Same Industry, Different Leverage: CoreWeave vs IREN
One point apart — 55 vs 56. The anatomy of that point is a lesson in how the model treats borrowed speed.
5 min read · 2 companies · Head-to-head scoring · Not investment advice
One point
CoreWeave: 55/C. IREN: 56/C. Same industry — neocloud, GPU infrastructure for the AI buildout. Same thesis — contracts signed, buildings being built, revenue recognized later. The Blitz Index separated them by exactly one point, and where that point comes from is worth studying.
"An A-grade business dragged to C by its balance sheet."
"The crossover quarter: AI Cloud revenue passed mining."
Where CoreWeave wins: the business
On business quality alone, CoreWeave is ahead on every axis the model measures before deductions. Speed 21 vs 17: quarterly revenue of $2.575B growing 112% year over year, a $104.2B backlog up 246%. Moat 37 vs 34: larger scale, faster growth, a deeper customer roster. Speed converts to 84.0 on the 100-point scale — among the highest readings in the entire coverage set. If the Blitz Index scored only the business, CoreWeave would be a B and the comparison would not be close.
IREN's business is younger and smaller — the quarter where AI Cloud revenue ($70.5M) first passed mining ($66.7M) was Q4 FY2026. But youth is not the story here. The balance sheet is.
Where IREN wins: the balance sheet
Deductions: −13 vs −6. That seven-point gap in D is the entire one-point story, with room to spare — it overcomes CoreWeave's four-point business lead and then some.
CoreWeave raised more than $35 billion in debt in the fifteen months after its IPO. Annualized interest — roughly $2.5B — runs about five times adjusted operating income. Q3 guidance put interest at $860–940M against adjusted operating income of $200–260M. The D1 deduction of −8 fires on this arithmetic directly.
IREN's capital is contract-backed project financing: the $9.7B Microsoft and $3.4B NVIDIA contracts fund the build, with $13.8B in committed capital against recognized revenue. Its D1 is −4 — a warning, not a verdict.
Blitzscaling's Rule 8 says raise too much money — and the model rewards it in A2 (capital velocity). But v1.0 draws a line the book never had to: equity dilution is forgiven; debt that compounds faster than delivery is not. Raising is speed. Borrowing is a claim on a future that may not arrive. D1 is where that distinction lives.
The leveraged blitz
CoreWeave is the leveraged blitz made concrete: speed bought with borrowed money. The score's verdict is not that the business is bad — 21/25 on speed says the opposite — but that the capital structure charges rent on the growth. Every quarter the $104B backlog takes to convert into recognized revenue, the interest compounds. The thesis and the risk are the same number with opposite signs.
Honesty requires the other side. IREN's leverage is real too — it just lives in commitments rather than drawn debt: $13.8B committed is 19.5× revenue, and 52.7% dilution in fourteen months is Rule 8 taken to its edge. Meanwhile CoreWeave's concentration is genuinely easing (largest customer from ~85% to ~35% of backlog, press-reported) and the contracts are signed with counterparties that pay. Neither company is safe; they are unsafe in different ways, and the one-point gap is the model's attempt to price the difference.
The backlog clock
The $104.2B backlog is the most important number in either company's story — and the least scored. v1.0's gates pass on "contracts and market size," but no single item measures how fast signed contracts become recognized revenue. For CoreWeave, that conversion rate is the entire investment case: data centers must be built, chips installed, capacity delivered — all while interest accrues quarterly.
The securities class actions filed in early 2026 allege the company overstated its ability to meet demand and concealed construction delays — charges that map precisely onto the B2 discount already in the score file. Whether the allegations have merit is for courts to decide; what matters for the model is that the market is pricing the backlog-to-delivery gap as the central risk. IREN faces the same physics at smaller scale: its Horizon 2–4 data centers must deliver within the year, and its $13.8B in commitments is 19.5× its revenue.
What v1.1 should study
This pair exposes a gap in v1.0 itself. The model scores recognized revenue and penalizes drawn debt — but neocloud risk actually lives in the space between: $104B signed, buildings unbuilt. The backlog is the thesis and the interest is the clock, and right now the model has no formal rule for the conversion rate between them. A v1.1 "commitment stack" criterion — already under open review — would score backlog-to-delivery velocity directly. CoreWeave vs IREN is the case study for why it matters.
Method: Blitz Index v1.0 (frozen Oct 2026). CRWV 55/C and IREN 56/C scored 2026-10-12/13 at T=2026-10-11. Customer-concentration figures are press-reported.
Not investment advice. A comparative methodology study. It contains no buy, sell, or hold opinion on any security.