STACK Infrastructure

Phase A — Understand the business

Lens 1 · Company Overview

What it is. STACK Infrastructure is a global developer, owner and operator of wholesale and hyperscale data centers — Denver-headquartered, launched in 2019 by rolling up several T5 Data Centers assets and three Infomart facilities into a single national platform under one brand. It is not a public company and never has been: it is the flagship operating platform of IPI Partners, a digital-infrastructure fund manager originally a JV of ICONIQ Capital and Iron Point Partners.

The business model in plain terms. STACK builds big buildings full of power, cooling and floor space, then leases that capacity to two customer types: (1) hyperscalers (cloud/AI platforms) under long-dated, single-tenant, build-to-suit leases on massive campuses; and (2) enterprises under multi-tenant wholesale colocation. The economic engine is: secure land + grid power → sign a creditworthy 10–15-year lease (the "offtake") → finance construction against that lease (green bonds / ABS / project facilities) → collect contracted rent. The lease is the collateral. Revenue is contracted and recurring; the scarce input is not capital but energized, grid-connected land.

Key products/services. (a) Hyperscale build-to-suit campuses (single-tenant, ≥100 MW, sometimes >1 GW); (b) wholesale colocation (multi-tenant); (c) "powered shell" and turnkey delivery options; (d) AI-optimized high-density designs with liquid/closed-loop cooling for accelerated compute.

Main customers. The one named anchor is Amazon Web Services: STACK is developing the $10B "Resilient Technology Park" campus in Shreveport / Caddo–Bossier Parishes, Louisiana for AWS — seven two-story buildings, first facility 120 MW with a second 100 MW to follow, approved by Shreveport City Council Dec 2025, first facilities live late-2027/early-2028, and AWS is covering 100% of the utility-infrastructure cost. Beyond AWS, tenants are described only generically by the parent (IPI): "primarily investment-grade corporations, including several Fortune 100 companies" — no other hyperscaler is publicly named to a specific STACK lease.

Main suppliers / partners. Utilities (for power — the binding constraint), general contractors, and in APAC a JV partner ESR (72 MW Osaka development). Capital "suppliers" are its lenders and the ABS/green-bond market.

Main competitors. Other wholesale/hyperscale developers: Vantage Data Centers, Aligned Data Centers, CloudHQ, QTS (Blackstone), CyrusOne (KKR/GIP), EdgeConneX, Switch, NTT, Compass, Sabey — plus the public REITs Digital Realty (DLR) and Equinix (EQIX) at the colocation end.

Contract structure / payment terms. Long-dated (typically 10–15-yr) single-tenant hyperscale leases; increasingly the tenant bears power/PUE risk above a 1.3–1.4 PUE cap, and offtake contracts must now be "financeable" (balanced termination rights) because the developer, not the hyperscaler, holds the scarce resource. Concentration is real: a single AWS-scale campus is a single counterparty on a single asset — investment-grade, but concentrated.

What STACK actually is (the "analyze it for what it is" instruction): not a stock, not an operating company with a P&L you can model, and not an imminent IPO. It is a capital-intensive, contracted-cashflow infrastructure platform wholly inside private fund vehicles, currently being actively carved up and monetized by its owner. The investable question is never "buy STACK" — it is "what does STACK tell us about (a) Blue Owl/OWL, the listed manager that now controls it, and (b) the private-market clearing price of energized megawatts."

Lens 2 · Supply Chain

Map: grid power + land → construction → STACK campus → hyperscaler/enterprise tenant → end AI/cloud workload. Named stakeholders along the chain:

• Upstream — power (the chokepoint): regional utilities. In STACK's largest market, Northern Virginia, that is Dominion Energy (and NOVEC substations, e.g. the NVA05 campus sits next to high-voltage lines fed by a NOVEC substation). Power availability — not capital, not land — is the single-source dependency for the entire model.
• Upstream — land: STACK controls large land banks in Loudoun/Prince William/Stafford (VA), Phoenix (AZ), Portland (OR), Toronto (ON), Doña Ana County (NM), Shreveport (LA), plus EMEA/APAC metros.
• Upstream — equipment: the AI-relevant bottleneck items are generators, switchgear, transformers, and liquid-cooling gear (industry-wide 12–24-month lead times in 2025–26; not STACK-specific but STACK-binding).
• The company: STACK does design, development, construction management and facility operations in-house across three regional units — STACK Americas (the core retained business), STACK EMEA (hyperscale-only after the colocation carve-out), STACK APAC.
• Downstream — tenants: AWS (named, Shreveport); other unnamed investment-grade/Fortune-100 hyperscalers and enterprises.
• End demand: AI training/inference and cloud workloads — the demand signal driving 2025–26 record-low <2% vacancy.
• Capital "supply chain": the ABS master trust (Stack Infrastructure Issuer LLC, now under Blue Owl Digital Infrastructure Trust), green-bond investors, and Blue Owl's own $39B digital-infra strategy.

Chokepoint verdict: the chain has exactly one true single-source dependency — grid interconnection/transmission — and it is currently binding in STACK's most concentrated market (see Lens 10/13). Everything else (equipment, capital, land) is tight-but-multi-sourced.

Lens 3 · Competitive Advantages (moats)

STACK's moat is not brand or IP — it is the pre-assembled bundle of {energized land + a track record hyperscalers underwrite + a proven, cheap financing machine}. Durable elements:

1. Land + power bank secured ahead of demand. STACK controls multi-GW of grid-adjacent land in the exact markets that are now supply-constrained (largest private developer in Virginia; multi-GW pipeline in Phoenix, Stafford, NM, LA). In a market where power is the scarce good, holding energized/near-energized land is the moat. This is the "time-to-power" advantage the Aligned deal just repriced (Lens 7).
2. A financing flywheel that peers can't cheaply replicate. STACK has raised ~$20B in debt capital for its global portfolio and runs a serial ABS master-trust rated A- by S&P that priced (per its own March-2024 disclosure) at the lowest blended spread over Treasuries of any data-center issuer since Oct 2021. Cheaper capital → ability to underwrite bigger single-tenant builds → wins the marquee leases → feeds the ABS. That loop is a genuine cost-of-capital moat, now supercharged by sitting inside Blue Owl's $39B platform.
3. Hyperscaler switching costs / underwriting trust. Once a hyperscaler co-designs a >100 MW build-to-suit campus (as AWS has at Shreveport), the relationship is sticky — the tenant has bespoke power, cooling and expansion rights on that site. Repeat business is the norm in this industry.
4. Bargaining power has shifted to STACK. Per 2026 lease-term analysis, "the era in which hyperscalers could unilaterally dictate terms is largely over" — developers who control grid-connected, shovel-ready land now negotiate from strength. That is a structural improvement in STACK's position vs. 2020–22.

Moat limits (honest): the product is fundamentally a commodity — a megawatt of power in a compliant building. There is no network effect and no IP wall; every large PE/infra platform (Blackstone/QTS, KKR/CyrusOne, DigitalBridge/Silver Lake/Vantage, BlackRock-MGX/Aligned) is chasing the same land and the same tenants with the same or deeper capital. STACK's edge is execution + cost-of-capital + a land head-start, not a defensible technology.

Lens 4 · Segments

No audited segment financials exist (private). Segmentation is operational/geographic, and the story since 2025 is one of active portfolio surgery — the segment mix is being sold down, which is itself the signal:

• STACK Americas (core, retained): the growth engine — Virginia (largest private developer in-state), Phoenix (~230–250 MW downtown campus), Stafford (1+ GW, 19 buildings across 4 sub-campuses), Portland, Toronto, Doña Ana County NM (>1 GW, on-site power + closed-loop cooling), Shreveport LA (AWS, $10B).
• STACK EMEA (halved): the pan-European colocation business (Stockholm, Oslo, Copenhagen, Milan, Geneva) was sold to Apollo Global Management (announced Apr 2025, reported May 2025), leaving STACK EMEA "focused primarily on hyperscale customers" per EMEA CEO John Eland. Origins: DigiPlex (Nordics) + SUPERNAP Italia (Milan) unified as STACK EMEA in Mar 2022; Safe Host (Switzerland, 56 MW / ~270k sq ft) acquired 2022.
• STACK APAC (on the block): Australia (Melbourne MEL01 180 MW; Sydney Erskine Park ~450 MW ~US$405M initial; Canberra 28 MW), Japan (Tokyo/Inzai 36 MW flagship + Osaka 72 MW ESR JV), South Korea, Malaysia. Blue Owl is exploring a partial or full sale of the entire APAC portfolio at $30B+ (preliminary, early-stage).

Trend + cause: the shape is shifting from "global diversified colocation + hyperscale" toward "US-centric hyperscale core, with EMEA-colo and APAC monetized to crystallize AI-era valuations." Cause: (i) the AI demand surge has made these platforms "trophy" assets commanding premium marks, and (ii) Blue Owl, as a fund manager, is doing its job — harvesting value for LPs by selling into a hot bid rather than holding forever. Accelerating dispositions, not accelerating consolidation.

---

Phase B — Measure performance

Private-overlay note: Phase B swaps the earnings/comps/catalyst lenses for the funding-trajectory / cap-table / traction lenses per the +private battery. There is no earnings print to analyze.

Lens 5 → Funding & Valuation Trajectory (replaces Earnings Result)

STACK finances at the project and platform level, not via equity rounds, so the "funding trajectory" is a debt-issuance and ownership-transaction history. All ``, unaudited:

Ownership trajectory (the equity story):

• 2019 — formed by IPI Partners (ICONIQ + Iron Point) from T5 + Infomart assets.
• 2022 — EMEA build-out: DigiPlex + SUPERNAP Italia → STACK EMEA (Mar); Safe Host (Switzerland) acquired (May).
• Oct 2024 → closed Jan 6, 2025 — Blue Owl Capital acquires IPI Partners' fund-management business from ICONIQ/Iron Point for ~$1.0B (~80% Blue Owl equity / ~20% cash); IPI = ~$10.5B AUM, 82 data centers, >2.2 GW leased as of Jun 30 2024. ICONIQ stays on for IR/investment-analysis services (payable 2026–28).
• Apr–May 2025 — Apollo buys STACK's pan-European colocation business (value undisclosed).
• 2026 (in process) — Blue Owl exploring $30B+ sale of STACK APAC (Australia/Japan/Malaysia; partial or full; early-stage, may be abandoned).

Debt-financing trajectory (the capital-raise story):

• ~$20B total debt capital raised for the global portfolio to date.
• 2025 alone: $6B+ green financing — $900M (Jan, NVA05 200 MW, Manassas VA) + $4B (Mar, 1+ GW Stafford VA + Portland OR + Toronto ON) + $1.4B (May, pool of 10 stabilized revenue-generating DCs).
• ABS master trust (Stack Infrastructure Issuer LLC → Blue Owl Digital Infrastructure Trust): serial A- (S&P) issuance — cumulative ~$2.83B by Mar 2024 (up from $2.30B Aug-2023, $2.59B early-2024), and a fresh Series 2026-1 of ~$850M now running through Blue Owl's Trust vehicle.

Valuation marks (the only ones sourceable):

• IPI management co. bought at ~$1.0B for the manager, on ~$10.5B AUM. That is a manager multiple, not an enterprise value of the assets.
• The APAC-only portfolio is being shopped at $30B+ — a partial-portfolio figure that implies the global STACK enterprise value is a multiple of that (see Lens 7 for the per-MW cross-check).
• Burn signals: n/a in the venture sense — this is contracted-cashflow infra funded by non-recourse-style project debt, not equity burn. The relevant solvency question is debt-service coverage on contracted leases, which the A- ABS rating and record-tight <2% market vacancy currently support.

Lens (added) → Traction & Unit Economics (Phase B, +private)

• Scale / capacity (reconciled across sources, dates flagged — figures are not consistent and should be read as a range):
• IPI parent portfolio: 82 DCs / >2.2 GW leased (Jun 30 2024).
• STACK-branded pipeline (2025 vintage): ~2.5 GW built or under development + ~4 GW further pipeline. (An earlier, narrower framing cited "400 MW under development + 1,000 MW potential" — clearly superseded by the 2025 figures; noted as a conflict, the 2025 numbers are used.)
• Virginia alone: ~1 GW built/under development + ~2 GW planned across Loudoun/Prince William/Stafford.
• Unit economics signal: the ~$8M-per-MW clearing price set by the Aligned/BlackRock-MGX deal is the best available proxy for what STACK's energized megawatts are worth (Lens 7). STACK's own A- ABS at record-tight spreads implies rating agencies view the contracted cashflows as high-quality.
• Demand backdrop: US data-center vacancy ~1.2% Q1 2026, net absorption 5.29 GW Q1 2026, ~92% of under-construction capacity pre-committed — i.e. STACK builds into a market that is absorbing supply as fast as it delivers. Unaudited, but the tape is unambiguous.

Lens 6 → Founder / Management Communications (replaces Earnings Calls)

No earnings calls (private). The "communications" signal comes from executive appointments and platform-owner messaging:

• Dec 2025 — Matt VanderZanden named CEO, STACK Americas (was President 2025, COO before) — a promotion that consolidates the retained US-hyperscale core under a domain-native operator (see Lens 9).
• Brian Cox — interim CEO, STACK EMEA (and STACK Americas board); Preet Gona — leads STACK APAC; John Eland — EMEA CEO who framed the Apollo carve-out as a deliberate "consolidation of commitment to hyperscale".
• Blue Owl's messaging positions STACK explicitly as an "execution layer" / operating partner inside a diversified AI-infra capital platform — not a standalone brand headed for IPO. Consistent theme across 2025–26: focus the core on US hyperscale, monetize the periphery, present the whole as infrastructure-for-AI. Tone is expansionary and disciplined; the "things they stopped saying" is any pretense of being a diversified global colocation player.

Lens 7 → Cap Table & Secondary Marks / Mechanism Comps (replaces Comps)

Cap-table quality (the +private tell): STACK sits inside Blue Owl (NYSE: OWL) fund vehicles — specifically Blue Owl Digital Infrastructure funds (Fund III ~$7B; total digital-infra strategy $39B raised, 100+ facilities, ~30 markets mid-2025), with ICONIQ retained for services and sovereign capital (e.g. QIA) partnering the strategy. This is the highest-quality syndicate tier — a listed alternatives manager + sovereign wealth — which is exactly the "IPO-proximity / permanent-capital" signal the +private overlay looks for, but here it points the other way: with permanent institutional capital and a live $30B+ private bid for APAC, the LPs have no need to IPO STACK — they can monetize asset-by-asset at premium private marks.

Comps — this asset class is now priced on "time-to-power" ($ per MW), not P/E. The Aligned deal explicitly reset the market multiple. Provenance-critical table (all ``; per-MW figures are deal-implied):

Read: energized-megawatt valuations roughly doubled (from ~$3–4M/MW in 2021–24 to ~$8M/MW in 2025) because control over near-term power became the scarce good. STACK — largest private developer in Virginia, multi-GW grid-adjacent bank, AWS anchor — is squarely in the Aligned-comp bucket. A public P/E multiple for STACK is n/a and is the wrong lens; the right lens is $/MW, and it says the private bid for these assets is at cycle-high marks.

Lens 8 → Funding / Product Events that Moved Value (replaces Stock-Price Catalysts)

No stock, so the ">5% move" analog is the events that repriced the platform, in order of value-impact:

1. Oct 2024 / Jan 2025 — Blue Owl acquires IPI (~$1.0B): moved STACK from a niche PE-JV asset into a listed-manager's flagship AI-infra platform.
2. Oct 2025 — Aligned sold for $40B / ~$8M-MW: not STACK's deal, but it repriced the entire comp set STACK sits in — the single most important external mark for STACK's implied value.
3. Dec 2025 — AWS Shreveport ($10B) approved: converted "unnamed Fortune-100 tenants" into a named, marquee hyperscaler anchor on a >$10B campus — a credibility and pipeline event.
4. Mar 2025 — $4B single green-financing raise: demonstrated the financing flywheel at unprecedented scale; "largest private developer in Virginia."
5. Apr–May 2025 — Apollo buys EMEA colocation: crystallized value on the periphery and signaled the hyperscale-focus pivot.
6. 2026 — $30B+ APAC sale process reported: the live catalyst; if it prints, it sets a hard private mark on a third of the platform.

Pattern: value for this name moves on (a) M&A comps that reset the per-MW clearing price, (b) named hyperscaler anchor wins, and (c) financing scale — not on quarterly operations. The market "reacts to" power-secured scale and marquee counterparties.

---

Phase C — Judge people & books

Lens 9 · Management

• Matt VanderZanden — CEO, STACK Americas (Dec 2025). The standout. Path: real-estate lawyer (Paul Hastings) → Meta/Facebook Director of Global Data Center Site Selection (built and ran Meta's global DC location strategy) → STACK 2019 (Chief Strategy Officer → COO → President 2025 → CEO Americas Dec 2025). ~15 years in digital infrastructure; domain-native, tenant-side pedigree — he sat on the buyer's side (Meta) of exactly the site-selection and power-procurement decisions STACK now sells into. That is close to the ideal résumé for a US-hyperscale developer CEO. Tenure at STACK: 6 years — continuity through the whole IPI→Blue Owl transition.
• Regional bench: Brian Cox (interim CEO EMEA), Preet Gona (CEO APAC), John Eland (EMEA CEO), plus Americas C-suite (CFO Brad Berkley, CCO Rada Flom, COO Mike Casey, CLO Tim Kuester).
• Track record — quantified: raised ~$20B debt capital; built a ~2.5 GW built/under-development fleet with ~4 GW pipeline; won the AWS Shreveport $10B anchor; ran a serial A- ABS at sector-tightest spreads. This is a demonstrably capable development-and-finance machine.
• Skin in the game / insider ownership: n/a — private, not disclosed. Economics ultimately accrue to Blue Owl's funds and LPs (incl. ICONIQ, sovereign capital), not to a founder cap table. Management is professional, not founder-owner — appropriate for a late-stage, institutionally-owned infra platform.
• Capital-allocation history: disciplined and pro-cyclically sensible — build into a supply-constrained market, finance cheaply against contracted leases, and sell peripheral assets (EMEA colo, potentially APAC) into a hot bid to crystallize value. No evidence of value destruction; the pattern is textbook infra-platform harvesting.
• Red flags (management): none specific surfaced. The generic caution is that the true principal is a fund manager optimizing for LP exit timing, so "management's" incentives are to maximize disposition value, not to build an enduring independent company — fine for the assets, but it means there is no "STACK equity" for an outside investor to align with.
• Archetype: professional-manager / institutional-platform (not founder-led). Correct for the stage.

Lens 10 · Forensic Red Flags

No audited financials exist, so classic forensic-accounting analysis (revenue recognition, receivables vs. revenue, SBC flattering non-GAAP, goodwill) is n/a — private, unaudited per public sources. The forensic lens re-points, per the +private overlay, to structural/financing risk:

• Leverage & structure. STACK is heavily debt-financed (~$20B raised) via green bonds, project facilities and a serial ABS master trust. The mitigant is that this debt is secured against long-dated, investment-grade, contracted leases and rated A- by S&P — i.e. the rating agency has done third-party diligence on the cashflow quality. Still, a rising-rate or spread-widening shock raises refinancing cost on a capital-intensive book; the Series 2026-1 $850M issuance shows the machine still runs, but ABS markets can gate quickly in stress.
• Off-balance-sheet / structured opacity. The ABS master-trust and multiple SPV project facilities mean the consolidated economics are not publicly visible — a legitimate transparency limitation, not evidence of wrongdoing. An outside analyst cannot see leverage-per-asset or true portfolio DSCR.
• Concentration. A single AWS-scale campus concentrates counterparty and single-asset risk; and STACK's outsized Virginia weighting concentrates the portfolio in one grid region facing transmission constraints (Lens 13).
• Related parties. ICONIQ's retained services agreement with Blue Owl (IR/investment analysis, payable 2026–28) is a disclosed related-party arrangement tied to the acquisition — normal for such deals, but worth noting.

Regulatory findings (required sub-section). Per regulatory/regulatory-findings.md (generated 2026-07-07 by fetch-regulatory-findings.ts):

• SEC (EDGAR EFTS — LR + AAER): total_sec_findings = 0. STACK has no CIK and is not required to file with the SEC; no EDGAR enforcement search is possible.
• Non-SEC (FTC/DOJ/FDA/CFPB/consent-decree/settlement/fine/penalty) web search: no material enforcement action against STACK Infrastructure surfaced. The only adjacent regulatory friction is utility/permitting (Virginia data-center power-load reviews, Dominion interconnection policy, local siting/zoning) — sector-wide, not company-specific enforcement.
• 10-K Item 3 (Legal Proceedings): n/a — no 10-K exists (private).
• Conclusion: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER, zero results due to no-CIK) and non-SEC web search as of 2026-07-07. All findings are web-derived and unaudited per public sources.

---

Phase D — Project & stress-test

Lens 11 → IPO-Readiness & Path-to-Tradeable (replaces Forward Projection)

Per the +private overlay, this replaces the EPS projection. No forecast.ts EPS line is logged (there is no EPS; and unattended-loop rules skip the forecast create step).

IPO readiness: LOW — and, more importantly, IPO is not the likely exit. STACK is a wholly-owned platform inside Blue Owl's permanent-capital funds, being monetized asset-by-asset in the private market at cycle-high marks, not groomed for a public listing. The evidence:

• EMEA colocation → sold to Apollo (2025). APAC → $30B+ private sale process (2026). These are trade sales / fund exits, the opposite of an IPO path.
• The tradeable expression already exists: Blue Owl Capital (NYSE: OWL) is the listed manager that captures fees and carry on the STACK platform; and Blue Owl Digital Infrastructure Trust is a (non-traded) vehicle now housing STACK's ABS. An investor wanting STACK exposure buys OWL, or the public data-center REITs (DLR/EQIX) / listed neoclouds as thematic proxies — not a future "STACK IPO."

Estimated path-to-tradeable window / milestones:

• Base case — no STACK IPO; monetization continues. Milestones that would unlock further value crystallization (not an S-1): (a) close the APAC sale (would set a hard $30B+ mark); (b) stabilize + season the US hyperscale core (AWS Shreveport and Stafford reaching operation 2027–28); (c) potentially recycle the US core into a REIT or a continuation vehicle at a later cycle peak. ``
• Low-probability IPO scenario: only plausible if Blue Owl chose to float the US hyperscale core as a standalone REIT after stabilization — a 2027+ event at the earliest, and there is no public evidence it is planned. ``

Recommended research/private-watch.json entry (proposed, not written — unattended wave boundary):

"stack-infrastructure": {
 "beat": "datacenters",
 "stage": "owned-platform",
 "ipo_readiness": 1,
 "lead_investors": "Blue Owl Capital (owns via IPI), ICONIQ (services), sovereign LPs (e.g. QIA)",
 "catalyst": "not IPO-bound; value crystallized via asset sales — EMEA→Apollo (2025), APAC $30B+ process (2026); tradeable proxy = OWL",
 "dossier": "companies/stack-infrastructure/deep-dive-2026-07-07.md"
}

(ipo_readiness: 1 on the 1–5 scale = early/not-IPO-track — because the exit is trade sale, not listing. Connor to add via privates.ts if desired.)

Lens 12 · Bull vs Bear

Bull case. STACK owns the scarcest thing in the AI buildout — energized, grid-adjacent land in the exact constrained markets (Virginia foremost), with a named AWS anchor, a ~$20B / A- financing flywheel, and a domain-native, ex-Meta operator running the core. The Aligned deal just repriced energized megawatts to ~$8M/MW (roughly 2x the 2021–24 level), and STACK sits squarely in that comp bucket — so the assets are marked at cycle highs, with a live $30B+ private bid validating APAC alone. Demand is structural: <2% vacancy, 5+ GW/quarter net absorption, ~92% pre-committed. Inside Blue Owl's $39B permanent-capital platform, STACK can keep out-financing rivals. The bull expression is long OWL (fee + carry on rising private marks) and/or long the power/cooling picks-and-shovels the whole cohort must buy.

Bear case (2–3 permanent-impairment risks).

1. Power-timing failure. The moat is interconnection; if Dominion's Eastern-Loudoun transmission pinch (deliveries delayed "perhaps years," 15-yr large-load queue) hits STACK's Virginia-heavy pipeline, campuses sit built-but-dark — capex sunk, rent deferred, DSCR stressed. This is the most plausible thesis-breaker.
2. Comp-driven repricing / oversupply in the back half. The ~$8M/MW mark and the $30B+ APAC ask are cycle-peak, AI-demand-dependent valuations. Analysts flag a potential power/cooling oversupply in H2-2026 into 2027 if enterprise AI adoption lags; Moody's has modeled a 40% AI-valuation drawdown contagion. A private mark can compress fast when the marginal buyer steps back.
3. Rate / spread shock on a levered book. ~$20B of debt against a capital-hungry pipeline; a spread-widening or rate shock raises refi cost and can gate ABS issuance — squeezing the financing flywheel that is the moat.

Pre-mortem (18 months out, thesis broke): it's early 2028 — an AI-capex air-pocket softened hyperscaler leasing just as STACK's Stafford/Shreveport capex peaked; Virginia transmission slipped again; the APAC sale was pulled for lack of price; ABS spreads widened; and Blue Owl marked the STACK platform down, dragging OWL. The assets were fine long-term but the timing mismatch between sunk capex and delayed power/rent broke the near-term numbers.

Are multiples too high? For the assets, arguably yes on an absolute basis (~$8M/MW is a doubling in ~18 months) — but they are demand-justified while vacancy stays sub-2%. The risk is not the level, it's the fragility of the level if demand blinks. There is no STACK equity multiple to call rich or cheap; the tradeable multiple to watch is OWL's.

Contrarian view (what the market refuses to see): the crowd treats "STACK" as a way to play AI data centers. It isn't investable as such — it's a fund asset being sold, not bought. The real, tradeable insight is that STACK's disposals (EMEA→Apollo, APAC→$30B+) are Blue Owl converting the AI-infra boom into realized carry — so the cleanest expression of "STACK is winning" is long the manager (OWL), and the cleanest risk signal is watching whether that APAC sale actually clears at $30B+ or gets quietly abandoned.

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

Dismantling the bull case:

• The product is a commodity and the moat is a head-start, not a wall. A megawatt in a compliant building is fungible. Every deep-pocketed rival — Blackstone/QTS, KKR/CyrusOne, DigitalBridge-Silver Lake/Vantage ($9.2B equity), BlackRock-MGX/Aligned ($40B) — is buying the same land, the same power queue, the same tenants with equal or deeper capital. STACK's cost-of-capital edge narrows the moment a bigger balance sheet decides to win a lease.
• Revenue concentration is real and worsening as it "focuses." By shedding diversified EMEA colo and (maybe) APAC, STACK becomes more concentrated in US hyperscale + Virginia + a handful of mega-tenants (AWS named). Lose or delay one anchor and a multi-hundred-MW asset underperforms.
• The moat inverts into the top risk: power timing. Bulls call grid-adjacent land a moat; the short calls it a bet on utilities STACK doesn't control. Dominion has admitted it can't meet Virginia demand on time; the large-load queue runs up to 15 years. Built-but-unpowered capacity is a balance-sheet bomb, not an asset.
• The valuation is a momentum artifact. ~$8M/MW and a $30B+ APAC ask exist only because AI capex is at a 70%+ YoY sprint. That is the definition of a priced-for-perfection, reflexive mark: cheap ABS → aggressive builds → premium disposal marks → cheaper ABS. Reverse any link (a capex pause, a spread shock, a failed APAC auction) and the whole flywheel runs backward.
• Worst capital-allocation angle: none egregious, but the principal is a fund manager maximizing exit price — there is a structural incentive to sell the best assets at the top and leave the harder-to-power pipeline in the vehicle. An outside OWL holder inherits the residual, not the trophies that were sold.
• What must hold for today's marks: (1) hyperscaler AI capex stays at/near current levels through 2027; (2) Virginia/US transmission actually delivers on schedule; (3) ABS/green-bond spreads stay tight; (4) the private bid for megawatts stays ~$8M/MW. If growth disappoints 20–30%, premium per-MW marks compress toward the 2021–24 $3–4M/MW band — a potential ~50% haircut on implied asset value, and a markdown that would hit OWL. ``
• Single scenario that permanently impairs: a sustained AI-capex retrenchment (enterprise adoption stalls, hyperscalers pause) coinciding with STACK's peak-capex build years (2026–28) — stranding Stafford/Shreveport/NM capacity as demand and financing both tighten. Plausibility: moderate — the base case is continued growth, but this is the tail that actually kills the numbers.

Lens 14 · Management Questions (ordered by information value)

1. Is Blue Owl's intent to IPO the US hyperscale core, hold it as permanent capital, or sell it too — and on what trigger? (Determines whether any tradeable STACK equity ever exists.)
2. What share of the ~2.5 GW built/under-development pipeline has a firm, energized power delivery date vs. a queue position — and how much is exposed to the Dominion/Eastern-Loudoun transmission delays?
3. What is the actual tenant concentration — top-3 tenants as a % of contracted rent, and how much of that is AWS after Shreveport?
4. Did the $30B+ APAC sale process clear, get repriced, or get abandoned — and at what implied $/MW?
5. What is the portfolio-wide, contracted-lease-weighted DSCR, and how much refinancing comes due in 2026–28 against the ~$20B debt book?
6. What blended $/MW did recent US hyperscale leases (incl. AWS Shreveport) actually price at, and how does it compare to the ~$8M/MW Aligned mark?
7. How much of the ~$4B pipeline is shell/land-banked vs. leased/committed — i.e. speculative exposure if demand softens in H2-2026?
8. What is the on-site power strategy for the >1 GW campuses (Doña Ana NM "on-site power," any behind-the-meter gas/nuclear/PPA), and does it de-risk grid dependency or add execution risk?
9. What termination/PUE-cap terms do the marquee single-tenant leases carry — are they "financeable" enough to protect the ABS if a tenant's AI plans change?
10. Post-Apollo and post-APAC, what is the pro-forma STACK — capacity, markets, EV — that Blue Owl is actually retaining?
11. What is the equipment-lead-time exposure (transformers, switchgear, cooling) on the 2026–28 delivery schedule, and how much is pre-ordered?
12. How does cost-of-capital inside Blue Owl compare to standalone STACK's prior ABS spreads — is the platform benefit real and quantifiable?
13. What ROIC/yield-on-cost is STACK underwriting to on new US hyperscale builds, and how has it moved as land/power got scarcer?
14. What is the water/PUE/sustainability compliance risk in Virginia and Phoenix given tightening state scrutiny of data-center resource use?
15. What is management's own base/bear view on data-center oversupply in H2-2026–2027, and what leasing signals would make you slow the build?

---