BlitzLab Scoring · Oct 2026 · Model v1.0 (frozen)

Applied Digital — Blitz Index 51 (C)

The AI factory landlord — sold the future, still building it.

51Grade C

Confidence: High · Category: ai-infra

Score breakdown

Section	Score	Detail
A · Speed	19 / 25	FY2026 +167%; Q1 FY27 +322% YoY; $6.88B financing in one year (~1x market cap)
B · Moat	31 / 60	1.41GW contracted; 175MW online (~12%); investment-grade tenants; no software
C · Creative destruction	10 / 15	Hyperscalers rent instead of build — the landlord layer of AI
D · BlitzFailing	−9	Interest $77.4M/qtr exceeds adj. EBITDA $64.4M (0.83x); debt $6.38B = 7.3x TTM; no prepayments

Radar axes (0–100): Moat 52 · Speed 76 · Creative 67.

Watch item

Gates: G1 pass · G2 pass. 14 below threshold — first-ever score. May fiscal year-end.

Strengths

Contracted revenue $36B (base) / $86B (with renewals) — 15-year take-or-pay (SEC)
Q1 FY27 revenue +322% YoY; adj. EBITDA $64.4M vs $0.5M a year ago (SEC)
HPC Hosting NOI margin 89% — the recurring-rent core is high-margin (reported)
75MW building delivered on schedule (2026-07-01) — early execution proof

Concerns

Interest exceeds adj. EBITDA (coverage 0.83x) — sharpest leverage of the four
53.7% of revenue is one-off tenant fit-out (low-margin) — headline growth quality
175MW online of 1.41GW contracted — PF3/DF1/DF2 financing not yet closed
Prepayments $39M vs capex $2.87B — built with borrowed money

Research notes

1. Current Position — The Extreme Form of the Construction Blitz ("The Contracts Are Sold, the Buildings Are Under Construction")

Applied Digital is **the lowest score of the four-way neocloud race**. NBIS 59/C · IREN 56/C · CRWV 55/C · **APLD 51/C** (score file, confirmed). The business belongs to the same "heavy-equipment blitz" family as CRWV and IREN, but APLD is its extreme form — the state of **"the contracts are sold, the buildings are under construction"** is the score itself.

2. Blitzscaling Lens

**Rule 8. "Raise Too Much Money" — but APLD's money differs from NBIS's.** By execution speed of Rule 8, APLD is the cohort's top tier. FY2026 financing cash flows of +$6.88B — ~1x market cap raised in a year (A2 4 points). But Rule 8's essence is "secure runway," not "secure expensive runway." **The money is of different quality.** NBIS builds with $9B+ of customer prepayments — demanders share part of the construction risk. APLD builds with 7%-coupon senior secured notes and planned project debt (PF3·DF1·DF2 — uncommitted, press-reported) — bondholders collect interest and the company bears all the construction risk.

3. Competitive Landscape — The Four-Way Neocloud Race, and the Difference of "Stage"

**The designated competitors** sit inside the neocloud cohort. NBIS 59/C · IREN 56/C · CRWV 55/C · **APLD 51/C** (score file, confirmed) — an empirical case of v1.0 distinguishing "same industry, different stage." IREN is the vertically integrated GPU-cloud type; APLD is the hosting/campus developer type (candidate reference) — **different layers of the same destruction**.

4. Inflection Points

**Inflection Point 1. 600MW online + PF3·DF1·DF2 financing closes — the point that is everything.** What to watch: at FY Q2 earnings (expected 1/6–1/12, estimate — **no Q3-season slot**), progress on the 600MW online target (within 12 months, press-reported) and whether the PF3·DF1·DF2 financings close (Macquarie preferred equity + project debt, press-reported — currently uncommitted). The score has exactly one direction: whether the 600MW lights up on time. Repeating the on-time 75MW delivery (2026-07-01) proof is the upside trigger; construction delays meeting deteriorating financing terms point downward. Link: measured confirmation of B2 10 points is the upside condition. v1.0's structural blind spot — the $36B backlog is not directly reflected in the score. The 600MW measured reading is the first observation point filling that blind spot.

5. One-Line Risk

In a structure with quarterly interest of $77.4M exceeding adjusted EBITDA of $64.4M and no prepayments, if the delivery speed of the 600MW cannot outrun the accumulation speed of interest, "the interest clock ticks first" — that is the fundamental risk of the construction blitz.

1. Current Position — The Extreme Form of the Construction Blitz ("The Contracts Are Sold, the Buildings Are Under Construction")

The numbers make the extremity vivid. Of 1.41GW contracted, only 175MW is online (Polaris Forge 1, SEC confirmed) — **~12%**. Recurring rent is base rent of $65.8M per quarter (SEC confirmed) — less than **3%** of the 15-year take-or-pay contracted revenue of $36B annualized (~$2.4B/year, estimate). FY2026 revenue of $611.3M (+167% YoY, SEC confirmed) → Q1 FY27 of $341.9M (+322% YoY, SEC 8-K·10-Q confirmed) — the headline is explosive, but **53.7% ($183.5M) of Q1 revenue was tenant fit-out** (SEC confirmed; one-off in nature, margin ~17.6%). The 322% means "part of the sold future has started being recognized as revenue," not "the recurring business quadrupled."

Hence 51/C. Speed (A) is 19/25 — speed converted at 76.0, backed by 1.41GW contracted, 600MW+ online within 12 months, and the 2030 target of 3.5–4GW (earnings call, press-reported). The core of the recurring-rent engine is genuinely high-margin — HPC Hosting NOI of $58.8M / base rent of $65.8M = **89% margin** (press-reported). But moat (B) 31/60 — a pure leasing business with no software stack (B1 network effects 5 points, the cohort's lowest) and under-demonstrated execution at 12% online (B2 10 points) hold it back. And deductions (D) −9 (D1 −8 · D3 −1) finish the job. Total debt of ~$6.38B (SEC 10-Q, confirmed) = **7.3x** TTM revenue (~$872M, estimate). **Quarterly interest of $77.4M exceeds adjusted EBITDA of $64.4M** — coverage of 0.83x (SEC confirmed). The only buffer is that 80%+ of debt matures after FY2031 (press-reported).

**GATE verdict:** G1 (sustained growth or a second growth curve) PASS (confirmed) — Q1 +322% and $36B of contracted revenue ("sold futures") are the evidence. G2 (new markets' 10-year ceiling ≥ 10x current revenue, bar ≈ $8.7B) PASS (confirmed) — the $36B contract alone is ~4.1x the bar, so strong that no TAM estimate is needed. The evidence of growth is signed futures, not recognized revenue — the same way CoreWeave passed.

**What differs is the "quality of money" (confirmed/press-reported):** NBIS builds "with the customer's money" via $9B+ of prepayments (score file). APLD's deferred revenue is **a mere $39M** (SEC 10-Q, confirmed) — the take-or-pay contracts are strong, but cash comes later. In other words, APLD **"builds with borrowed money and gets paid later."** FY2026 financing cash flows of +$6.88B (SEC confirmed) — roughly 1x the market cap (~$7B, press-reported as of 2026-10-12) raised in a single year — centered on expensive 7.00%-coupon debt ($1.59B senior secured notes, 2031, ComputeCo 3 — SEC confirmed). More expensive money than NBIS's convertible-and-prepayment structure. Sensitivity to rate and financing conditions is the highest of the four.

**Two-sided assessment (per the score file):** - **On the upside:** Contracted revenue $36B (base) / $86B (with renewals) — ~41x TTM, 15-year take-or-pay (SEC confirmed) · all 3 anchor tenants investment-grade (CoreWeave + 2 hyperscalers, SEC confirmed) · adjusted EBITDA $64.4M (turnaround from $0.5M a year ago, SEC confirmed) · HPC Hosting NOI margin of 89% (press-reported) · 75MW Building 2 delivered on time (2026-07-01, earnings-call, press-reported) — early proof of execution capability · planned rent escalation of +15% plus 250MW expansion leasing by year-end — pricing power (press-reported) · revenue per employee ≈ $3.4M (~256 employees, SEC confirmed) — CRWV-class lean productivity. - **On the downside:** Interest > adjusted EBITDA (coverage 0.83x, SEC confirmed) — the sharpest leverage quality of the four · 53.7% of revenue is fit-out (one-off, low-margin, SEC confirmed) — the quality of the headline growth · 175MW online / 1.41GW contracted — all the realization risk lies ahead (PF3·DF1·DF2 financing uncommitted, press-reported) · effectively 100% of HPC leasing revenue from 3 tenants (v1.0 does not directly score customer concentration — a v1.1 candidate) · $39M of prepayments vs FY2026 capex of $2.87B (SEC confirmed) — the "build with borrowed money" structure · CEO's shares pledged to JPMorgan (DEF 14A, confirmed) — a governance yellow flag · GAAP net loss of −$221M per quarter (common-shareholders, SEC confirmed), dilution continuing.

The directional assessment is **"construction blitz (execution sprint)"** — the score has exactly one direction: whether the 600MW lights up on time. In a structure where the interest clock ticks first, building speed is the speed of survival.

2. Blitzscaling Lens

This difference detonates as the number 0.83x interest coverage. Interest expense exceeding adjusted EBITDA means even the time bought with borrowed money cannot be afforded — total dependence on a single future income stream called "future rent." Rule 8 said running out of money mid-race is fatal, so accept dilution. APLD did accept dilution (weighted-average 291.6M shares, ongoing — SEC confirmed), but hung a larger weight on "interest" than on dilution. That is why the score file proposes v1.1 candidate 1 ("distinguishing contract-backed project financing from debt-market-dependent financing," the leasing variant — "customer-prepayment type vs pure borrowing type"). **Borrowed money also has quality.**

**Rule 5. "Let It Burn" — the big fire is construction and financing; the small fires are the fit-out mix and the CEO's shares.** Hoffman's fire priority is "extinguish only the fires that can kill the company; let the small ones burn." For APLD, the true big fire is **lighting up the 600MW on time and closing the PF3·DF1·DF2 financings** (press-reported — uncommitted). The on-time 75MW Building 2 delivery (2026-07-01) and the first 100MW building going live (2025-10) are early proof of "can execute" — the big fire is still burning. But Rule 5's paradox: the big fire of construction delay is not one APLD can extinguish itself — power-delivery schedules, grid, materials, and financing closes are entangled. The only way to extinguish it is "keep building."

Meanwhile, the fit-out mix (53.7% of revenue, SEC confirmed) and the CEO's share pledge (D3 −1) are "small fires" — distracting but not lethal. Fit-out is a disclosed mix issue, not accounting fraud (D2 −0). But the small fires blur the big narrative of "the contracts are sold, the buildings are under construction" — missing the recurring revenue ($65.8M/quarter base rent) behind the 322% headline. Rule 5's lesson: not "extinguish the small fires," but "don't let the small fires take your eyes off the big fire (construction speed)."

**A variant of Rule 4 — "buildings cannot be patched like software," and APLD is their landlord.** Blitzscaling says "launch an embarrassing product and patch it" (Rule 4), but **a 75MW AI factory cannot be patched like software**. The same "heavy-equipment blitzscaling" grammar shared with the CRWV research — operating one level below in APLD's case. CRWV sells compute — APLD **sells not compute but sites, power, and buildings** (score file: "the landlord of the AI factory"). One level down the stack has two meanings. First, there is no technology risk — no exposure to GPU architecture changes. Second, **the only defenses are "power, sites, and permits"** — B1 network effects 5 points, the cohort's lowest, is the structural limit of a pure infrastructure lease with no software stack.

But Rule 4's variant also has a side favorable to APLD. The landlord has nothing to "patch" — the tenants (CoreWeave, hyperscalers) patch GPUs with their own money. The 89% HPC Hosting NOI margin (press-reported) is this landlord model's high margin demonstrated in numbers. The problem is **the high-margin core is still too small** — base rent $65.8M per quarter vs fit-out $183.5M per quarter (SEC confirmed). The five-growth-lever verdict (score file) — primary lever distribution (power, sites) + capital, no secondary lever, weak lever network effects — is the declaration of "the AI factory's landlord." The landlord model is beautiful when it runs at 89% margins. But when empty buildings are built, that beauty collapses the WeWork way — this time not in real-estate rents but in bond interest.

3. Competitive Landscape — The Four-Way Neocloud Race, and the Difference of "Stage"

**Not business quality but "stage" decides the ranking.** If the CRWV research's one-point gap was "the business belongs to CRWV, the balance sheet to IREN," APLD's last place is "the gap between promise and operation." ~12% of 1.41GW contracted is online — unlike CRWV (850MW+ online → 1.7GW target, press-reported) or IREN, APLD is at the earliest part of the "building the sold future" stage. D1 −8 is on par with CRWV but for a different reason — in the score file's words, **CRWV is "a compute blitz at 5x interest," APLD is "a construction blitz at 6x capex"** (Q1 FY27 investing cash flow of −$2.08B = 6.1x quarterly revenue, SEC confirmed; FCF −$2.02B per quarter).

**The financing-quality gap with NBIS** is this research's sharpest interpretive point. NBIS builds with the customer's money via $9B+ of prepayments; APLD builds with borrowed money on $39M of deferred revenue (SEC confirmed). Both hold 15-year long-term contracts (C2 4 points — the lock-in is contractual, not cash), but **the direction of cash is opposite**. The reason rate- and financing-sensitivity is the highest of the four — the formalization target of v1.1 candidate 1.

**The structural competitor** is power and permitting themselves. The bottleneck is not GPUs but grid interconnection, power delivery, and construction schedules — why B2 10 points reflects the "on-time 75MW delivery" proof while being discounted for "~12% of 1.41GW contracted" (score file). The interesting paradox: local moratoriums and tighter permitting become entry barriers for established campuses (earnings call, press-reported) — regulation as the destroyer's wall, not the destroyer's enemy. Large-scale power procurement (1,200MW gas-generation contract, 2030 delivery — press-reported) and four-year preemption of factory production capacity are the entirety of the supply-side moat (B1 5 points).

**The direction of creative destruction:** C1 (displaced order) 3 points — the "AI factory" lease model is a structural inversion in which hyperscalers lease capacity long-term from neocloud/DC developers instead of building themselves, but as a lessor of sites·power·buildings rather than compute, APLD's destruction is one step less direct than NBIS (4). C2 4 points — the irreversibility of 15-year take-or-pay is certain, but unlike NBIS there are almost no prepayments, so the lock-in is "the contract" rather than "cash." Combined with Ledger insight ③ (the real threat is fratricide), APLD's most direct threat is kinsmen attempting the same destruction (order competition from IREN, Nebius, et al.) and the hyperscalers' shift to "doing it themselves" — the moment tenants start building houses in the landlord model.

4. Inflection Points

**Inflection Point 2. Interest coverage — the trajectory of "the interest clock ticks first."** What to watch: the quarterly 10-Q ratio of interest expense (currently $77.4M per quarter — SEC confirmed) to adjusted EBITDA ($64.4M per quarter — SEC confirmed). The interest-coverage direction is this company's lifeline. That 80%+ of debt matures after FY2031 (press-reported) buffers maturity risk, but interest itself is a clock ticking every quarter. Whether quarterly FCF burn (−$2.02B — SEC confirmed) improves, and whether additional financing terms (rate, structure) deteriorate — observed together. Current D −9 is 6 points below the FLAG bar (15 points of deductions) — one rate hike and one quarter of delivery delay could trigger the watch flag. Link: cash of $2.95B + restricted cash of $0.31B (SEC confirmed) supports the score file's judgment of "not default-imminent," but carries the condition "as long as demand does not turn and building speed holds." None of blitzscaling's 9 rules guarantees construction schedules — this inflection point overlaps with the maximum risk.

**Inflection Point 3. Scaling of recurring rent — the quality of the 53.7% fit-out.** What to watch: whether the fit-out share ($183.5M of $341.9M in Q1 FY27, 53.7%, SEC confirmed — one-off, low-margin at ~17.6%) declines and base rent ($65.8M per quarter, SEC confirmed) expands. With the 89% HPC Hosting NOI margin (press-reported) as proof of the "high-margin core," the speed at which that core scales is the observation point. If the fit-out share falls below 50% and recurring rent crosses $100M per quarter, the A1·B4 discount factors lift. Link: before v1.1 candidate 2 ("adjust A1·B4 when one-off, low-margin revenue is 50%+") is formalized, only headline growth moves the score — APLD's 322% is the exemplar case. The essence of this inflection point is the transition from the "selling the future" stage to the "selling the recurring" stage.

**Observation checklist (quarterly):** ① 600MW online progress + PF3·DF1·DF2 close status ② direction of the interest-expense ($77.4M — SEC confirmed) / adjusted-EBITDA ($64.4M — SEC confirmed) ratio ③ declining fit-out share (53.7% — SEC confirmed) and expanding base rent ($65.8M — SEC confirmed) ④ improvement in quarterly FCF burn (−$2.02B — SEC confirmed) ⑤ maintaining the 89% NOI margin of recurring rent (press-reported) ⑥ easing of the 3-tenant concentration in HPC leasing revenue. ①·③·⑤ are "execution" metrics, ②·④ are "firewall" metrics, ⑥ is a "diversification" metric. If execution metrics trend up and the firewall holds, 51/C rises in the direction of the "construction blitz" assessment. If one quarter of delivery delay meets a rate hike, D1's (−8) warning converts into real leverage pressure.

5. One-Line Risk

*BlitzLab v1.0 (frozen). Unconfirmed figures must be labeled estimate / press-reported.* *For education and research purposes only. Not a recommendation to buy, sell, or hold any security.*

### Verification Log - Verification review waived.

*Translated from the Korean edition (2026-10-15). For education and research purposes only. Not a recommendation to buy, sell, or hold any security.*

Next earnings

Jan 6, 2027 · After close · Expected

Coverage

As of October 2026 · Model v1.0 (frozen) · Confidence: High (listed, SEC filings) · Not investment advice — a scholar's assessment.

BlitzLab scores businesses against a scholar’s standard — not investment advice. No price targets, no buy/sell recommendations. Figures are from SEC filings (XBRL) as labeled in the source scorecard.