Core Scientific — Blitz Index 52 (C)

The landlord of the AI builders — a second blitz after Chapter 11.

Score breakdown

Radar axes (0–100): Moat 55 · Speed 72 · Creative 67.

Watch item

Gates: G1 pass · G2 pass (estimated). 13 below threshold — first-ever score. CoreWeave $9B takeover rejected by shareholders (Oct 2025).

Strengths

• Colocation revenue +1,194% YoY to $136.7M/quarter — 83% of total (SEC)
• Contracted revenue $24B+ (~55x TTM) + AMD 2.5GW expansion option
• Gross margin 6.4% → 42.7% in one year; colocation-only margin 58.5% (SEC)
• AMD 529MW 15-year lease — first proof of the independent route

Concerns

• Negative equity: $7.68B total debt vs $5.26B assets (SEC)
• Single customer = 83% of revenue (CoreWeave, Q2 2026, SEC)
• GAAP net loss −$1.16B/quarter (mostly non-cash warrant/CVR remeasurement)
• AMD volumes contribute from 2027 — execution gap in H2 2026

Research notes

1. Current Position — A Fallen Blitzer's Second Blitz (Recovery Blitz)

Core Scientific is at the **start of Act 2 of its blitz**. Act 1 was bitcoin self-mining expansion — that act ended with the Chapter 11 filing on 2022-12-21 (revived 2024-01-23, ~$400M of debt extinguished, existing shareholders diluted 40% — press-reported). Act 2 is AI-dense colocation: Q2 2026 colocation revenue of $136.7M (+1,194% YoY, SEC confirmed) was 83.3% of consolidated revenue (SEC confirmed), and total revenue flipped from $78.6M to $164.2M (+108.8% YoY, confirmed). Gross margin improved from 6.4% to 42.7% over a year (confirmed), colocation-only margin 58.5% (confirmed), adjusted EBITDA from $28.5M to $41.1M per quarter (+44% YoY, confirmed). The pivot "from miner to landlord" is confirmed in the numbers — also the basis for the G1 (sustained growth or a second growth curve) pass (confirmed).

2. Blitzscaling Lens

**Rule 8. "Raise Too Much Money" — burned once by that very money.** CoreWeave is the aggressive execution of Rule 8 (total debt $35B+), and for IREN the qualitative distinction of contract-backed financing was key. CORZ goes one step further — **a company that actually collapsed while executing Rule 8**. The 2022 Chapter 11 is Rule 8's paradoxical proof — "running out of money mid-race is fatal": it raised too much money, expanded on leverage, and when the cycle turned, the money consumed the company. ~$400M of debt extinguished (note-to-equity conversion), existing shareholders diluted 40% (press-reported) — Rule 8's "secure runway at the cost of dilution" was realized, in the end, as "paying off debt with dilution."

3. Competitive Landscape — Coordinates Inside the "Landlord" Cohort, and the Tenant's Irony

**The designated competitors** sit inside the small-neocloud/hosting cohort. APLD (contracted revenue ~$36B, 1.41GW+), WULF (Anthropic 20-year ~401MW, $19B), CIFR (Fluidstack + AI lab $9B+) — all companies of the same grammar: "power-and-land preemption → long-term take-or-pay → high-capex delivery" (provided reference). CORZ's distinction is the **"AI landlord" position with the chipmaker (AMD) itself as tenant** — a different species of blitz from CRWV·NBIS, which does not buy GPUs but sells power and sites. GPU-neutrality (hosting both NVIDIA and AMD) is another singularity.

4. Inflection Points

**Inflection Point 1. Diversifying the 83% single customer — the speed of adding tenants.** What to watch: at the Q3 2026 earnings (expected Friday 10/23 — an estimate; no company announcement), whether CoreWeave's share of colocation revenue falls from Q2's 83% (SEC confirmed), and whether the AMD 529MW deployment progress (deployments expected 2027 — press-reported) becomes concrete. The speed at which the "landlord" escapes dependence on one tenant is this company's survival speed. AMD volumes actually go live from 2027 — until then the 83% structure persists. Link: the risk is already reflected in the score file's D1 −9 — debt service capacity itself is tied to a single customer's contract performance. Before v1.1 candidate 2 (a quantitative customer-concentration anchor) is formalized, this risk is not additionally reflected in the score — v1.0's structural blind spot, and the reason 52/C does not fully capture the "landlord's fate."

5. One-Line Risk

In a capital-deficiency position with 83% of revenue riding on a single tenant (CoreWeave), if that tenant's execution speed bends, the AI landlord's second blitz could rewrite its first blitz (Chapter 11) with the same ending — that is the fundamental risk of the fallen blitzer.

1. Current Position — A Fallen Blitzer's Second Blitz (Recovery Blitz)

Core Scientific is at the **start of Act 2 of its blitz**. Act 1 was bitcoin self-mining expansion — that act ended with the Chapter 11 filing on 2022-12-21 (revived 2024-01-23, ~$400M of debt extinguished, existing shareholders diluted 40% — press-reported). Act 2 is AI-dense colocation: Q2 2026 colocation revenue of $136.7M (+1,194% YoY, SEC confirmed) was 83.3% of consolidated revenue (SEC confirmed), and total revenue flipped from $78.6M to $164.2M (+108.8% YoY, confirmed). Gross margin improved from 6.4% to 42.7% over a year (confirmed), colocation-only margin 58.5% (confirmed), adjusted EBITDA from $28.5M to $41.1M per quarter (+44% YoY, confirmed). The pivot "from miner to landlord" is confirmed in the numbers — also the basis for the G1 (sustained growth or a second growth curve) pass (confirmed).

And yet the score is 52/C, 13 points below the threshold, because **the past and the present coexist on the same balance sheet**. Speed converted at 72.0 (A 18/25) ranks among the top of the small-neocloud candidate cohort, and moat (B) 33/60 is solid. But D1 −9 — a capital deficiency with total debt of $7.68B exceeding assets of $5.26B (6/30, SEC confirmed); long-term debt of $4.30B is ~10x TTM revenue (~$440M, estimate). It is −9 rather than −10 because of two buffers: cash of $1.77B (confirmed) and CoreWeave absorbing part of capex ($180.9M per 6 months, SEC confirmed) — a structure of "buying speed with the customer's money," a cheaper variant of Rule 8. But the warrant structure (30M shares for AMD, strike $23.47 — press-reported) creates the paradox where a rising share price detonates GAAP net losses through non-cash remeasurement (Q2 warrant/CVR loss of $1,045.5M, confirmed) — the cost of financing passed on to shareholders as dilution and volatility. The core basis of D1 is the bankruptcy history itself: a company that once collapsed on leverage in 2022 is expanding again on the same leverage pattern.

**GATE verdict:** G1 PASS (confirmed) — the +1,194% YoY colocation and the Q2 total-revenue +108.8% YoY reversal are the evidence. G2 (new markets' 10-year ceiling ≥ 10x current revenue, bar ≈ $4.4B) PASS (estimate) — the AI infrastructure TAM runs to the hundreds of billions (industry estimates), and contracted revenue of $24B+ alone is ~5.5x the bar (press-reported). The AMD option exercisable up to 2.5GW is the size of the upside option.

**The largest risk is "one tenant is the company":** 83% of revenue comes from a single customer, CoreWeave (Q2 2026, SEC confirmed). None of blitzscaling's 9 rules guarantees demand's permanence, yet CORZ's demand rests on a single decision. Easing customer concentration is this company's survival condition.

**Two-sided assessment (per the score file):** - **On the upside:** Colocation +1,194% YoY (SEC confirmed) — empirical proof of the second growth curve · gross margin 6.4% → 42.7% (confirmed), colocation-only margin 58.5% · contracted revenue $24B+ (~55x TTM, press-reported) + AMD 2.5GW expansion option — demand visibility · adjusted EBITDA +$41.1M per quarter (+44% YoY, confirmed) — operations in the black, unlike GAAP losses · diversification starting with AMD and Neocloud 529MW (deployments expected 2027, press-reported) · revenue per employee ~$1.35M ($440M TTM estimate ÷ 325 employees, press-reported/estimate) — lean-organization productivity · shareholder check working (acquisition voted down) — a positive governance signal. - **On the downside:** Capital deficiency — total debt $7.68B > assets $5.26B (SEC confirmed), long-term debt alone ~10x TTM · 83% single-customer dependence (CoreWeave, Q2 SEC confirmed) — chain risk on contract disruption · Chapter 11 history (2022) — empirical proof of leverage failure, repeating the same pattern · GAAP net loss of −$1,155.3M per quarter — non-cash (warrant/CVR remeasurement −$1,045.5M + mining-asset impairment −$266.5M, confirmed), but warrant-structure volatility is structural · AMD 529MW revenue contribution from 2027 (press-reported) — H2 2026 is an execution-risk window · −50% from the 52-week high of $30.46 — reflecting market doubt on execution (press-reported).

The directional assessment is **early Act 2 of the second blitz (execution sprint)** — Q2 2026 was the inflection point, and AMD 529MW (deployments starting 2027, press-reported) is the full-track entry point. The H2 2026 in between is the execution window: "the contracts are sold, the buildings are under construction."

2. Blitzscaling Lens

**Rule 8. "Raise Too Much Money" — burned once by that very money.** CoreWeave is the aggressive execution of Rule 8 (total debt $35B+), and for IREN the qualitative distinction of contract-backed financing was key. CORZ goes one step further — **a company that actually collapsed while executing Rule 8**. The 2022 Chapter 11 is Rule 8's paradoxical proof — "running out of money mid-race is fatal": it raised too much money, expanded on leverage, and when the cycle turned, the money consumed the company. ~$400M of debt extinguished (note-to-equity conversion), existing shareholders diluted 40% (press-reported) — Rule 8's "secure runway at the cost of dilution" was realized, in the end, as "paying off debt with dilution."

The current D1 −9 is this history's score. Spending $954.2M of capex per 6 months (confirmed) from a capital-deficiency position (total debt $7.68B > assets $5.26B, SEC confirmed) does not erase the suspicion of "the second run following the same pattern." The score file's v1.1 candidate 1 ("a D1 anchor for companies with bankruptcy history — CORZ −9 as the base case") aims at exactly this point. There are, however, two buffers now: cash of $1.77B (confirmed) and CoreWeave absorbing capex of $180.9M (confirmed) — "buying speed with the customer's money" is a cheaper variant of Rule 8. But the warrant structure (30M shares to AMD at a $23.47 strike, press-reported) creates the paradox where a rising stock price detonates GAAP net losses through non-cash remeasurement (Q2 warrant/CVR loss $1,045.5M, confirmed) — the cost of financing passed on to shareholders as dilution and volatility. Rule 8 said "raise too much money," not "raise it again the way you once collapsed."

**Rule 5. "Let It Burn" — the big fire is one tenant; the small fire is the GAAP loss.** In Hoffman's fire priority, CORZ's true big fire is that **83% of revenue is a single customer, CoreWeave** (Q2 SEC confirmed). If any one of CoreWeave's demand slowdown, contract renegotiation, or in-house DC shift materializes, the capital-deficiency balance sheet comes under the pressure of a second Chapter 11 — in the score file's words, "one tenant is the company." The small fire is the GAAP net loss of −$1,155.3M per quarter — mostly non-cash (warrant/CVR remeasurement −$1,045.5M + mining-asset impairment −$266.5M, confirmed), distracting but not lethal. The −50% discount from the 52-week high (press-reported) is also a small fire — a reflection of the market's execution doubt.

Rule 5's paradox works here too. The big fire (customer concentration) is not one CORZ can extinguish itself — only delivery performance can put it out, namely the AMD 529MW deployments starting 2027 (press-reported). And a company with many small fires gets no credit from the market when the big fire breaks out — the −50% discount from the 52-week high (press-reported) is the price of that absent credit. In the score file's phrase: "the contracts are sold, the buildings are under construction" — the promises are signed, the proof is still pending.

**The limits of Rule 7 — "Ignore Your Customers" only holds when there are many customers.** Rule 7 says "support customers only to the extent it doesn't slow you down." It holds on the premise of many customers, where some dissatisfaction can be absorbed as cost. **CORZ inverts this rule's premise.** When 83% of revenue is one customer, not the customer's "complaint" but the customer's single "decision" determines the company's survival. An un-ignorable customer — not just un-ignorable, but a customer whose disregard ends the company.

The genuinely applicable interpretation here runs the opposite direction. Rule 7's spirit is "reject unnecessary demands for the sake of speed"; what CORZ needs is "speed unswayed by a single customer's demands" — dispersing customers with long-term take-or-pay contracts like the AMD 529MW (15 years, 2.5% annual escalator — press-reported). The C2 4-point structural lock-in of the 15-year leases (up to 30 years with extensions) is the empirical proof of this prescription. AMD·Neocloud 529MW is still at the "prescription" stage with revenue contributions from 2027, but the prescription points the right way — the only way for the "landlord" to escape dependence on one tenant is to add tenants.

3. Competitive Landscape — Coordinates Inside the "Landlord" Cohort, and the Tenant's Irony

**The designated competitors** sit inside the small-neocloud/hosting cohort. APLD (contracted revenue ~$36B, 1.41GW+), WULF (Anthropic 20-year ~401MW, $19B), CIFR (Fluidstack + AI lab $9B+) — all companies of the same grammar: "power-and-land preemption → long-term take-or-pay → high-capex delivery" (provided reference). CORZ's distinction is the **"AI landlord" position with the chipmaker (AMD) itself as tenant** — a different species of blitz from CRWV·NBIS, which does not buy GPUs but sells power and sites. GPU-neutrality (hosting both NVIDIA and AMD) is another singularity.

**Scale coordinates:** ~1.1GW leased is small among the 7-candidate cohort (vs CRWV's multi-GW). Contracted revenue of $24B+ (press-reported) is ~55x TTM revenue (~$440M, estimate) — the cohort's highest ratio of demand visibility. But the structural risk of "the contracts are sold, the buildings are under construction" is common to the cohort — the gap between actually delivered volume and contracted volume, against 437MW billing at ≈ $635M annualized (press-reported), is this company's execution risk.

**The tenant's irony:** the tenant accounting for 83% of CORZ's revenue is none other than CoreWeave (CRWV 55/C) — "the tenant (CORZ 52/C's tenant) scores higher than the landlord (CORZ 52/C)." More sharply: CORZ's fate rides on CRWV's execution speed — if CRWV research's maximum risk ("the speed of borrowed money") materializes, its shock transmits immediately into CORZ's 83%. The landlord's and the tenant's leverage are entangled in the same direction — the point v1.1 candidate 2 ("a quantitative anchor for customer concentration — CORZ's 83% as the most extreme case") aims to formalize.

**The direction of creative destruction:** C1 (displaced order) 3 points — the business displacement from bitcoin self-mining to AI-dense colocation is underway (self-mining revenue −66% YoY, confirmed). But CORZ is less a "destroyer" than the "landlord" of the destroyers (CoreWeave, AMD) — indirect participation as an infrastructure lessor, one axis of the neocloud supply chain. C2 (irreversible behavioral change) 4 points — the structural lock-in of 15-year leases (up to 30 years with extensions), 2.5% annual escalators, long-term take-or-pay contracts. C3 (cost-structure change) 3 points — recycling stranded mining assets into HDC, evidenced by gross margin 6.4% → 42.7% (confirmed).

4. Inflection Points

**Inflection Point 1. Diversifying the 83% single customer — the speed of adding tenants.** What to watch: at the Q3 2026 earnings (expected Friday 10/23 — an estimate; no company announcement), whether CoreWeave's share of colocation revenue falls from Q2's 83% (SEC confirmed), and whether the AMD 529MW deployment progress (deployments expected 2027 — press-reported) becomes concrete. The speed at which the "landlord" escapes dependence on one tenant is this company's survival speed. AMD volumes actually go live from 2027 — until then the 83% structure persists. Link: the risk is already reflected in the score file's D1 −9 — debt service capacity itself is tied to a single customer's contract performance. Before v1.1 candidate 2 (a quantitative customer-concentration anchor) is formalized, this risk is not additionally reflected in the score — v1.0's structural blind spot, and the reason 52/C does not fully capture the "landlord's fate."

**Inflection Point 2. Delivery speed of contracts — "the contracts are sold, the buildings are under construction."** What to watch: whether the 437MW billing ≈ $635M annualized (press-reported) expands in Q3, facility utilization across 10→11 sites, and construction progress on the AMD 529MW deployment (starting 2027). The measured execution of the "convert all MW to HDC over 3 years" plan (10-K, confirmed) is the observation point. The discount factor of B2 (distribution scalability) 11 points — "no measured record of on-time delivery" — is what lifts or deepens here. Link: of the $954.2M/6-month conversion capex (confirmed), CoreWeave absorbs $180.9M (confirmed) — whether the "buy speed with the customer's money" structure holds is the barometer of capital pressure. The progress of the AMD warrant's 30M shares (strike $23.47, press-reported) vesting and exercise is also the observation metric for shareholder dilution.

**Inflection Point 3. Deleveraging from capital deficiency — the pressure of a second Chapter 11.** What to watch: whether the gap between total debt $7.68B and assets $5.26B (6/30, SEC confirmed) narrows each quarter, the repayment trajectory of long-term debt $4.30B (≈ 10x TTM ~$440M estimate), and the burn rate of cash $1.77B (confirmed). Whether the divergence between adjusted EBITDA +$41.1M per quarter (+44% YoY, confirmed) and GAAP net loss −$1,155.3M (non-cash-centered, confirmed) persists — observed together. Link: current D −9 is 6 points below the FLAG bar (15 points of deductions) — "not blitz-failing watch," but the cohort's highest deduction. From an 83% single-customer structure, a single contract disruption could trigger the watch flag — where the maximum risk overlaps.

**Observation checklist (quarterly):** ① easing of CoreWeave's share (83% — SEC confirmed) of colocation revenue ② expansion of the 437MW billing ≈ $635M annualized (press-reported) ③ AMD 529MW deployment progress (deployments expected 2027) ④ direction of the total-debt/assets gap ($7.68B vs $5.26B — confirmed) ⑤ burn rate of cash $1.77B (confirmed) ⑥ dilution progress of the AMD warrant 30M shares (press-reported). ①·②·③ are "execution" metrics, ④·⑤ are "firewall" metrics, ⑥ is a "shareholder" metric. If execution metrics trend up and the firewall holds, 52/C rises in the direction of the "recovery blitz" assessment. If a crack appears in the single tenant's contract — the capital-deficiency balance sheet comes under the pressure of a second bankruptcy. In a fallen blitzer's second blitz, failure costs more than the first.

5. One-Line Risk

In a capital-deficiency position with 83% of revenue riding on a single tenant (CoreWeave), if that tenant's execution speed bends, the AI landlord's second blitz could rewrite its first blitz (Chapter 11) with the same ending — that is the fundamental risk of the fallen blitzer.

*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

Nov 5, 2026 · Before open · Confirmed

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.