DataBank

Phase A — Understand the business

Lens 1 · Company Overview

What it is. DataBank is the largest edge/enterprise colocation and interconnection operator in the United States by facility count — 65+ data centers, ~20 major interconnection hubs, across 25–27+ metros. It sells (a) retail and wholesale colocation (cabinets, cages, private suites) to enterprises, carriers, content and cloud providers; (b) interconnection (cross-connects, cloud on-ramps, IX access) — the high-margin, sticky layer; (c) managed services (managed hosting, cloud, backup/DRaaS, compliance-grade hosting for regulated verticals); and, since 2024–25, (d) hyperscale build-to-suit wholesale capacity for AI/cloud anchors (the Oracle/Red Oak pivot — see Lens 4/5).

Business model in plain terms. It is a digital-infrastructure landlord. It builds/acquires data-center real estate in tier-1 and, distinctively, tier-2 metros (Salt Lake City, Kansas City, Minneapolis, Pittsburgh, Cleveland, Nashville-adjacent, etc.), signs multi-year leases, and collects contractual recurring revenue (MRR) with escalators. The economics are a REIT-style spread business: cost of capital (ABS + construction debt + sponsor equity) vs. stabilized yield-on-cost of the assets. It is structured as a taxable operating platform, not a listed REIT, but the asset logic is identical.

Customers. Highly granular on the enterprise/colo book — 1,757 customers in the Series 2026-1 securitized pool, with the top customer at 3.9% of annualized MRR and the top 5 at 15.3%. That diversification is the historical selling point. But the new hyperscale layer inverts it: Oracle is the single anchor tenant for the first four of eight buildings at the 480MW Red Oak (Dallas) campus. So DataBank is becoming a barbell — a diversified 1,700-tenant enterprise annuity on one side, a concentrated hyperscale-lease build on the other.

Contract structure & payment terms. Enterprise colo = recurring MRR, multi-year, with renewal stickiness driven by interconnection and migration cost. Hyperscale = long-dated triple-net-style leases (the ABS pool shows leasehold expirations stretching 2032–2054). Weighted-average remaining term on the securitized pool is 48 months — respectable but not ultra-long, reflecting the enterprise mix; the hyperscale leases lengthen it going forward.

Ground-truth caveat. No positioning.md / customers.csv on the shelf — all of the above is ``. Company self-describes as "the largest edge data center platform in the U.S." Independent market reports (ResearchAndMarkets, MarketsandMarkets) consistently list DataBank among the top-tier US colo names alongside Equinix, Digital Realty, CyrusOne, Aligned, QTS, Vantage, STACK, Compass — but as a mid-market/tier-2 specialist, not a global-scale leader.

Lens 2 · Supply Chain

Map: upstream inputs → DataBank → end customer.

Upstream (what DataBank buys):

• Land & shells — greenfield sites (Red Oak 292 acres) and acquired facilities. Site selection gated by power availability + fiber density (the two real chokepoints in 2026).
• Power — the binding constraint. DataBank contracts utility power and (increasingly) negotiates directly with utilities/IPPs for multi-hundred-MW campus loads. Named exposure: the ~850MW build-out (Red Oak 480MW, Culpeper VA 192MW, Atlanta 120MW) depends on grid interconnection queues in ERCOT (Dallas), PJM/Dominion (Culpeper VA), and Georgia Power (Atlanta) — the same congested queues throttling every operator.
• Critical electrical & cooling gear — generators, switchgear, UPS, chillers/liquid-cooling. Supplier universe is the named datacenters-beat peer set: Vertiv (VRT), Eaton (ETN), Schneider, Caterpillar (CAT)/Cummins (CMI)/Generac (GNRC) for gensets, Trane (TT)/Johnson Controls (JCI)/Modine (MOD) for thermal. Switchgear and transformer lead times (18–24+ months) are a real gating item industry-wide.
• Fiber/connectivity — cross-connects and cloud on-ramps to AWS/Azure/GCP/Oracle Cloud; DataBank's roots trace partly to the zColo (ex-Zayo) fiber-colo assets acquired 2021, giving it carrier-hotel density.
• Construction & EPC — GCs and electrical contractors (the beat's Quanta (PWR), Comfort Systems (FIX), IES (IESC), Sterling (STRL) cohort captures this labor pool).
• Capital — arguably DataBank's most important "input." Its supply chain of money (ABS investors, construction lenders led by MUFG, sponsor equity from DigitalBridge/AustralianSuper/Swiss Life) is as load-bearing as its supply chain of steel and power.

Downstream (who buys from DataBank): 1,700+ enterprises (regulated verticals — healthcare, finance, government — are a DataBank strength via compliance certifications), carriers, SaaS/content providers, and now hyperscalers/AI (Oracle).

Chokepoints / single-source dependencies:

1. Power interconnection — #1 constraint; multi-year utility queues gate every new MW.
2. Anchor-tenant concentration on the hyperscale build — Oracle for 4 of 8 Red Oak buildings is a single-source revenue dependency on the growth engine (contrast the diversified legacy book).
3. Capital markets access — the model requires continuous ABS + construction debt issuance. A frozen securitization market (2022-style) would stall growth. (Mitigant: 2025 data-center ABS volume ~tripled to $30B+, demand > supply — the window is wide open now.)

Names or it didn't happen — delivered above.

Lens 3 · Competitive Advantages (moats)

Where the moat is real:

• Tier-2 metro density + interconnection. DataBank's genuine edge is being the incumbent interconnection point in ~25 markets, many of them tier-2 where Equinix/DLR are thin. Once an enterprise lands cross-connects and cloud on-ramps in a DataBank facility, switching cost is high (physical migration + re-cabling + re-establishing the ecosystem). This is the classic colo moat — weaker than Equinix's global IBX network effect, but real at the metro level.
• Regulated-vertical compliance stack. HIPAA/PCI/FedRAMP-adjacent certifications and managed-services wrap make DataBank sticky for healthcare/finance/gov workloads that can't easily lift-and-shift to a commodity wholesaler.
• Sponsor + capital-formation machine. Access to DigitalBridge's playbook, AustralianSuper/Swiss Life equity, and a proven, repeat, investment-grade ABS program (5 series, $3.23B, now dual-rated S&P/Moody's) is a structural cost-of-capital advantage over sub-scale private colo operators. This is a balance-sheet moat, not a product moat.

Where the moat is thin:

• Against hyperscale wholesale peers (Aligned, Vantage, CyrusOne, QTS, STACK, Compass), DataBank is a late, smaller entrant to the 100MW+ AI build game. Winning Oracle validates it, but it's competing on speed-to-power and capital scale against players who have been building GW-scale campuses for years.
• Bargaining power. Over enterprise customers: strong (fragmented, sticky, 1,757 of them). Over hyperscale customers (Oracle): weak — a single 480MW tenant dictates terms, and DataBank needs the anchor more than Oracle needs any one landlord. Over suppliers of power and gear: weak — everyone is queued behind the same utilities and switchgear OEMs.

Verdict on moat: a durable mid-market interconnection/compliance moat (defensible, cash-generative) grafted onto a contested, capital-intensive hyperscale land-grab (validated but not yet advantaged). The moat protects the base; it does not yet protect the growth story.

Lens 4 · Segments

No segments.csv on the shelf — DataBank does not publicly break out revenue by segment (private). What the ABS disclosure and press let us reconstruct ``:

Geography: US-only, ~25 metros. Concentration is moderate at the metro level — no single market >~20% of pool revenue — but Dallas is becoming the center of gravity via Red Oak (up to 480MW, 8 buildings, 3.4M sqft when complete ).

The segment story in one line: DataBank is mid-transformation from a diversified enterprise-colo cash machine (~330MW, 1,700 tenants) into an enterprise-colo-plus-hyperscale-AI platform (~1,200MW+ pipeline, anchored by Oracle) — tripling power capacity, and in doing so trading some of its prized diversification for hyperscale scale and lease duration. Whether that's accretive or dilutive to the moat is the central debate (Lens 12/13).

---

Phase B — Measure performance

No "earnings print" exists (private, no P&L on the shelf). Phase B is reconstructed from the +private overlay: Lens 5 → Funding & valuation trajectory, Lens 7 → Cap table & marks, plus a Traction & unit-economics read off the ABS pool. Lens 6 → founder/exec commentary (not calls). Lens 8 → funding/financing/product catalysts.

Lens 5 · Funding & Valuation Trajectory (+private swap)

Round / recap history (seed of the equity story):

Implied valuation (the one clean anchor). DBRG's 7.8% residual = $486M implies a ~$6.2B equity value for DataBank at the Jan-2025 mark ``. PitchBook/CBInsights headline "$12.4B total funding" is cumulative capital raised (equity + debt across the history), not equity value — do not read it as a valuation. A precise current enterprise value is n/a — not disclosed; the Series 2026-1 pool alone carries a $4.2B third-party appraised value on just 36 of 65+ assets, so the total asset base materially exceeds that.

Burn / capital-intensity signal. This is a capital-devouring compounder by design: ~$2B equity (2024) + $2.65B Red Oak construction debt (2026) + $3.23B+ ABS + $725M credit facility. Growth is funded, not organic — the "burn" question is really a spread + access question (Lens 12/13).

Lens 6 · Founder / Executive Commentary (sentiment) (+private: interviews, not calls)

No earnings calls. Signal from exec commentary and financing narrative:

• CFO Kevin Ooley has publicly framed data-center ABS as "an AI investment vehicle" — positioning DataBank's securitization program as the retail/institutional on-ramp to AI infrastructure cash flows. Tone = confident, capital-markets-forward, disciplined.
• 2026 house view (DataBank's own published forecasts): titles like "Tighter Capital, Constrained Power, and the Return to Fundamentals" and "Data Center Construction Predictions for 2026" signal management is explicitly flagging capital tightness and power scarcity as the binding 2026 constraints. That is a management-is-clear-eyed tell — they are not pretending the environment is frictionless.
• Succession messaging (see Lens 9) is framed as "entering the next phase of scale and expansion" — the standard growth-narrative wrapper, but the substance (a 10-year CEO stepping up to Exec Chairman, the CFO taking the wheel) reads as continuity, not upheaval.

Sentiment shift over time: narrative has moved from "largest edge/interconnection platform" (2021–23, the diversification story) → "AI-infrastructure + hyperscale capital machine" (2024–26, the Oracle/ABS/scale story). The pivot is deliberate and on-message.

Lens 7 · Cap Table & Secondary Marks / Comps (+private swap)

Cap table quality (the +private tell):

• Sponsor: DigitalBridge (~7.8% residual after the 2024–25 secondaries) — soon a SoftBank subsidiary. This is a strategic upgrade to the cap table's backing: SoftBank's "ASI"/Stargate-adjacent ambitions make DataBank an indirect piece of the largest AI-infra balance sheet in the world.
• Anchor institutional: AustralianSuper (~$1.5B, board seat) — a top-tier crossover/sovereign-scale pension, its first US data-center investment. A pension of that caliber taking a board seat is a governance-and-IPO-proximity signal (patient capital, but exit-motivated on a 5–7yr horizon).
• Other institutionals: Swiss Life Asset Managers, EDF Invest, IMCO — infrastructure specialists, long-duration.
• Read: this is a blue-chip, exit-capable syndicate (sovereign pension + insurance/infra funds + a mega-strategic sponsor). The presence of AustralianSuper + Swiss Life is exactly the "crossover-fund-entered → IPO-proximity" pattern the overlay looks for.

Comps — public digital-infra landlords (the tradeable proxies). DataBank has no listed equity; the honest comp is how the market values its listed peers, since an eventual DataBank IPO/sale would price off these. Multiples are `` where I could anchor them and n/a otherwise. I did not find clean, dated, current EV/EBITDA or P/AFFO prints for each peer within this run and will not fabricate them:

Best available valuation anchors (use these, not fabricated multiples):

• Implied DataBank equity ~$6.2B at Jan-2025 mark ``.
• Series 2026-1 pool: $4.2B appraised / $433M annualized revenue = ~9.7x revenue on the securitized real estate `` — a rich, AI-era digital-infra appraisal multiple.
• ABS LTVs: A2 at 48.4%, Class B at 50.1% — the rating agencies are lending to ~50% of appraised value, i.e. the pool supports ~2x asset coverage on the senior debt.
• CyrusOne's ~$15B take-private (2022) is the cleanest private-transaction comp for a US wholesale-colo platform.

Lens 8 · Financing & Product Catalysts (>5% "moves") (+private: no stock, so capital/asset events)

The events that would move a listed DataBank / that repriced the private mark:

• Oct 2024 — $2B equity raise (AustralianSuper). The single biggest re-rating: tripled the power pipeline, brought in a sovereign-scale anchor. Direction: strongly positive.
• Jan 2025 — DBRG secondary at +24% step-up, 2.2x MOIC. A third-party-validated markup — the private equivalent of a beat. Positive.
• Sep 2025 — $1.1B hyperscale ABS, dual-rated S&P A- / Moody's A3 (first Moody's DC ABS), first green hyperscale bond. Proved DataBank can term-fund hyperscale (not just enterprise) assets at IG cost. Positive — de-risks the growth funding.
• Dec 2025 — SoftBank to buy DigitalBridge (~$4B). Upgrades the ultimate parent to the world's most aggressive AI-infra allocator; plausibly accelerates capital into DataBank. Positive, but adds strategic-ownership complexity.
• Jan 2026 — Series 2026-1 $665M ABS (36 assets, $4.2B appraised, $433M rev). Routine-but-large; refreshed the market's appraisal of the base portfolio. Neutral-positive.
• Apr 2026 — $2B (→$2.65B) Red Oak construction loan, MUFG-led; Oracle leases DFW9/10/11 (180MW). The largest financing in company history and the hyperscale-execution proof point — capital and an anchor tenant, together. Positive on execution; raises concentration + leverage risk.

Pattern: the "tape" for a private is its funding cadence, and DataBank's cadence is relentless and successful — 5 ABS series, two mega-equity/recaps, a record construction loan, all inside ~24 months, every one oversubscribed or IG-rated. What the "market" reacts to for this name is capital access and anchor wins — and both are currently green.

---

Phase C — Judge people & books

Lens 9 · Management

• Raul Martynek — CEO since 2017 (~10-yr tenure). A 20+ year telecom/internet-infrastructure operator: CEO of Net Access (NJ data centers), CEO of Voxel.net (managed hosting, sold to Internap 2012), Chief Restructuring Officer of Smart Telecom (Ireland), Senior Advisor to DigitalBridge before taking the DataBank seat. Archetype: professional operator + sponsor insider — exactly the profile a PE-backed roll-up platform wants. His decade covers the zColo integration (doubled the footprint), the 2022–23 recaps, and the 2024 AustralianSuper raise — i.e. he executed the buy-and-build that got DataBank from a regional colo to a top-tier national platform.
• Succession (telegraphed, orderly). Effective May 19, 2026 Martynek becomes Executive Chairman; Kevin Ooley (current President & CFO) becomes CEO effective Jan 1, 2027. Ooley is the architect of the ABS/capital-markets strategy (the "ABS as an AI vehicle" thesis) — putting the balance-sheet engineer in the CEO chair at the moment the company's entire growth model is capital formation. That is a coherent, non-alarming hand-off (contrast a surprise exit). Continuity risk: low; it's a promotion-from-within with the founder staying as Chairman.
• Capital-allocation history. The record is strong and consistent: disciplined buy-and-build (zColo at an accretive moment), repeated ability to raise equity at step-ups (24% in 2025) and 2.2x MOIC for the sponsor, and a best-in-class IG ABS program that lowers cost of capital. No visible value-destruction. The open question is whether the hyperscale pivot (a lower-return, higher-capital, more-concentrated business than legacy colo) is the right allocation of the new $2B — bulls say scale + duration, bears say margin dilution (Lens 12/13).
• Skin in the game. As a private, management equity is not publicly quantified — n/a — private, not disclosed. Sponsor alignment is via DigitalBridge/carry and the institutional syndicate's board oversight (AustralianSuper board seat).
• Red flags: none material found. The one structural governance item is the DigitalBridge relationship (sponsor is also promoter, manager, and — via the zColo/Zayo deal — a related-party counterparty). See Lens 10.

Assessment: a high-quality, aligned, continuity-strong management team with a demonstrated capital-formation edge. This is one of the stronger operator/sponsor pairings in private digital infra.

Lens 10 · Forensic Red Flags

Standard caveat: no audited financials on the shelf (private). Forensic analysis is therefore structural, not line-item — I cannot test receivables-vs-revenue or SBC-vs-non-GAAP because there is no public P&L. What the structure itself flags ``:

1. Leverage is the model, and it is high. ~$3.23B+ securitized notes + $2.65B Red Oak construction debt + $725M credit facility + more construction lines, against a ~$6.2B implied equity value ``. This is an aggressively financed platform — investment-grade at the asset/tranche level (S&P A- / Moody's A3 on senior ABS), but the consolidated entity runs meaningful structural leverage. Refinancing risk is real: ABS Series 2026-1 has an anticipated repayment of Feb 2031 with legal maturity Feb 2056 — i.e. the debt must be refinanced at ARD or the coupon steps up; the model depends on perpetual capital-markets access.
2. Appraisal-based valuation. The $4.2B pool value and the LTVs rest on third-party appraisals of data-center real estate at AI-cycle peak cap rates. If cap rates normalize (rates stay higher / AI capex cools), appraised values — and thus LTV headroom and refi capacity — compress. This is the single biggest "accounting-adjacent" risk: the whole capital stack is anchored to appraised asset values, not GAAP earnings.
3. Related-party history (DigitalBridge). The zColo assets were acquired from Zayo, a DigitalBridge-controlled entity — a sponsor-to-portfolio-company (related-party) transaction. Common in PE roll-ups and not per se improper, but it is exactly the category (related-party asset transfers at sponsor-set prices) a forensic analyst flags. With DigitalBridge now heading into SoftBank, watch for further related-party dealings (SoftBank-affiliated tenants, Stargate-linked capacity, intercompany financing).
4. Anchor-tenant concentration on the growth engine. Oracle = 4 of 8 Red Oak buildings. If Oracle's AI-capacity plans slip, the marginal-return campus (the one the new equity funded) is directly exposed. The legacy book is diversified (top tenant 3.9%); the incremental book is not.
5. Green-bond / ESG labeling. DataBank's ABS are marketed as "green bonds." Not a red flag, but green-labeled DC debt is drawing scrutiny (real energy intensity vs. the label) — a reputational/regulatory watch item, not a finding.

Regulatory findings (required sub-section):

• SEC (EDGAR EFTS — LR + AAER): Zero findings. DataBank has no CIK (private, not an SEC registrant), so no EDGAR enforcement search is possible. (Note: DataBank's financing SPVs have filed Form ABS-15G as asset-backed issuers — those are securitization disclosures, not enforcement actions.)
• Non-SEC (FTC/DOJ/FDA/CFPB/etc.): ran the prescribed web search — "DataBank" (FTC OR DOJ OR... OR settlement OR fine OR penalty) enforcement. No material enforcement action, consent decree, fine, penalty, major outage, or data-breach class action attributable to DataBank Holdings / DataBank Ltd. found. (Search surfaced unrelated companies — NTT/RagingWire's 2020 Privacy Shield settlement, Blackbaud, Kochava — none of which are DataBank.)
• Item 3 Legal Proceedings (10-K): n/a — no 10-K exists (private).
• Conclusion: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER), FTC case index + web search, and confirmed there is no 10-K to review, as of 2026-07-06. Unaudited per public sources.

Net forensic read: the risks are structural (leverage + appraisal + related-party + anchor concentration), not evidentiary. There is no smoke on fraud, litigation, or enforcement. The thing to underwrite is the capital structure and the sponsor relationship, not the honesty of the numbers.

---

Phase D — Project & stress-test

Lens 11 · IPO-Readiness & Path-to-Tradeable (+private swap — the be-early payoff lens)

No EPS projection (private, no earnings). The overlay question is: when does DataBank become tradeable, and what unlocks an S-1?

Readiness score: 3.5 / 5 `` (Scale: 3 = late-stage · 4 = pre-IPO/secondary-active · 5 = S-1 filed / IPO imminent.)

Why 3.5:

• Toward tradeable (pull): blue-chip exit-capable syndicate (AustralianSuper board seat, Swiss Life, EDF, IMCO); active secondary market in the shares (the 2024–25 DBRG secondaries); IG capital-markets fluency (5 ABS series) means the disclosure/reporting machinery for a public listing is largely built; scale (65+ DCs, $4B+ appraised on a subset) is IPO-relevant; and a red-hot data-center-IPO/AI-infra window.
• Holding it private (drag): DataBank's sponsors have repeatedly chosen private recaps + ABS + secondaries over an IPO — the private capital-formation machine is working too well to force a listing; AustralianSuper/Swiss Life are patient infra capital (they don't need the liquidity soon); and the SoftBank–DigitalBridge deal (closing 2H 2026) injects a strategic-owner variable that could point the exit toward SoftBank/private continuation vehicles rather than a public IPO.

Path-to-tradeable — the realistic routes (ranked):

1. Continuation / private-continuation vehicle or further secondary (most likely near-term) — the sponsors keep compounding privately; liquidity via periodic secondaries. Tradeable for insiders, not for us.
2. IPO — plausible 2027–2029 if the AI-infra listing window stays open and AustralianSuper/Swiss Life want a mark-to-market exit. This is the be-early prize: DataBank would list as a scaled, IG-rated, AI-levered US digital-infra REIT/operator.
3. Strategic sale / absorption — SoftBank (via DigitalBridge) or another mega-allocator takes it whole into an AI-infra platform. The Dec-2025 SoftBank deal makes this more likely than a year ago, and would remove the public-IPO path.

Milestones that unlock an S-1: (a) SoftBank–DBRG deal closes and clarifies DataBank's role (2H 2026); (b) Red Oak/Oracle stabilizes, giving a clean hyperscale + enterprise blended-yield story to sell; (c) a decision by AustralianSuper/Swiss Life to seek liquidity; (d) a still-open AI-infra IPO window.

Estimated window: no S-1 imminent. Watch 2027–2029. Near-term liquidity is far more likely private (secondary/continuation) or strategic (SoftBank) than a public IPO.

Write-back: DataBank should be added to research/private-watch.json with beat: datacenters, stage: late-stage/pre-IPO-optional, ipo_readiness: 3 (I'd hold at 3, not 4 — the private-continuation bias is strong), lead_investors: "DigitalBridge (→SoftBank), AustralianSuper, Swiss Life, EDF Invest, IMCO", catalyst: "SoftBank–DBRG close 2H2026; Red Oak/Oracle stabilization; potential 2027-29 IPO", dossier: this file. (Per wave boundaries I am not editing that file in this run — flagging for the master.)

Lens 12 · Bull vs Bear

Bull case. DataBank is the best-managed mid-market US digital-infra platform, sitting on a defensible tier-2 interconnection/compliance moat (1,757 sticky enterprise tenants, top one only 3.9%) that throws off IG-rated, appraisable, term-financeable cash flow — a cash annuity growing at the pace of enterprise cloud + AI edge. On top of that annuity it is stapling a hyperscale AI growth engine (Red Oak 480MW, Oracle anchor) funded by a proven capital machine (5 ABS series, $2B sovereign equity, record MUFG construction loan) — tripling power capacity into the tightest supply environment in data-center history, where power + speed-to-market are the scarce goods and DataBank has both land and capital. The SoftBank–DigitalBridge tie-up upgrades its ultimate backer to the world's most aggressive AI-infra allocator, plausibly accelerating capital and tenant flow (SoftBank/Stargate-adjacent). Every financing in 24 months has been oversubscribed or IG-rated — the market is paying up to lend to and invest in this asset. Secular tailwind (AI + cloud), scarce inputs it controls, elite capital access, orderly succession into the balance-sheet architect.

Bear case (permanent-impairment risks).

1. It's a levered spread business at a cyclical peak in asset values. The entire capital stack rests on AI-cycle appraisals ($4.2B on 36 assets) and perpetual ABS refinancing (Series 2026-1 must refi by 2031 ARD). If rates stay high, cap rates normalize, or AI capex cools, appraised values and refi headroom compress simultaneously — the classic levered-real-estate impairment. This is not an earnings-miss risk; it's a balance-sheet risk.
2. Margin/return dilution from the hyperscale pivot. Wholesale hyperscale build-to-suit is lower-yielding and more capital-intensive than DataBank's high-margin interconnection colo. Pouring $2B of new equity + $2.65B of debt into a contested land-grab against bigger incumbents (Aligned/Vantage/QTS/CyrusOne/STACK) could dilute blended returns and concentrate the book on one anchor (Oracle). DataBank may be trading its best asset (diversified, sticky, high-margin colo) for a commoditized, capital-hungry one.
3. Never-tradeable-for-us risk. The sponsors keep compounding privately and may exit via SoftBank / continuation vehicle, never IPO-ing — the "investment" is a mark you can't buy.

Pre-mortem (18 months out, thesis broke — what happened?): AI-capacity demand from hyperscalers softened just as Red Oak's later buildings needed pre-leasing; Oracle deferred phases 5–8; higher-for-longer rates widened ABS spreads and forced a refi at punitive coupons, compressing equity returns; a data-center appraisal reset cut pool values ~15–20%, squeezing LTV headroom; the SoftBank deal introduced related-party/strategic complexity that spooked the co-investors. None of these is a fraud — all are cycle + leverage + concentration.

Are the marks too high? The ~9.7x revenue implied on the securitized pool `` and the 24% step-up are AI-peak digital-infra pricing. Justifiable if AI demand and low cap rates persist; rich if either reverts. On an IG-lending basis (~50% LTV), the debt is conservatively marked; the equity is priced for continued AI-infra euphoria.

Contrarian view (what the market is refusing to see): the consensus treats every AI-data-center operator as a pure growth call. The subtle truth on DataBank is that its durable value is the boring part — the 1,757-tenant, tier-2, compliance-grade interconnection annuity — and the exciting part (hyperscale/Oracle) is where it is least advantaged and most levered. The market may be over-paying for the growth graft and under-appreciating (or, on a bad-tape day, over-levering) the annuity. DataBank is best understood as a fixed-income-like infrastructure credit with an equity kicker, not a hyper-growth AI equity.

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

Dismantling the bull case:

• Where the money is concentrated & what shifts it: the growth revenue is concentrating on Oracle (4/8 Red Oak buildings). Oracle's own AI-capacity ambitions (OCI, Stargate) are aggressive and unproven at the announced scale — if Oracle re-trades or slows, the marginal, equity-funded campus is impaired, and DataBank has already spent the $2.65B to build it.
• Why the moat is weaker than bulls think: the durable moat (interconnection/compliance) is tier-2, mid-market — it does not extend to the hyperscale wholesale business, which is a speed-and-capital game DataBank enters late and sub-scale vs. Aligned/Vantage/QTS/CyrusOne/STACK/Compass, all with more GW-scale experience. Bulls conflate a real colo moat with a hyperscale advantage DataBank hasn't earned.
• The most dangerous competitor bulls underestimate: not Equinix — it's the well-capitalized wholesale pure-plays (Vantage, Aligned, QTS-under-Blackstone, CyrusOne-under-KKR/GIP) who can out-build and out-bid on power. On the enterprise side, hyperscalers' own edge/regional zones and cloud on-ramps slowly erode the "why colo" question for some workloads.
• Worst capital-allocation / governance items: the related-party zColo-from-Zayo (DigitalBridge) transaction; a capital structure that requires perpetual ABS access (a strategy that is also a dependency); and now a sponsor being absorbed by SoftBank — an owner famous for aggressive, sometimes value-destructive, cross-portfolio bets (WeWork, Vision Fund vintage). SoftBank ownership is a double-edged catalyst: more capital, but more strategic risk and related-party entanglement.
• What must hold for today's mark: (1) AI/cloud demand stays strong enough to pre-lease Red Oak phases 5–8; (2) ABS spreads stay tight and the 2031 refis clear at reasonable cost; (3) data-center cap rates don't normalize; (4) the SoftBank deal doesn't disrupt DataBank's governance/capital plan.
• Valuation if growth disappoints 20–30%: the equity (levered, ~50% LTV on peak appraisals) is where the pain lands. A 20–30% cut to appraised values + a spread-widening refi could halve or worse the equity value while leaving the senior ABS money-good — i.e. the debt survives, the equity gets crushed, the classic levered-infra outcome.
• Single scenario that permanently impairs: a sustained AI-capex air-pocket coinciding with a higher-for-longer rate regime — appraisals reset, refis get punitive, hyperscale pre-leasing stalls, and the $2B of 2024 growth equity is trapped in half-let, low-return campuses. Plausibility: moderate (not base case in mid-2026's demand environment, but the tail that matters).

Short-seller's one-liner: "A superbly financed colo annuity is being re-cast as a levered, late, single-anchor hyperscale bet at peak cap rates, under an owner about to become SoftBank — I don't doubt the numbers, I doubt the return on the incremental dollar."

Lens 14 · Management Questions (ordered by information value)

1. Return on incremental capital: what stabilized yield-on-cost do you underwrite for the Red Oak hyperscale buildings vs. your legacy interconnection colo — and is the hyperscale pivot accretive or dilutive to blended unlevered returns?
2. Oracle concentration: what portion of contracted forward (2026–2029) revenue is Oracle, what are the take-or-pay / termination / phase-commitment terms on Red Oak buildings 5–8, and what happens to campus economics if Oracle defers later phases?
3. Refinancing: across your $3.23B+ ABS and construction debt, what is the maturity/ARD wall by year, and how do you stress the 2031 Series 2026-1 refi under a "spreads +200bps, cap rates +150bps" scenario?
4. Appraisal sensitivity: how much LTV headroom do you have if third-party appraised values reset −20%, and at what point does that constrain new issuance?
5. SoftBank/DigitalBridge: post-close (2H 2026), how does SoftBank ownership change your capital access, tenant pipeline (Stargate-adjacent?), and related-party governance — and what guardrails protect co-investors (AustralianSuper, Swiss Life)?
6. Exit intent: what is the base-case liquidity path and timeline for AustralianSuper/Swiss Life — IPO, continuation vehicle, or strategic sale — and what milestones gate an S-1?
7. Power: how much of the ~1,200MW pipeline has secured, energized grid interconnection vs. still in utility queue, and what are the in-service dates by campus?
8. Hyperscale competitive position: against Vantage/Aligned/QTS/CyrusOne/STACK, what is your durable edge in the wholesale AI build market — or is this a scale-and-speed business where you're a price-taker?
9. Blended lease duration: as hyperscale grows, where does WA remaining lease term (48 months today on the ABS pool) go, and how does that change the credit/refi profile?
10. Margin trajectory: as the mix shifts toward lower-margin wholesale, what happens to platform EBITDA margin over 2026–2028?
11. Enterprise-book durability: what is churn / net revenue retention on the 1,757-tenant enterprise base, and is interconnection revenue still growing as a share (the moat) or shrinking?
12. Green-bond substance: what PUE / renewable-sourcing metrics underpin the green-bond label, and what is your exposure to tightening EU/US energy-disclosure regimes?
13. Capital-allocation guardrails: what leverage ceiling (consolidated, through-the-cycle) will you not cross, even to win an anchor?
14. Succession: with the CFO becoming CEO (Jan 2027), does capital-markets strategy become more central to the platform — and who owns tenant/operations as Ooley moves up?
15. Downside plan: if the securitization market froze for 12 months (2022 redux), what is the self-funding runway and which pipeline projects get paused first?

---