Switch

Phase A — Understand the business

Lens 1 · Company Overview

Switch, Inc. is a Las Vegas-headquartered hyperscale colocation and data-center developer — it builds, owns, and operates very large, very dense, very redundant data-center campuses and leases capacity (space + power + cooling + connectivity) to enterprises, cloud providers, and now AI neocloud tenants. Founded in 2000 by Rob Roy (still Founder/CEO/Chairman), it built the SUPERNAP facilities and coined its own reliability grade, Tier 5® Platinum.

Business model — in plain terms. Switch is a landlord for computers, at the top of the quality spectrum. Unlike a wholesale developer that hands over a shell, or a retail colo that rents by the cabinet, Switch's pitch is the most resilient, most efficient, densest facility in the market, at scale — "fault-sustainable," 100% renewable, PUE 1.18. Revenue is recurring colocation on multi-year contracts (historically ~good retention; eBay has been an anchor tenant for a decade). The economics are infrastructure-REIT-like: heavy up-front capex, long-lived assets, contracted cash flows, financed increasingly with asset-backed securitizations (green ABS/CMBS).

Key products / campuses (the "PRIME" exascale campuses):

• Las Vegas — "The Core": ~2M sq ft, up to ~275MW. The original SUPERNAP cluster and eBay's anchor home.
• Tahoe Reno — "The Citadel": designed for up to ~7.2M sq ft and, per company claims, >2GW at full build-out — pitched as the world's most powerful colocation campus, sited next to Tesla's Gigafactory.
• Grand Rapids — "The Pyramid" (Michigan): ~1.8M sq ft, ~120MW.
• Atlanta — "The Keep": Southeast US hub.
• Austin + Houston (Texas): added post-take-private; AI-factory build-out underway.

The AI pivot — "AI Factories." Post-2022 Switch has re-platformed around EVO, Rob Roy's hybrid air+liquid cooling design claiming up to 2MW per cabinet and modular density "to gigawatt scale". The marquee proof point: Switch hosts CoreWeave's landmark NVIDIA GB300 NVL72 deployment — the first commercial Blackwell Ultra racks — on the EVO design. It is also integrating NVIDIA's Omniverse DSX blueprint for AI-factory design.

Customers / suppliers / competitors — detailed in Lens 2. Headline: >1,000 customers historically, 40+ cloud providers, anchored by eBay; now leaning hyperscale + neocloud (CoreWeave). Key supplier relationship: Schneider Electric ($1.9B supply-capacity agreement to power the AI factories). Competitors: Equinix, Digital Realty, CyrusOne, QTS, Vantage, Aligned, CoreWeave (as both customer and increasingly a self-builder).

Contract structure. Historically recurring multi-year colocation with the standard colo mix of base rent + power pass-through/metered power; concentration was real (see Lens 2). The AI-factory deals appear to be large, long-duration, build-to-suit-style anchor leases (the CoreWeave GB300 host arrangement), which is the industry-wide shift from many small tenants to few enormous ones. Exact take-or-pay terms are not disclosed — private [n/a — private, not disclosed].

Lens 2 · Supply Chain

Map: power + land + long-lead equipment → Switch (design/build/operate) → enterprise / cloud / AI tenant. Named stakeholders along the chain:

Upstream — power & land (the binding constraint):

• Renewable power: Switch Station 1 & 2 solar (Apex Industrial Park, NV) — ~179MW combined; long-term renewable PPAs delivering power at a cited ~4.9¢/kWh. Nevada / Texas siting chosen partly for power availability and cost. Power — not chips — is the true chokepoint for the whole sector.
• Land: large contiguous parcels (Tahoe Reno Industrial Center, ~2,000 acres). Historical single-source flag: land at the Las Vegas campus was tied to Beltway Business Park, an entity affiliated with founder Rob Roy — a related-party dependency (Lens 10/13).

Upstream — critical equipment (long-lead, AI-era bottleneck):

• Schneider Electric — $1.9B supply-capacity agreement for the power infrastructure (switchgear, UPS, prefabricated power) of the AI factories. This is the single most important named supplier relationship — locking Schneider capacity is itself a moat in a market where electrical gear lead-times blew out.
• NVIDIA — reference-design partner (Omniverse DSX blueprint); the GB300 racks Switch cools are NVIDIA silicon.
• Dell — historically both a customer and a server supplier; supplied the servers in CoreWeave's Switch-hosted GB300 deployment.
• Cooling/mechanical, generators, transformers — Switch's EVO is proprietary but relies on the same strained CDU/liquid-cooling and transformer supply chain as every peer.

The company (midstream): Switch's own IP is the design (500→950+ patent claims per company figures) and the construction/operations org (Terri Borden, Chief Construction Officer; in-house build).

Downstream — the buyers (the demand side):

• Anchor / hyperscale / neocloud: eBay (legacy anchor, was ~13.5% of revenue pre-buyout), CoreWeave (marquee AI tenant), plus historical roster of Google, Cisco, VMware, Microsoft (Xbox), Intel, Apple, PayPal, FedEx, Box.
• The sell-side agent: Switch retained CBRE as exclusive leasing agency for the Tier-5 PRIME colocation product.

Chokepoints: (1) grid interconnect / megawatts — the whole model gates on securing power; (2) electrical gear (mitigated by the Schneider lock); (3) capital — the build is financed by continuous ABS issuance, so credit-market access is a supply-chain input (Lens 5). A frozen securitization market would choke the build as surely as a transformer shortage.

Lens 3 · Competitive Advantages (moats)

Real moats:

1. Land + power + interconnect at scale, already permitted and energized. In 2026 the binding constraint on data-center growth is power you can actually plug in. Switch's Tahoe Reno and Nevada footprint — contiguous acreage, dedicated solar, cheap long-term PPAs, existing substations — is a genuinely scarce, hard-to-replicate asset. DigitalBridge explicitly framed the acquisition as solving the California/Western capacity crunch. This is the durable moat.
2. Density / engineering reputation. EVO's claimed 2MW/cabinet and the Tier 5 brand let Switch win the hardest AI workloads (GB300-class) that lower-tier facilities physically cannot cool. Being the site CoreWeave chose for the first commercial Blackwell Ultra racks is a credible signal, not just marketing.
3. PUE 1.18 + 100% renewable + green-financing flywheel. Best-in-class efficiency (vs. Equinix ~1.39, Digital Realty targeting ~1.5) both lowers operating cost and unlocks cheaper capital — every Switch ABS is a designated green bond, widening the buyer base and cutting the coupon. Efficiency → cheaper capital → more build is a self-reinforcing loop.
4. Switching costs. Migrating a live production estate (especially a dense AI cluster with custom networking) out of a facility is expensive and risky; anchor tenants like eBay stayed a decade.

Bargaining power. Over suppliers: strong and growing — a $1.9B Schneider commitment gives Switch priority allocation rivals can't match. Over customers: weakening at the margin as the buyer mix concentrates into a handful of hyperscalers/neoclouds who can (and do) build their own. A CoreWeave or a Microsoft has real leverage on price and terms; a 1,000-tenant enterprise book had none. This is the central tension in the bull/bear.

Where the moat is thin: the physical assets are excellent but the category is capital — anyone with land, power, and $10B (Blackstone/QTS, KKR/CyrusOne, BlackRock/Aligned, Vantage) can build a competing Tier-4+ campus. Switch's edge is time and reputation, not a patent nobody can design around. The "Tier 5" standard is self-certified — the Uptime Institute does not recognize a Tier V — so the brand premium rests on operating record, not third-party imprimatur.

Lens 4 · Segments

Switch does not report segment financials as a private company [n/a — private, not disclosed]. The last public segmentation (FY2021 10-K era) was effectively single-segment (US colocation) with the meaningful cuts being by campus and by customer:

Trend & cause: the mix is shifting from many enterprise tenants to fewer, larger AI/hyperscale anchors and from Nevada-centric to a five-campus national footprint. That raises revenue scale and duration but increases single-customer concentration — the opposite of diversification. Directionally accelerating on revenue, but the quality of that revenue (concentration, counterparty) is the thing to watch, and it is unauditable from outside.

---

Phase B — Measure performance

Lens 5 · Funding & Valuation Trajectory (+private lens-swap for "Earnings Result")

Switch has no earnings prints. The equivalent tape is its capital-raising trajectory, which is unusually legible for a private and tells the story cleanly.

The last audited baseline (FY2021, the final full public year):

• Revenue $592.0M; Adjusted EBITDA $315.1M (~53% margin); Net income $14.8M.
• Pre-buyout debt included $600M 3.75% senior notes due 2028 + $500M 4.125% senior notes due 2029.

The take-private (the entry mark):

• December 6, 2022 — DigitalBridge + IFM completed the take-private at $34.25/share, ~$11B including debt. That is ~35x FY2021 adjusted EBITDA — a full data-center-platform multiple in line with the 25–30x era (Lens 7).
• Ownership post-deal: DigitalBridge 55.8% / IFM 37.2% / Switch management 7.0%; Aware Super (Australian super fund) took a minority stake in 2023.

The re-marking (2024 → 2026, the story):

• Sept 10, 2024 — Reuters: owners in preliminary talks to IPO Switch at ~$40B including debt, as early as 2025. ~3.6x the 2022 entry in <2 years.
• 2025–2026 — a new $2B private round led by Andreessen Horowitz ($400M commitment), Goldman Sachs + JPMorgan arranging financing, at ~$50B including debt / ~$19B equity pre-money. >4x the 2022 price in ~4 years.

Balance-sheet / financing signals (the "print" that matters most for a levered developer):

• ~$20B raised since 2024 across sustainability-linked/green loans, green bonds, and securitizations.
• Green ABS/CMBS: a $2.4B SASB CMBS + $1.1B ABS in 2024–25; $659M fourth ABS (Oct 2025); a subsequent $768M issuance — Switch is now the largest single issuer of data-center ABS in the period.
• Credit facilities / borrowing base upsized to ~$10B (Jul 2025); stated goal to retire 100% of the bank debt from the 2022 take-private and lower cost of capital.

Read: this is a company being aggressively marked up by its sponsor on AI-factory demand, funding a very large build with continuous green-securitization rather than equity dilution, and lining up a re-IPO / crossover round. The trajectory is genuinely impressive — but every number here is a sponsor-influenced private mark or a debt raise, not an audited P&L. The absence of a disclosed current revenue/EBITDA figure is itself the most important fact in this lens. Burn signal: raising ~$20B of mostly debt against undisclosed EBITDA is the classic AI-infra "build ahead of contracted cash flow" posture — powerful in a bull tape, fragile if leasing or credit markets stall.

Lens 6 · Founder / Management Signal (+private lens-swap for "Earnings Calls")

No earnings calls. The sentiment proxies are sponsor commentary + press cadence + founder posture:

• DigitalBridge consistently frames Switch as a crown-jewel AI-infrastructure asset inside a ~$108B-AUM, ~5.4GW platform with a "record 5GW US hyperscale commitment". Sponsor tone = uniformly bullish (they are marking and monetizing the asset).
• Press cadence since 2024 is relentlessly positive and capital-formation-heavy: ABS after ABS, the Schneider $1.9B, the CoreWeave GB300 host, NVIDIA DSX — a deliberate drumbeat consistent with pre-IPO narrative-building.
• What to discount: there is no adversarial voice in the flow — no short-seller, no quarterly Q&A, no independent audit commentary — because there's no public float. Sentiment here is management-controlled by construction, which is exactly why the governance lenses (9/10/13) carry more weight than usual.

Lens 7 · Cap Table, Secondary Marks & Comps (+private lens-swap for "Comps")

Cap table / syndicate quality — an IPO-proximity read:

• Sponsors: DigitalBridge (55.8%, control), IFM Investors (37.2%), management (7.0%), Aware Super (minority, 2023).
• Crossover tell: the reported a16z-led ~$2B round with Goldman + JPMorgan arranging is precisely the late-stage-crossover pattern that precedes an S-1. A tier-1 growth investor coming in at ~$50B is an IPO-runway signal.
• Ultimate-parent flux (critical): DigitalBridge itself is being acquired by SoftBank for ~$4B ($16.00/share), announced Dec 29 2025, expected to close H2 2026 — which sweeps Switch (with AtlasEdge, Yondr, Vantage) into the SoftBank orbit and, per SoftBank, "DigitalBridge will continue to operate as a separately managed platform led by Marc Ganzi". SoftBank ownership + Stargate ambitions could accelerate a Switch monetization — or complicate the IPO timeline.

Secondary / step-up marks (the valuation tape):

Mechanism comps — recent data-center platform valuations (the "is ~$50B sane?" check):

Lens 8 · Valuation / Funding Catalysts (+private lens-swap for "Stock-Price Catalysts")

No stock, so the events that re-rate the private mark:

• Nov 2017 → 2022 (as SWCH): IPO'd at $17 (Oct 2017, jumped >22% day one), then de-rated to ~$11.85 by Jun 2018 on the profitability/capex disappointments (Lens 10), traded rangebound, and exited at $34.25 in the 2022 take-private. The public-market lesson: the market punished Switch hardest on capex overruns and the Grand Rapids/Atlanta build economics, and rewarded it on the take-out.
• Dec 2022 — take-private closes (entry mark, ~$11B).
• 2024–2025 — AI-factory demand (the CoreWeave GB300 host win, Schneider $1.9B) is the re-rating engine; the ABS drumbeat validates the build.
• Sep 2024 — Reuters IPO-at-$40B report (first public up-mark).
• 2025-26 — a16z ~$2B round at ~$50B (crossover up-mark).
• Dec 2025 / H2 2026 — SoftBank buys DigitalBridge → new ultimate parent, potential accelerant or complication for a Switch exit.
• Forward catalysts that would confirm/break the mark: a Switch S-1 filing; a named hyperscaler anchor on a multi-hundred-MW AI lease (à la the Applied Digital/CoreWeave $7B–$11B deals); the SoftBank deal closing and its stance on monetizing Switch; the next ABS spread (a widening tells you credit markets are pricing more risk).

---

Phase C — Judge people & books

Lens 9 · Management

Rob Roy — Founder, CEO, Chairman, chief inventor. The company is Rob Roy. He conceived the SUPERNAP design, holds 500 → 950+ issued/pending patent claims (the figure inflates in company copy over time — itself a mild promotional tell), and defined the Tier 5 standard.

• Track record: built one of the most respected data-center engineering franchises in the US from a small Las Vegas colo, IPO'd it (2017), and exited public holders at a premium (2022). Genuinely built something hard and durable — quantified: ~$592M revenue / ~$315M EBITDA by 2021 from a standing start.
• Skin in the game: management held 7.0% post-buyout; as founder Roy retains substantial equity. High alignment — but see the founder-control flags.
• Capital allocation: aggressive reinvestment into ultra-high-spec builds; the quality of that allocation was the public-market controversy — bulls saw a durable premium product, bears saw capex-heavy campuses (Grand Rapids/Atlanta) that never matched Las Vegas returns.
• Archetype: founder-engineer with 10:1 super-voting control — visionary product builder, not a capital-markets professional. Implication at this stage: brilliant at building the asset, but the governance (super-voting, related-party history) is exactly what a public-market re-listing would have to clean up.

Thomas Morton — President & Chief Legal Officer. Joined 2008; served as CFO and General Counsel through the 2017 IPO. The institutional/legal counterweight to Roy; deep tenure, significant equity. Continuity through the private era.

Other execs: Chris Donnelly (Chief Connectivity Officer), Terri Borden (Chief Construction Officer) + a broad bench. The construction capability being in-house is a real operating asset in a build-constrained market.

Sponsor overlay: DigitalBridge (Marc Ganzi) controls the board post-2022; IFM + Aware Super as institutional co-owners; SoftBank (Masayoshi Son) incoming as ultimate parent H2 2026. The people question shifts from "trust the founder" to "trust the founder plus an infra-PE sponsor plus SoftBank's appetite" — a more complex, more conflicted cap table than a standalone public company.

Lens 10 · Forensic Red Flags

(Forensic view. No audited post-2022 financials exist — the biggest red flag is the absence of them. What follows is the public-record forensic history + the structural risks a diligence analyst must price.)

From the public-company era (documented):

1. The 2017 IPO capex-disclosure controversy → 2018 securities class action. Plaintiffs alleged Switch made false/misleading statements around the IPO: that Grand Rapids/Atlanta would never be as profitable as Las Vegas; that high-redundancy capex was less profitable than represented; that >$64M of unbudgeted Q3-2017 capex wasn't disclosed until after the IPO; and that $9.4M of FY17 revenue was recognized for services not to be delivered until FY18. This is a revenue-recognition + capex-transparency flag — squarely the kind of thing to re-diligence in any S-1.
2. Related-party real estate — Beltway Business Park. Land/leases at the Las Vegas campus involved Beltway Business Park, an entity affiliated with Rob Roy — a classic founder related-party dependency (the company leased/purchased property from a founder-linked entity). Exact economics not re-verified here; flagged as a governance item, not a proven abuse.
3. Founder super-voting control / "controlled company." Roy held ~67.7% of voting power via 10-votes-per-share Class C stock and Switch elected "controlled company" exemptions from NYSE independence rules. Minority holders had structurally weak governance rights.
4. Tax Receivable Agreement (Up-C structure). Switch used the standard Up-C/TRA structure at IPO, which typically routes ~85% of certain tax benefits to the pre-IPO owners (Roy et al.) — a common but real value-leakage-to-founder mechanism.

From the private era (structural, unauditable): 5. No audited financials since FY2021. All post-2022 valuation and revenue signals are sponsor-influenced private marks and press releases — the single largest forensic caveat. 6. Heavy, continuous securitization against undisclosed cash flow. ~$20B raised, largely ABS/CMBS/green-loan debt, with no public EBITDA to size leverage against. Green-bond framework has a Sustainalytics second-party opinion (a positive), but ABS structures can flatter headline capacity while embedding real refinancing/lease-up risk. Watch the ABS spread trend as the honest market signal. 7. Self-certified "Tier 5" standard — a marketing construct the Uptime Institute does not recognize — a (minor) promotional-behavior flag consistent with the pattern.

Where cash flow could diverge from the story: impossible to test directly (no statements). The structural worry is capex and interest outrunning contracted lease revenue during the AI build — the exact failure mode the 2017–18 episode rehearsed, now at 10x the scale and with far more leverage.

Regulatory findings (required sub-section). Per regulatory/regulatory-findings.md: Switch has no CIK and is not an SEC filer — zero EDGAR LR/AAER results possible. Non-SEC web check ("Switch" (FTC OR DOJ OR FDA OR CFPB OR consent decree OR settlement OR fine) enforcement) surfaced no material government enforcement action against Switch, Inc. the data-center operator. The only material litigation on record is shareholder litigation, both resolved in the company's favor: the 2018 securities class action, and a stockholder challenge to the 2022 take-private that Latham & Watkins got dismissed with prejudice (Aug 2023). Item 3 (Legal Proceedings) from a current 10-K is unavailable — no filer. Summary: no regulatory/enforcement findings; the only legal history is shareholder litigation, all resolved for Switch — verified via the regulatory-findings file (no SEC record possible), web search, and public case records as of 2026-07-07. Private company: unaudited per public sources.

---

Phase D — Project & stress-test

Lens 11 · IPO-Readiness & Path-to-Tradeable (+private lens-swap for "Forward Projection")

(No EPS model — private. The +private question is: when does this become tradeable, and at what mark?)

IPO-readiness assessment:

• Stage: late-stage private, sponsor-owned, actively pre-IPO — the most IPO-ready posture short of a filed S-1. Reuters-reported IPO talks (2024) + a crossover round (2025-26) + a decade of prior public-company reporting machinery (they were public 2017–2022, so the disclosure controls and audit history exist and can be dusted off).
• Readiness (my scale, 1–5): ~4/5 — filing-capable within 2–4 quarters if the sponsor chooses; the gating items are market window and the SoftBank/DigitalBridge deal, not internal readiness.
• Milestones that unlock the S-1: (1) SoftBank–DigitalBridge deal closes (H2 2026) and SoftBank signals it will monetize rather than hold Switch; (2) a named multi-hundred-MW hyperscaler anchor contract to underwrite the AI-factory revenue story; (3) AI-infra IPO window stays open (CoreWeave's public reception is the read-through bellwether); (4) a couple of clean audited years post-buyout to reset the numbers.
• Estimated window: 2026–2027 for an S-1, contingent on the SoftBank deal — SoftBank could equally decide to keep Switch private inside its AI-infrastructure empire (Stargate-adjacent), which would push the tradeable event out or convert it to secondary-only liquidity.

Valuation path (all /, unaudited):

• Entry (2022): ~$11B. IPO talk (2024): ~$40B. Crossover (2025-26): ~$50B EV / ~$19B equity.
• Base case: an IPO/monetization in the ~$40–55B EV range if the AI-factory lease-up is real and the window holds — i.e., the crossover mark, plus or minus the tape. This is a bet on EBITDA compounding into the multiple, not a discount to it.
• Bull: >$60B if Switch lands a marquee hyperscaler anchor and prints CoreWeave-like contracted-revenue backlog before listing; the Aligned $40B and the AI-infra re-rating support the upside.
• Bear: $20–30B (still ~2–3x the 2022 entry) if AI-infra sentiment cools, ABS spreads widen, the anchor-tenant concentration spooks IPO buyers, or SoftBank's own leverage forces a slower/cheaper monetization. A down-round from ~$50B is entirely possible — private marks are not floors.

No Brier forecast logged (private, no binary EPS/readout line, and --watchlist skips the create step). The scoreable question to track: "Switch files an S-1 or completes a monetization event at ≥$40B EV by end-2027." I'd put that at ~45–55% — real but hostage to the SoftBank deal and the AI-capex cycle.

Lens 12 · Bull vs Bear

Bull case. Switch is a scarce, already-energized, best-in-class US colocation platform at the exact moment power-constrained AI demand is the tightest bottleneck in tech. It owns contiguous gigawatt-scale campuses with dedicated cheap renewable power, the industry's best PUE (1.18), a proprietary 2MW/cabinet design that wins the hardest AI workloads, a $1.9B Schneider supply lock, the marquee CoreWeave GB300 deployment as proof, and a sponsor (DigitalBridge → SoftBank) with the capital and ambition to fund a multi-gigawatt build via the deepest green-ABS program in the sector. The comp set (Aligned $40B, CyrusOne 26x, QTS, AirTrunk) says top-tier platforms clear ~25–30x and $40B+ marks. A re-IPO or SoftBank-driven monetization at $40–55B is a credible, arguably conservative, outcome. Contrarian bull: the market is treating Switch as "just another colo" being marked up in a frothy cycle; it is actually one of a tiny number of operators who can physically deliver 2MW-cabinet AI factories at scale with power in hand today — the CoreWeave host win is the tell that it's a capacity oligopolist, not a commodity landlord.

Bear case (permanent-impairment risks).

1. Customer concentration into counterparties that can build their own. The revenue base is shifting from 1,000 diversified enterprises to a handful of hyperscalers/neoclouds (CoreWeave chief among them). If CoreWeave (itself a leveraged, customer-concentrated, Microsoft-dependent AI-cloud) stumbles, or if hyperscalers inssource, Switch's growth and its anchor cash flows are impaired simultaneously. Concentration risk compounds counterparty risk.
2. Leverage into an unaudited, cycle-sensitive build. ~$20B raised, heavily debt, against undisclosed EBITDA, to build ahead of contracted demand. If AI-capex digests (even a pause), lease-up slows while interest and refi come due — the 2017–18 capex-overrun episode at 10x scale and far more leverage. ABS-spread widening is the canary.
3. The mark is narrative, not cash flow. ~$50B EV sits on an FY2021 audited base of ~$315M EBITDA. The entire valuation is forward AI lease-up. A cooling AI-infra IPO tape (a weak CoreWeave stock, a stalled Stargate) could reset the private mark hard — down-rounds from ~$50B are on the table.

Pre-mortem (18 months out, thesis broke): The AI-capex cycle took a breather in late 2026; a marquee neocloud tenant renegotiated or defaulted; Switch's next green ABS priced 150bp wider; the SoftBank–DigitalBridge deal closed but SoftBank — stretched by its own AI commitments — shelved the Switch IPO and pursued a cheaper secondary. The ~$50B mark quietly became a ~$30B mark, and the "story" holders learned that private marks are opinions, not prices.

Are multiples too high? At a multiple-on-today's-cash-flow basis, self-evidently yes (n/a, but ~$50B on ~$315M last-audited EBITDA is not a running-yield valuation). At a platform-scarcity, forward-lease-up basis, in line with Aligned/CyrusOne-era prints. You are not buying cash flow; you are buying optionality on the US AI-power build. Price it as such.

What the market refuses to see: on the bull side, that power-in-hand at gigawatt scale is a harder moat than the "commodity colo" framing admits. On the bear side, that a private mark set by the seller (DigitalBridge, who is itself being sold) is a conflicted number, and that the same super-voting founder + related-party history that dogged the 2017 IPO hasn't gone anywhere.

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

Dismantling the bull case:

• The valuation is a sponsor's mark, and the sponsor is a forced-ish seller. DigitalBridge marked Switch up ~4.5x in ~4 years and is itself being acquired by SoftBank. Every "up-round" here benefits the people setting the price. There is no audited cash flow, no public float, no short interest, no independent price discovery — the $50B is an assertion, not a transaction you can sell into.
• Revenue concentration is the structural break. The AI pivot concentrates the book into a few tenants — and the flagship, CoreWeave, is one of the most fragile counterparties in the entire complex (leveraged, ~62% Microsoft-dependent, GPU-collateral financing). Switch is building gigawatts of highly specific capacity for customers who (a) can build their own and (b) may not survive an AI-capex air pocket. That's not diversified recurring colo; it's a concentrated bet on a handful of neocloud balance sheets.
• The moat is capital, and capital is everywhere right now. Blackstone (QTS/AirTrunk), KKR (CyrusOne), BlackRock/MGX ($40B Aligned), Vantage, plus the hyperscalers' own build — the most dangerous competitor bulls underestimate is the customer. When Microsoft/Meta/CoreWeave self-build (as they increasingly do), Switch's pricing power and lease-up both erode.
• Governance is unreformed. 10:1 super-voting founder, "controlled company" history, related-party Beltway real estate, an Up-C/TRA that routes tax benefits to insiders, and a 2018 securities class action over exactly the capex/rev-rec issues that recur in a heavy build. None of this is disqualifying, but it all resurfaces in an S-1 and caps the multiple public buyers will pay.
• Assumptions that must hold for ~$50B: AI-capex stays vertical; the AI-infra IPO window stays open; ABS markets keep funding the build at tight spreads; the anchor tenants stay solvent and don't insource; SoftBank chooses to monetize at a premium. If growth disappoints 20–30%, this is a forward-lease-up story with no running yield to cushion it — the mark compresses fast, plausibly to $25–35B, and "story" holders discover the exit is illiquid.
• The single scenario that permanently impairs: a neocloud anchor default coincident with an ABS-market freeze — capacity built, tenant gone, refinancing shut. Plausibility: moderate, not remote — it is the sector's defining tail risk in 2026-27.

Lens 14 · Management Questions (ordered by information value)

1. What are current (last-twelve-month) audited revenue, adjusted EBITDA, and unlevered FCF — the single number every valuation above depends on and the only thing not disclosed?
2. What share of contracted revenue is CoreWeave and the top-3 tenants, and what is the weighted-average remaining lease term and counterparty-credit profile of that concentration?
3. Of the ~$20B raised since 2024, how much is drawn vs. committed, what is the blended cost of capital and near-term maturity wall, and how does net leverage compare to EBITDA?
4. What are the terms (take-or-pay? MW-based? escalators? exit rights?) of the AI-factory anchor leases underwriting the growth story?
5. How does the SoftBank acquisition of DigitalBridge change Switch's IPO timeline, governance, and monetization path — and does SoftBank intend to list Switch or hold it?
6. What is the contracted vs. speculative split of the multi-gigawatt build pipeline — how much is leased before you pour concrete?
7. How have your green-ABS spreads moved across the last four issuances, and what does the trend say about how credit markets price your lease-up risk?
8. What is the status of the Beltway Business Park related-party arrangements today, and what related-party transactions with the founder or sponsors remain?
9. Will the 10:1 super-voting Class C / controlled-company structure survive a re-listing, and if so why should public minority holders accept it a second time?
10. What is actual delivered capacity today (energized, contracted MW) vs. the "designed for" gigawatt headline numbers — the gap between marketing and metered?
11. What are your power-interconnect queue positions and timelines across the five campuses, and where is grid access the binding constraint on the build?
12. How exposed is the model to a hyperscaler-insourcing shift, and what stops your largest tenants from building their own next-generation capacity?
13. What is the customer-retention / renewal rate in the private era, and has it changed as the mix shifted from enterprise to hyperscale/neocloud?
14. How real is the "net-positive water" / 100%-renewable claim on a metered, third-party-verified basis across all campuses — or is it partly REC/offset accounting?
15. What would trigger you to slow the build, and what is the downside financing plan if the AI-infra capital markets tighten?

---