Applied Digital

Refresh delta (since 2026-08-10)

What changed — six material moves:

1. Delivery kept to schedule: live capacity went 100 → 175 → 250 MW. Building 2 Phase 1 (75 MW) at Polaris Forge 1 reached Ready for Service on 2026-06-30, and Building 2 Phase 2 (three 25 MW halls, 75 MW) on 2026-10-02, taking PF1 to 250 MW of 400 MW. The prior dossier's "100 of 1,410 MW earning rent" is now 250 of 1,410 MW (18%). The June step was disclosed in the Q4 release but missed by the prior dossier.
2. The tenant-concentration discrepancy is resolved — and it is real concentration, not a garble. The CEO said ~$20B of contracted revenue comes "from a single world-class customer"; the Q4 release confirms Polaris Forge 3, Delta Forge 1 and Delta Forge 2 are all with "the same high investment-grade hyperscaler". That is 810 MW / ~$20.2B = 57% of contracted MW and ~56% of contracted value, and 80% of the 1,010 MW that is not CoreWeave — which is almost certainly where the ">80%" in the call summary came from.
3. The proxy put a number on who gets paid. FY26 Summary Compensation Table: Cummins $189.2M (of which $185.9M stock awards), Zhang $100.0M, Mohmand $40.5M — $329.7M of reported FY26 compensation (Summary Compensation Table; ~98% is grant-date value of equity awards, mostly five-year grants). The stock-awards column is $323.9M of the $329.7M; footnote 2 says these grant-date (ASC 718) amounts "do not correspond to the actual value that may be received." Most of it is the January/February 2026 five-year grants (Cummins alone: PSUs $122.1M, RSUs $45.4M), which vest over five years or, for the CEO's PSUs, not at all unless the stock averages $50 for 90 days (it is ~$24.80). The FY25 say-on-pay passed with only 52.6%. Cummins beneficially owns 7.8%, of which 17.59M shares are pledged to a JPMorgan line of credit.
4. Dilution re-accelerated. Shares outstanding 287.88M (2026-05-31) → 302.32M (2026-09-23), +14.44M / +5.0% in under four months. Cumulative Series G converted rose from 995,146 shares / ~52.8M common (10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes. date, 2026-07-29) to 1,197,050 / ~60.6M (2026-09-25) — i.e. PEPA puts continued after year-end. The POSASR prospectus covers 53.1M shares for resale — 25.53M carried over from the 2025 registration, 24.56M newly registered (Fifth/Sixth PEPA amendments) and 3.0M warrant shares; only the 24.56M is new supply, which the tape read as overhang.
5. A related-party correction. The prior dossier said the B&W guarantee ended on a $37.0M payment "by or on behalf of Base Electron." The 10-K's own Note 7 is unambiguous: APLD paid the $37.0M, and it was added to a demand grid note receivable from Base Electron. Base Electron is "owned and managed by certain officers and directors of the Company, including Messrs. Cummins and Zhang and Dr. Nottenburg, acting in their individual capacities"; Cummins is its CEO; APLD owns ~10%; the grid note stood at $59.2M on 2026-09-01 (cap $100M).
6. The market kept selling. Close $29.06 (2026-08-10) → $25.38 (2026-10-02, last full close before this run), −12.7%; low close $23.42 (2026-09-15); twelve days with moves ≥5% through 2026-10-02. Intraday 2026-10-05 the stock was ~$24.6–24.8; the $24.80 quote is used for the multiples below. Short interestShort interestHow many shares have been borrowed and sold by people betting the price falls. 23.1% of float (from 27.0%). Consensus PTConsensus price targetThe average of what published analysts think the share price should be. An opinion poll, not a forecast. $66.43, 15 analysts; initiations in the window: Morgan Stanley Hold/$38 (2026-09-18), Rothschild & Co Redburn Hold/$22 (2026-09-21), UBS Buy/$38 (2026-09-23), Jones Trading Buy/$70 (~2026-09-25; one secondary citation only).

What held. The structural frame — a build-to-suit, take-or-pay landlord whose game is tenant credit and cost of capital — is untouched. Nothing in the window weakened a lease, a tenant or a financing. The execution leg actually strengthened (two on-time RFS events). What moved is the governance and dilution leg of the bear case, which the proxy made concrete, and the concentration leg, which is now quantified: CoreWeave was ~59% of FY26 full-year revenue (10-K), one unnamed hyperscaler is ~56% of contracted value tomorrow.

---

Phase A — Understand the business

Company Overview — REFRESHED (partial)

The business description in the previous dossier Lens 1 stands (two segments; five campuses; 1,410 MW; ~$36.2B contracted; ~$86B with options). Three updates:

• Live capacity by campus:

• Tenant map, resolved. Three counterparties: CoreWeave (400 MW, PF1), an investment-grade hyperscaler (200 MW, PF2), and a single high-investment-grade hyperscaler "H" holding 810 MW across three campuses. Zhang's PSU disclosure gives an independent cross-check: at his grant date (2026-02-06) APLD had 700 MW of ≥15-year hyperscaler contracts, 300 MW of it investment-grade; by 2026-05-29 the Second Hurdle (1.6 GW) was 88.125% achieved — i.e. exactly 1,410 MW.
• Data Center Hosting still runs 286 MW (Jamestown 106 MW, Ellendale 180 MW) at full capacity for one crypto customer; management calls it "the highest return-on-assets business in the company".

Supply Chain — CARRIED

Unchanged; see the previous dossier Lens 2 (ABB, BASX, Caterpillar; power the binding constraint). Nothing in the window moved it.

Competitive Advantages — CARRIED

Unchanged; see the previous dossier Lens 3. The two on-time RFS events (2026-06-30, 2026-10-02) add evidence to the "delivery reliability" claim management now leads with, but one campus with one tenant is still a thin sample.

Segments — CARRIED

No new segment data (next disclosure is the FY27-Q1 10-QForm 10-QThe quarterly version of the annual report. Lighter, and not audited.). See the previous dossier Lens 4. One quarterly cross-check now on the shelf: HPC base rent was $44.1M in both FY26-Q3 and FY26-Q4 — flat, because only the first 100 MW was paying in both quarters. Q4 HPC revenue was $203.0M = $44.1M base rent + $152.4M tenant fit-out + $6.5M recoveries.

---

Phase B — Measure performance

Earnings Result — REFRESHED (no new print; corrections + pre-print setup)

No quarter has been reported since the boundary. The FY27-Q1 print lands 2026-10-07. What this lens adds:

Two corrections to the prior read of Q4 FY26:

• The "+407%" is year-over-year, not sequential. The secondary transcript summary carried in the prior dossier said "jumped 407% sequentially"; the release says Q4 revenue of $258.7M vs the fiscal fourth quarter 2025. Sequentially it was $126.6M → $258.7M, +104%.
• The "$1M prior quarter" adjusted EBITDA is FY25-Q4. Q4 FY26 adjusted EBITDA $42.4M "compared to an Adjusted EBITDA of $1.0 million for the fiscal fourth quarter 2025". Sequentially it fell: Q3 FY26 was $44.1M. This closes the prior dossier's open conflict.
• NOI run-rate, not FY NOI, is the right base for the $1B target. Q4 FY26 NOI was $39.9M (91% margin), i.e. ~$160M annualized. The prior dossier compared $1B to FY26 NOI of $90.4M and called it "11x"; against the exit run-rate it is ~6.3x.

Pre-print setup for FY27-Q1 (quarter ended 2026-08-31):

• Consensus: revenue ~$116–135M depending on source — ~$116.3M (MarketBeat), ~$124.6M (Futu), ~$134.9M (scanx.trade preview; Yahoo "Applied Digital's Revenue Is Likely About to Plunge"); EPS ~$(0.28) to $(0.30). Options imply a ~12% move.
• What the contracts alone imply: base rent ≈ $66M (100 MW × full quarter at Q4's $44.1M/100 MW + 75 MW × 2 of 3 months from the 2026-06-30 RFS: $44.1M + $44.1M × 0.75 × ⅔), NOI ≈ $60M at 91%. Add Data Center Hosting at the FY26 run-rate ($39M/qtr: $154.4M ÷ 4) and ChronoScale ($18M/qtr: FY26 total $611.3M − adjusted $539.7M = $71.6M ÷ 4) and the ex-fit-out revenue floor is roughly $125M — inside the ~$116–135M consensus range before a dollar of tenant fit-out, which ran $152.4M in Q4. Assumptions: rent commences at RFS at PF1-Building-1 per-MW rates; DCH and ChronoScale flat. If fit-out continues at anything like Q4's pace (PF1 B3/B4 and PF2 are all mid-build), the headline will again "beat" on construction pass-through. Judge the print on base rent and NOI, not revenue.
• Balance sheet is still the audited 2026-05-31 set (see the previous dossier Lens 5). Post-year-end changes known from filings: $1.59B 7.000% notes closed 2026-06-16 (repaid the $300M bridge); $150.0M of Series G issued 2026-06-17/22; further Series G conversions to 2026-09-25 (move 4 above); $7.0M drawn by ChronoScale on a new $100M APLD grid note.

our figures now carries two quarters (FY26-Q3 added this run, from 10-q-2026-q1.md). FY26-Q1/Q2 were deliberately not added: their as-reported revenue predates the ChronoScale recast (the Q3 10-Q's nine-month revenue of $352.6M implies a recast six-month figure of $225.9M against $190.8M as originally reported), so those rows would not be comparable.

Earnings Calls — REFRESHED (thin)

transcripts/ still does not exist — five dossiers running. This pass tried fool.com / insidermonkey.com / Benzinga (no APLD post), Yahoo (404), and our model against Investing.com (HTTP 403). The only call text read was the Investing.com Q4 FY26 transcript via WebFetch (labelled; WebFetch reports it as the full transcript, but it was read through a summarizer, so the quotes below are the strongest available, not verified verbatim).

What it adds beyond the prior dossier:

• CEO, on concentration: "Together representing approximately $20 billion in long-term contracted revenue from a single world-class customer." This is the sentence that resolves the ">80%" discrepancy (see delta move 2).
• CEO, on the $1B NOI target: "We now expect to achieve that run rate goal a year from now. We're three years ahead of schedule." It is a run-rate claim dated ~July 2027.
• CFO, on capex: "around $600 million-ish for the upcoming quarter. That will take a step up as we enter more advanced stages of construction at the new campuses."
• CFO, on SBC: Q4 stock comp included $47.9M tied to one-time ChronoScale awards and $65.1M tied to PSUs. That is $113.0M of FY26's $219.3M SBC landing in Q4 alone, and it lines up with the proxy (PSUs vested 2026-06-22/07-01; MIP units in ChronoScale granted 2026-04-09).

Tone between calls: there was no call, but the company spoke twice through press releases (2026-09-28, 2026-10-02), and both lead with the same frame: execution, not contracts — "Securing power is only the starting point. The real work is designing, building, commissioning, and operating". No new lease has been announced since Delta Forge 2 (2026-06-05), after five campuses in ~10 months. The "1.7 GW actively marketed / 250 MW in advanced negotiation" from the July call has not converted in public. Watch whether the Oct 7 call still uses those numbers.

Comps — REFRESHED (partial — first sourced peer figures in three dossiers)

All. Every peer is loss-making, so P/E and EV/EBIT are n/a; dividend yields are n/a (none pay). 5-yr average ROE n/a.

* stockanalysis.com uses 291.47M shares — the 10-K cover count dated 2026-07-28, now superseded. On the 302.32M shares outstanding at 2026-09-23, APLD's market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. is ~$7.50B and EV ~$10.88B. Including the $1.956B Macquarie redeemable NCI as debt-like: ~$12.84B.

EV / audited contracted base-term rent ($35.845B, 10-K Note 17): ~0.30x ex-NCI (was ~0.33x at the boundary on the same method), ~0.36x with NCI. Trailing EV/Sales is close to meaningless here — trailing revenue is 44% fit-out pass-through — so the backlog multiple remains the comparison that means something. Peer backlog multiples are n/a; do not infer them from EV/Sales.

Stock-Price Catalysts — REFRESHED

Window 2026-08-10 → 2026-10-02: $29.06 → $25.38 close, −12.7% (last full close; ~$24.6–24.8 intraday 2026-10-05). Daily moves ≥5%:

Reading: most of the window's large moves have no company-specific primary-source cause and coincide with peer moves — the stock is trading as a high-beta member of the AI-infrastructure/neocloud basket (~23% short interest, ~12% implied earnings move). The two identifiable company events cut opposite ways and the dilution one dominated: an on-time delivery was worth +5%; a resale registration plus pre-earnings caution was worth −6.6% despite a $70 initiation the same week. What the market reacts to has shifted again — from lease signings (2025), to non-reaction on the backlog (July 2026), to supply of stock (September 2026).

Insider activity: director Nottenburg sold 75,000 shares at $31.15 on 2026-08-04 — before the 2026-08-10 boundary, and already in the prior dossier as the "$2.34M director sale"; listed here only to identify it; President Zhang's 2026-09-12 Form 4 is a code-F tax withholding of 49,188 shares on an RSU vest, not a sale. No open-market insider purchases found.

The print — Wednesday 2026-10-07, after the close (call 17:00 ET). What to watch, in order of information value:

1. Base rent and NOI for the quarter, not revenue. The contracts imply ~$66M base rent / ~$60M NOI. A number well below that means rent commencement lags RFS — the single most important thing to learn.
2. Cost-to-complete or a capex plan. Q1 capex vs the "~$600M-ish" guide; any campus-level cost-to-complete figure (still disclosed nowhere — 10-K Note 18 confirms construction contracts are milestone-based and cancellable but gives no total).
3. Liquidity at 2026-08-31: unrestricted vs restricted cash, Macquarie draws (was $1.8B of $5.0B), Series G issued in the quarter.
4. Share count at quarter-end and at the 10-Q cover date (302.32M at 2026-09-23 is the baseline).
5. PF2 timing: whether "start coming online in 2026" holds — PF2 plus the rest of PF1 is what the $1B NOI run-rate needs (Lens 11).
6. Any new lease against the "1.7 GW actively marketed" claim — none since 2026-06-05.
7. The $1B NOI run-rate by ~July 2027 — restated, softened, or dropped.

Next dated events: annual meeting 2026-11-04 (virtual; say-on-pay vote); McConnell motion-to-dismiss ruling still pending (see Lens 10).

---

Phase C — Judge people & books

Management — REFRESHED (the proxy)

The 2026 proxy is the most informative governance document this name has filed, and it moves this lens more than anything since the CoreWeave lease.

Pay:

The headline grants are five-year "sole award" mega-grants (Jan/Feb 2026): Cummins 4.5M PSUs + 1.5M RSUs, Zhang 1.5M PSUs + 0.5M RSUs, Mohmand 0.75M PSUs + 0.25M RSUs. The hurdles are the most useful thing in the document because they tell you what the board says it is solving for:

• Cummins: one third each at a 90-day average close of $50 / $75 / $100 within five years of the 2026-01-06 grant ($30.26 grant price). Forfeited in full if $50 is never reached. Stock today: ~$24.80. The company's own framing: max payout $450M against ~$21B of market cap created (a 2.15% "sharing ratio").
• Zhang: 375k PSUs each for 600 MW (IG only) and 1.6 GW (all) signed, and the same two thresholds reaching RFS. Signed was 88.1% of 1.6 GW at 2026-05-29.
• Mohmand: 375k PSUs each for $1B and $2B of trailing-twelve-month HPC NOI by 2031-02-28. Note that this is a much easier test than the CEO's public "$1B run-rate a year from now" (TTM lags run-rate by up to a year, and the deadline is 2031). Contracted NOI was 65.1% of the $2B hurdle (~$1.30B) at 2026-05-29 per the change-in-control math.

Response to the 2025 vote. Say-on-pay got 52.6% in 2025; the committee reached holders of ~36% of shares and met with five holders (~15% of shares), all of whom had voted against. Changes: FY27 annual bonus becomes formulaic on adjusted cash EBITDA (±25% individual modifier); expanded disclosure. The mega-grants were made after that vote.

Skin in the game — and the pledge. Cummins 23.36M shares (7.8%), of which 17.59M held by Cummins Family Ltd are pledged under a JPMorgan line of credit. All directors and officers: 9.5%. Other 5% holders: Hood River Capital 7.4%, Jane Street 7.0%. A pledge of ~75% of the CEO's holding is a standing forced-selling risk in a drawdown — this name has fallen ~48% from its $47.28 close on 2026-05-29 to ~$24.80.

Related parties:

• Base Electron: owned by Cummins, Zhang and Nottenburg individually plus third parties; Cummins CEO, Zhang President. APLD owns 10%, provides admin services ($0.1M incurred, nothing paid), paid $37.0M to B&W to end a guarantee of Base Electron's ~$2.4B design-build contract, and holds a demand grid note at $59.2M (2026-09-01), cap $100M, at the short-term AFR. A board committee for Base Electron transactions was created 2026-04-29.
• ChronoScale (APLD ~96%): 1%-of-revenue services agreement; a $100M APLD grid note, $7.0M drawn, demand right waived until 2027-08-20; MIP units equal to ~5.25% of ChronoScale's issued shares granted to four executives on 2026-04-09, fully vested.
• Six family members of Cummins were paid $3.3M in FY26 through 2026-09-01.
• Board: Chuck Hastings is ex-Chairman of B. Riley (to March 2025); Cummins was president of B. Riley & Co. 2002–2011.

Archetype read. Founder-dealmaker, unchanged — and the evidence that he can deliver (two on-time RFS events) improved this quarter. The evidence on whose interests the structure serves got worse: $189.2M of reported FY26 compensation (almost all grant-date value of five-year equity, much of it out of the money), a pledge, a power company co-owned by the CEO, President and a director (among numerous third parties) and financed by the issuer, and executive equity carved out of a 96%-owned subsidiary. None of it is unlawful or undisclosed. All of it compounds.

Forensic Red Flags — REFRESHED

Carried as live from the previous dossier Lens 10 (SBC as the dominant non-GAAP add-back; fit-out flattering revenue; lease-incentive assets; FY2031 maturity wall; "no critical audit matters"). New or sharpened:

• The $37.0M was APLD's cash, booked as a loan to a related party. 10-K Note 7: "the Guarantee was terminated, effective upon payment of $37.0 million by the Company on behalf of Base Electron which was added to the… Demand Grid Promissory Note. As such, the Company recorded a related party loan receivable". The commitments note (Note 18) uses the vaguer "by or on behalf of." The prior dossier quoted Note 18 and missed this; the substance is that the issuer funded the exit from a related party's guarantee and now carries it as a receivable from a pre-revenue company its CEO runs.
• SBC timing. $113.0M of FY26's $219.3M SBC landed in Q4 ($47.9M ChronoScale one-time + $65.1M PSUs). Whether FY27 SBC falls back depends on the PSU accounting for the January/February mega-grants (market-condition PSUs expense on a schedule whether or not the price hurdle is hit). FY27 SBC is the add-back to watch on Oct 7.
• Equity supply. Series G conversions of ~7.8M common shares between 2026-07-29 and 2026-09-25; 50.1M PEPA conversion shares now available for resale under the registration (25.53M carried from 2025 + 24.56M newly registered). The floor price remains board-adjustable downward.

Regulatory findings (required):

• SEC Litigation Releases / AAERs: none. regulatory/regulatory-findings.md re-run with --force on 2026-10-05, period 2021-10-05 → 2026-10-05, 0 findings.
• Non-SEC enforcement: web search "Applied Digital" (FTC OR DOJ OR "consent decree" OR settlement OR fine OR penalty) enforcement 2026 returned no hits naming the company.
• 10-K Item 3: McConnell v. Applied Digital, No. 3:23-cv-01805 (N.D. Tex.) — Section 10(b)/20(a) claims, including that the board was not "independent" under Nasdaq rules; stayed and administratively closed since 2025-09-08 pending the motion to dismiss. No ruling found as of 2026-10-05. Derivative suit dismissed 2024-06-05 (unchanged).
• Conclusion: no SEC or non-SEC enforcement action found as of 2026-10-05; one stayed securities class action.

---

Phase D — Project & stress-test

Forward Projection — REFRESHED

Model integrity. our model re-run 2026-10-05 with 2 quarters in our figures (FY26-Q3 added). Opening balance sheet sourced = yes (2026-05-31 audited). our model: "No computed values. The workbook has 51 formulas and no cached results." The workbook has never been opened and saved; our figures still holds untouched seeds (gross margin 50%, revenue growth 8%, capex 5% of revenue against an actual ~498%, share price 0). Nothing below comes from the workbook. No our model create (unattended).

The contractual schedule is unchanged — see the previous dossier Lens 11 and our figures (FY27 minimum rent $451.1M vs scheduled interest $377.5M; FY28 $1.454B vs $399M; FY29 $2.252B vs $359M; total $35.845B; $4.195B principal due FY2031).

New this pass — the $1B NOI run-rate, decomposed:

• PF1 Building 1 pays $44.1M per quarter per 100 MW of base rent, i.e. ~$1.76M/MW/yr. The hyperscaler leases imply ~$1.67M/MW/yr ($7.5B ÷ 300 MW ÷ 15 years; $5.0B ÷ 200 MW ÷ 15).
• $1B of NOI at a 91% margin needs ~$1.10B of base rent, i.e. ~620–660 MW live and paying.
• PF1 at full build (400 MW) + PF2 (200 MW) = 600 MW → ~$1.04B base rent → ~$0.95B NOI. So the CEO's "run rate a year from now" (~July 2027) is a claim that every remaining PF1 and PF2 megawatt — 350 MW from today's 250 — is live and rent-paying within ~9 months, with essentially no slack. PF3 and DF1 (CY2027 "initial operations") would have to contribute too if PF1/PF2 slip at all.
• This makes the $1B target falsifiable quarter by quarter: live MW should be ~400 (PF1 complete) plus first PF2 halls by the January print, and ~600 by the July 2027 print.

EPS path — carried with one adjustment: FY2027 base ~$(0.90), FY2028 ~+$0.10, FY2029 ~$1.60 from the previous dossier, re-based to ~305M shares (302.3M outstanding + continuing PEPA/RSU issuance; was ~288–300M). The per-share effect is small (−1% to −5%) but it is in the wrong direction and the PEPA has up to 50.1M more shares registered. The structure of the argument is unchanged: FY2027 is a GAAP loss year; FY2028 is the crossover; the equity is a bet on getting there without a dilutive raise into a weak tape.

Still the biggest missing number: cost-to-complete. Re-confirmed absent from the 10-K (Note 18 says only that construction contracts are milestone-based and generally cancellable). The prior dossier's bound ($7–13B remaining vs $9.6B identified capital) stands, with identified capital since reduced by Q1 capex ($600M guided) and increased by the $150M June Series G and later puts.

Bull vs Bear — REFRESHED

Bull case. Everything that is physical went right this quarter. Two RFS events on schedule, live capacity 2.5x in four months, PF1 at 250 of 400 MW. The audited $35.8B backlog is unchanged; no tenant, lease or financing weakened. The concentration question has an answer, and the answer is a high-investment-grade hyperscaler that chose APLD three campuses in a row. The stock fell ~13% to the 2026-10-02 close in a window with no adverse operating news, so EV/contracted rent fell to ~0.30x. The Oct 7 print (consensus revenue ~$116–135M against a contract-implied ex-fit-out floor of ~$125M) is, more importantly, the first quarter with 175 MW paying.

Bear case — what can permanently impair:

1. Governance as a cost of capital. $329.7M of reported FY26 compensation (Summary Compensation Table; ~98% is grant-date value of equity awards, mostly five-year grants) for three people, a 52.6% say-on-pay, a CEO who pledged ~75% of his stake and runs a power company he co-owns with the President and a director (among numerous third parties) that the issuer has lent $59.2M to, and executive equity in a subsidiary. A board that approved all of this will approve the next PEPA floor cut. That is what the equity is pricing.
2. Dilution is back on the accelerator. +5.0% shares in under four months, 50.1M PEPA shares registered for resale, and the CEO's PSUs pay nothing below a $50 average close — which aligns him with the price but not with per-share value if issuance funds the build.
3. Concentration rotated, it did not fall. CoreWeave was ~59% of FY26 full-year revenue per the 10-K (on an SPV with a springing guaranty); tomorrow one hyperscaler is ~56% of contracted value. A single counterparty decision dominates either way.
4. The $1B NOI run-rate needs ~350 MW in ~9 months with no slip. If it is restated on Oct 7, the stock loses the one near-dated number bulls were holding.

Pre-mortem (18 months out). PF2 energized two quarters late on transformer lead times; the $1B run-rate quietly became "FY2028"; rent at the IG campuses started on time but FY27 interest ate it; the PEPA was drawn at a lowered floor through the spring; the CEO's pledged block became a story in a down-tape. The buildings were fine. The equity was diluted and de-rated on governance, not on assets.

Are multiples too high? No GAAP multiple applies. On backlog, ~0.30x EV to audited contracted rent is cheaper than at any point in the prior two dossiers. The market is discounting the gap between contracts and per-share value — dilution, related-party leakage and the cost of the Macquarie preferred — rather than doubting the leases.

Contrarian view. At the boundary the contrarian point was that the market prices a net-lease receivable as a growth stock. The sharper version now: the market believes the buildings and the tenants, and is pricing the shareholder's share of them. If the Oct 7 print shows rent starting on time, a capex plan, and no new dilution in the quarter, the discount has little left to stand on. If it shows another quarter of puts and a softened $1B, the 0.30x was earned.

Devil's Advocate — REFRESHED

Dismantling the bull case.

• "On time and on budget" is one campus, one tenant. Both RFS events are PF1 buildings for CoreWeave, which has been on site since 2025. Nothing has been delivered to an investment-grade tenant, in Louisiana, or at PF2/PF3. The execution track record that bulls now lean on is narrower than the backlog it is asked to support.
• No new lease in four months after five in ten. The July call said 1.7 GW is being marketed and 250 MW is in advanced negotiation. Since 2026-06-05: nothing. Either demand paused or terms got harder. Neither is in the price narrative.
• The pay design rewards the wrong margin. The CFO's NOI hurdle is trailing-twelve-month by 2031 — a test he passes on contracted leases alone if they simply deliver late. The President's hurdles reward signing and energizing megawatts, not the return on them. Only the CEO's is tied to price — and price is what PEPA issuance at a board-adjustable floor dilutes. No NEO hurdle is per-share.
• The related-party pattern is escalating, not static. 2025: B. Riley ties, a stayed suit alleging non-independence. 2026: an IPP co-owned by the CEO, President and a director (among numerous third parties) whose ~$2.4B design-build contract the issuer guaranteed, then paid $37M to exit — that $37M sits inside the $59.2M grid note, one exposure of $59.2M, not two; a subsidiary whose equity was carved out to executives at grant; six family members on payroll. Each step is disclosed and committee-approved. A short does not need any of it to be improper — only for it to keep growing.
• The pledge is the trap door. 17.59M pledged shares at ~$24.80 is ~$436M of collateral. The stock fell ~48% from the May 29 close. Margin terms are not disclosed. A forced sale would land in exactly the tape where the PEPA is also selling.
• What must hold for today's price: PF1 complete and PF2 energizing on schedule; rent commencing at RFS; cost-to-complete inside identified capital; the PEPA used sparingly; one hyperscaler and CoreWeave both performing for 15 years; the FY2031 wall refinancing. Break the first two and the $1B run-rate dies on the January print.
• Single scenario that permanently impairs: an IG-campus delivery slip (PF2 or DF1) combined with a PEPA-funded bridge at a lowered floor — the same scenario as last time, now with a governance accelerant: a down-tape that triggers the pledge.

Management Questions — REFRESHED priority order

The prior 15 stand (see the previous dossier and the previous dossier Lens 14). Reordered top eight for the 2026-10-07 call and the 2026-11-04 annual meeting:

1. What is total remaining cost-to-complete for the 1,410 MW, campus by campus, and how much of it is funded today without PEPA or common?
2. The $1B NOI run-rate "a year from now" needs ~600 MW paying — PF1 complete plus all of PF2. On what dates do PF1 Buildings 3/4 and each PF2 building reach RFS, and does rent start at RFS?
3. How much Series G did you put between June 1 and today, and at what floor were the Aug–Sep puts converted? The ~201,904 preferred converted after the 10-K into ~7.8M common imply ~$26/share, below the $34 floor in the 10-K — so the floor appears already to have been lowered via the Conditional Notice mechanism (posasr). What is it now?
4. The CEO has pledged 17.59M shares to JPMorgan. What are the LTV and margin-call terms, and does the board's policy permit further pledging?
5. APLD paid $37.0M to exit a guarantee of Base Electron, a company you, the President and a director co-own (with third parties), and now carries $59.2M of demand notes from it. What is Base Electron's funding plan, when is the note repaid, and what has APLD received for its ~10%?
6. No new lease since 2026-06-05. Has the 250 MW "advanced negotiation" closed, and has pricing per MW moved?
7. FY27 SBC in dollars, given the January/February mega-grants are market-condition PSUs that expense regardless of the price hurdle.
8. Define "springing" in the CoreWeave ELN-02/ELN-03 guaranties (carried from the prior dossier — still unanswered).

---