO deduction: $284M (2025), $250M (2024), $219M (2023). The bear case was that more of the $4.3B gross capex is really maintenance and should reduce AFFO. Counter-evidence: (1) Audit Committee independent review found no error/restatement; (2) PwC audited ICFR and concluded it was effective as of 12/31/2025; (3) recurring-capex disclosure is granular and trending up sensibly with the asset base. The misclassification claim is judgment-laden (where to draw the maintenance/growth line is genuinely subjective) but has cleared audit + regulator.
• AFFO add-backs to watch (standard REIT, but flagged): SBC of $498M (FY25) is added back to AFFO — real economic dilution that non-GAAP AFFO ignores (shares 97.3M→98.2M YoY). Straight-line revenue/rent and contract-cost adjustments net out small. None is aggressive vs. REIT norms, but AFFO flatters the cash picture by the SBC add-back — a perennial REIT critique, not an EQIX-specific fraud.
• Cash flow vs. earnings: operating cash flow $3,911M comfortably exceeds net income $1,348M (as expected for a D&A-heavy REIT, D&A $2,066M) — no earnings-without-cash divergence.
• Leverage: ~$18.4B senior notes + $703M term/mortgage + $2.4B finance leases ≈ $21.5B gross debt; cash $3.2B → ~$18.3B net debt; PP&E $23.6B. Net debt/EBITDA ≈ 4.0x — below the 5.1x sector average. Plus $6.3B unaccrued capex commitments (mostly due within 12 months) — the real forward obligation. Manageable, investment-grade.
• Impairments: $68M (2025) and $233M (2024, mostly Equinix Metal + a Hong Kong IBX) — honest write-downs of sub-scale assets, not a pattern of serial charges.

Regulatory findings (required sub-section):

• SEC EDGAR EFTS (LR + AAER): No Litigation Releases and no AAERs naming Equinix in the 2021-06-21 → 2026-06-21 window.
• SEC investigation (the big one): opened via subpoena Apr 30, 2024; on November 19, 2025 the SEC notified Equinix it concluded the investigation and does NOT intend to recommend enforcement action. Clearing event.
• DOJ / US Attorney NDCA: subpoena Mar 20, 2024; company "does not expect any further related action from the NDCA".
• Securities class action (filed May 2, 2024; alleged false statements re: accounting/internal controls): motion to dismiss granted in part Jan 2025; settled, final approval Dec 19, 2025, dismissed with prejudice, covered entirely by insurance.
• Derivative suits: two NDCA derivative suits voluntarily dismissed (Apr-Aug 2025); one Delaware derivative suit (filed Aug 6, 2025) remains — motion to dismiss pending (adds an insider-trading allegation). This is the only live legacy item, and it is a nominal-defendant derivative claim, not a company-liability or regulatory action.
• Non-SEC (FTC/DOJ-antitrust/etc.): no material enforcement hits surfaced.
• ICFR: effective, PwC-audited, no material weakness, no restatement.

Verdict (Lens 10): Apart from one residual derivative suit, the entire 2024 forensic short-thesis has been adjudicated and closed without restatement or enforcement. The accounting is REIT-standard; AFFO's SBC add-back is the only legitimate ongoing quibble. Books: clean, post-clearing.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (FY2026 → FY2028, AFFO/share)

EQIX is a REIT — AFFO per share, not EPS, is the unit of value. Build is bottom-up off guidance + the disclosed growth algorithm. Output ``; inputs labeled.

Anchors:

• FY2025 actuals: revenue $9,217M; Adj EBITDA $4,530M (49.1%); AFFO $3,761M; AFFO/sh ≈ $38.33.
• FY2026 guidance (raised): revenue $10.12-10.22B; Adj EBITDA $5.14-5.22B (51%); AFFO/sh $42.31-43.11.

Driver assumptions (each labeled):

• Revenue growth: +9-10% cc sustained — supported by $1.6B bookings (+27%), AI = 60% of large deals, ~25% of 2026 capacity pre-sold. Recurring re-accelerated to +12% reported in Q1.
• Margin: +~200bps/yr of EBITDA-margin expansion is the explicit guide (51% in 2026 vs 49.1% in 2025) — operating leverage as fixed-cost IBX fill rises.
• Pricing: positive — power pass-throughs + pricing actions on renewals; interconnection pricing firm.
• Dilution: modest, ~+1%/yr from ATM + SBC (shares 97.3M→98.2M).
• Interest: rising absolute interest expense ($527M FY25, +15% YoY) as the $18.4B note stack grows with the build — a headwind to AFFO partly offset by EBITDA growth.

Three-year AFFO/share path:

Base assumes the disclosed ~+9-11% AFFO/sh algorithm holds for 2027 then mild deceleration. At ~$1,050, base FY2026 = 24.6x AFFO; FY2027 base ≈ 22.3x; FY2028 base ≈ 20.4x — the multiple compresses into the growth, which is the bull's core point.

Per --watchlist rules, no forecast.ts create is logged in the unattended loop. Suggested Brier seed for a later human pass: "EQIX FY2026 AFFO/share ≥ $42.71, p≈0.70, resolves 2027-02-15."

Lens 12 · Bull vs Bear

Bull case. Equinix is the one data-center REIT with a network-effect moat (500k+ interconnections), and that moat is being fed by AI's inference + distributed-AI phase exactly when the bears wrote it off. Recurring revenue is re-accelerating (+12% reported Q1), margins are expanding +200bps/yr on operating leverage, the AFFO/sh algorithm compounds ~10-12%, and the dividend grows off a comfortable ~53% payout. The Hindenburg overhang — the single thing that kept the multiple in check for two years — is now legally dead (SEC closed, class action settled by insurance). Capital allocation is sharpening (Metal pruned, xScale JVs, capital recycling). You are buying the toll-road of the interconnected AI economy with a clearing catalyst behind it.

Bear case (permanent-impairment risks). (1) AI's center of gravity is wholesale, not retail. Training clusters want contiguous multi-MW build-to-suit — DLR/CoreWeave/hyperscaler-self-build territory — where EQIX plays only via capital-light JVs. If "AI data-center demand" keeps meaning wholesale, EQIX captures the smaller, slower slice and the AI re-rating reverses. (2) Multiple + rates. 24.6x AFFO and a 1.9% yield is priced for sustained double-digit AFFO growth; a growth stumble and higher-for-longer rates (REITs are rate-sensitive; $18.4B of notes to refi over time) compress both the multiple and AFFO. (3) Power & cooling ceiling. Management itself says liquid-cooling retrofit pace — not demand — caps near-term AI supply; grid queues can stall the growth algorithm.

Pre-mortem (18 months out, thesis broke): AI demand concentrated in wholesale hyperscale; EQIX's retail bookings normalized to mid-single-digits; a rate back-up took the AFFO multiple from 24x to 18x; the stock round-tripped back toward the $710 Hindenburg low — not on fraud (that's closed) but on de-rating + decelerating growth. The CFO transition added a credibility wobble.

Are multiples too high? Full, not bubbly. 24.6x forward AFFO for ~10-12% AFFO/sh growth + a fortress moat is a fair-to-slightly-rich price (PEG-equivalent ~2x). The premium to DLR (~40% on P/FFO) is justified by mix/moat but leaves little margin for error.

Contrarian view (what the market refuses to see): The consensus narrative is "AI = wholesale = DLR/CoreWeave, EQIX is the boring REIT." The market is under-pricing that inference (not training) is the larger, more durable, more distributed workload over time — and inference wants exactly what EQIX sells: low-latency, interconnection-dense, metro-edge capacity next to enterprise networks. If inference out-grows training spend by 2027-28, EQIX is mis-bucketed as a laggard when it is the right vehicle.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case. Hindenburg already wrote the famous version; the accounting leg is dead, so here is the durable short.

• The moat protects the wrong 18%. Interconnection ($1.66B) is wonderful and defensible — but it is 18% of revenue. 70% is colocation, where wholesale players and hyperscaler self-build compete on $/kW. The AI capex supercycle is overwhelmingly flowing to multi-MW wholesale and to hyperscalers building their own — Equinix's retail+JV posture means it is renting a cabana at the edge of the AI ocean, collecting interconnection tolls but not the training-capacity gold rush.
• Concentration cuts both ways. Low single-customer concentration (great for stability) also means EQIX is not signing the mega 15-year, multi-hundred-MW AI anchor leases that re-rate DLR — its growth is a thousand small enterprise renewals, structurally capped near high-single-digits unless xScale scales (and xScale is JV, so EQIX only gets ~25-50% of the economics).
• Most dangerous competitor bulls underestimate: not DLR — the hyperscalers themselves (build-your-own) plus neoclouds (CoreWeave-type) that could disintermediate the colocation layer for the largest AI tenants, and DLR's wholesale scale for the contiguous blocks EQIX can't offer retail.
• Capital-allocation worst case: the $6.3B unaccrued capex commitment + rising interest expense ($527M, +15%) means the build is debt-and-equity-funded into a high-rate world; an equity raise to fund growth (ATM has $1.2B left) is dilutive at any wobble. The CFO who built the financing playbook is leaving.
• What must hold for $1,050: ~10%+ AFFO/sh growth through 2028 and a stable-to-falling rate backdrop and no AI-demand bypass to wholesale. Knock 20-30% off the growth (AFFO/sh CAGR 10%→6%) and re-rate to 18x (DLR-ish) → fair value roughly $800-850, i.e. ~20% downside, exactly the 52-wk-low zone.
• Single scenario that permanently impairs: structurally, very little — interconnection is a genuine annuity. The realistic de-rating (not impairment) scenario is AI-routes-to-wholesale + rates-up, which is a multiple/growth problem, not a broken business. That asymmetry (hard to break, easy to de-rate) is itself the case for patience over chase.

Lens 14 · Management Questions (ordered by information value)

1. What share of 2026-2028 incremental revenue do you expect from retail colocation + interconnection vs. xScale/hyperscale JVs — and how do the per-dollar economics (your share of EBITDA) differ between the two?
2. AI = ~60% of your largest Q1 deals — break that into training vs. inference, and tell us which you architecturally win and why.
3. With Keith Taylor retiring after 27 years, what specifically changes (and what is protected) in the AFFO methodology and capital-allocation discipline under the new CFO?
4. Liquid-cooling retrofit pace is your stated near-term ceiling — quantify the MW of AI-grade capacity gated by cooling/power today, and the timeline to clear it.
5. The Delaware derivative suit remains open — what is the realistic range of outcomes and cost, and does it touch the AFFO methodology again?
6. Defend the maintenance-vs-growth capex line: what objective test determines a dollar of capex is "recurring" (AFFO-reducing) vs. "growth," and has that test changed post-SEC review?
7. At ~24x forward AFFO and a 1.9% yield, how do you think about issuing equity (ATM) to fund the build vs. protecting per-share AFFO growth?
8. What is your power-procurement runway — contracted MW vs. pipeline MW — across your top 10 AI metros, and where are the grid-interconnect bottlenecks worst?
9. Interconnection is growing faster than colocation — is there a re-pricing or product cycle (Fabric, digital services) that structurally re-accelerates it, or does it track colocation long-term?
10. How do you defend the interconnection moat if hyperscalers and neoclouds build private interconnect fabrics that bypass neutral colocation for the largest tenants?
11. xScale is JV-funded — at what point does hyperscale demand justify bringing more of it on-balance-sheet, and what would that do to leverage (currently ~4x net debt/EBITDA)?
12. Net debt/EBITDA ~4x with $18.4B notes — what is your refinancing schedule and blended cost-of-debt sensitivity over 2026-2028?
13. Cabinet utilization is ~77% — is the ceiling power, cooling, or sales, and what utilization can the AI-era IBX design actually reach?
14. Which two markets/metros are nearest capacity, and what is the revenue you are leaving on the table for lack of supply?
15. Three years out, is Equinix primarily an interconnection company that happens to own real estate, or a real-estate company that happens to interconnect — and how should we value it accordingly?

---