Below the Line: An Atlas of Explosions Under 65
65 is the Blitz Index success line. We judged all 24 companies beneath it — and the territory below the line is not empty. 16 of 24 are exploding in some form.
In the first Atlas of Ignition Points we charted 14 companies above 65 — all provably ignited, 12 of 13 ignition windows pairing revenue acceleration with margin expansion at the same time. The natural question is what the territory below the line looks like. This is the second volume: every listed company in the 2026Q4 scored universe under 65, judged one by one — 24 in total.
The headline finding is a reinterpretation of the line itself. The difference between the two volumes is not whether the companies exploded. It is the level of qualitative proof. In volume one, the signature was acceleration and margin expansion occurring together. In volume two, the margins of the ignition group are unproven: Nebius OM −115.5%, CoreWeave diluted to −0.9%, Applied Digital −38.7%, TeraWulf −110.5%. The 65 line is closer to this: not the line where explosions start, but the line where explosions get proven by margins.
Method
Sources: SEC EDGAR XBRL financials (the corrected 2026-10-14 financial store), the 2026Q4 score files, and Q3 2026 research where available. No external collection. CBRS and FRMI have no financial store data (zero revenue / no XBRL frames) — score-file figures are used and labeled as such. Coverage: descriptive only. It makes no predictions. And crucially — a business explosion is not a stock explosion. Our H1 rejection (post-crossing +1y median −5.9%) and H3 rejection (ρ(Δ, forward 3y) = −0.091) already showed that score-derived signals cannot time prices. The windows below are records of when a business changed trajectory, nothing more.
Verdict categories (designed for this volume): ignition (explosion underway, proven by recognized revenue) · pre-ignition (contracts and backlog accumulate, revenue and margins not yet proven) · post-ignition cooling (a decisive window proven in recognized revenue, now cooled) · fake explosion, in decay (explosion accompanied by margin collapse, per the WeWork rule from volume one) · second-curve waiting (first business mature, next window not yet converted to revenue). Composite verdicts are allowed.
The full distribution
Ignition 12 · pre-ignition 4 · post-ignition cooling 6 · fake explosion 1 · second-curve waiting 1 (pure) — with TSLA, DBX, and ABCL carrying composite verdicts (cooling + waiting).
| Company | Score | Verdict | Decisive window | Evidence | Confidence |
|---|---|---|---|---|---|
| NBIS | 59/C | ignition | 2024–26 | FY2025 $529.8M (+479%) → TTM ~$1.36B, ARR $3.0B, backlog $40B+ | High |
| CBRS | 59/C | ignition | 2022–26 | $24.6M → FY2025 $510M (+76%), RPO $25.4B (~37× annual revenue) — no finData | Medium |
| SEI | 58/C | ignition | 2024–26 | TTM $762.1M (+70.5%), Power Solutions H1 +129%, OM 21.8% | High |
| TSLA | 57/C | cooling + waiting | 2018–22 (first) | +82.5%→+51.4%, OM −13.9→+16.8 → FY2025 first annual decline (−2.9%) | High |
| IREN | 56/C | ignition (second-curve type) | 2025–26 | AI Cloud +687% YoY, Q4 mining revenue crossed below AI Cloud | High |
| ORCL | 55/C | ignition (second-curve type) | 2025–26 | OCI +121%, RPO $638B→$664B, revenue +30% (Q1 FY27) | High |
| CRWV | 55/C | ignition | 2023–26 | FY2025 $5.13B (+168%) → TTM ~$7.6B, backlog $104B | High |
| SYM | 54/C | cooling (decelerating) | 2020–23 (first) | $92.1M → $2.25B (24×) → growth 52%→26%→+22% | High |
| CORZ | 52/C | ignition (second-curve type) | 2025–26 | colocation +1,194% YoY, 83.3% of consolidated revenue, Q2 +108.8% reversal | High |
| APLD | 51/C | ignition (entering, second-curve type) | 2026– | Q1 FY27 $341.9M (+322% YoY), contracts $36B/$86B | High |
| ASTS | 51/C | pre-ignition | not started | backlog ~$1.3B, 60+ MNOs vs $0 commercial D2D service revenue (delayed to early 2027) | Medium |
| HIMS | 49/C | cooling (decelerating) | 2020–24 (first) | $82.6M → $2.35B (28×), OM −90.1→+4.5 → Q2 +38.2% | High |
| IONQ | 47/C | ignition — acquisition-led | 2024–26 | FY2025 $130M (+202%), RPO $485M (+297%) — organic growth ~0 | Medium (organic unproven) |
| SERV | 45/C | fake explosion, in decay | 2025–26 | revenue +404% but GM −570%, guidance cut −63% | High |
| ABCL | 43/C | cooling (first decayed) + waiting | 2020–22 (first) | $485.4M (Lilly royalty) → $38M collapse in 2023 | High |
| RDW | 43/C | ignition — acquisition-led | 2025–26 | TTM ~$426M, backlog $542M all-time high — organic unproven | Medium (organic unproven) |
| WULF | 41/C | ignition (entering, second-curve type) | 2025–26 | HPC lease $0→$31.9M/quarter, Anthropic 20-year ~$19B contract | Medium (total revenue still falling) |
| TLN | 39/D | ignition (second-curve type) | 2024–26 | revenue $2.1B→$2.6B (+22%), AWS PPA 1,920MW — acquisition mix | Medium |
| LEU | 35/D | pre-ignition (conditional) | not started | backlog $3.7B (7.8× TTM) vs FY2025 revenue +1.5% | Medium (expansion funding incomplete) |
| DBX | 34/D | cooling + waiting | pre-2018 estimated (first, out of range) | revenue $2.52B (−1.1%), Dash AI rolling out, no revenue yet | Medium (first-window data thin) |
| CIFR | 32/D | pre-ignition | not started | contracted ~$11.4B vs recognized ~$191M, Black Pearl first delivery began | Medium |
| FRMI | 31/D | pre-ignition | not started | TensorWave lease $6.5B/15y (SEC-confirmed) vs $0 revenue — no finData | Medium (deepest confidence discount) |
| COIN | 58/C | cooling | 2020–21 (first) | FY2021 $7.84B (OM 39.2%) → FY2025 $7.18B, TTM −10.4% | High |
| LBRT | 30/D | second-curve waiting | not started | core −7% (fracking cycle), LPI power 400MW (2027 delivery) | Medium |
Seven observations
1. Sub-65 is not "the absence of explosions." 16 of 24 are exploding in some form. The case that looks like a counterexample — SEI at 58 with OM 21.8%, the only ignition company with proven margins — needs its explanation stated: single-customer concentration (47.3%) and the capital intensity of the power business hold its score down. Margin proven ≠ high score.
2. The WeWork rule works here too. The volume's only fake explosion, SERV (GM −570%, guidance cut −63%), has the WeWork structure — margin collapse plus a wound to demand credibility. The volume-one signature ("acceleration defines the explosion; the margin's direction judges real from fake") holds. Nebius and CoreWeave's margin dilution stays inside the N5 "normal" range — dilution is not collapse.
3. A type new to this volume: the acquisition blitz (IONQ, RDW). Volume one had no such type. Here revenue explosions are evidence of acquisitions, not organic growth — IONQ's 10-Q MD&A names acquisition-driven revenue growth outright; RDW grows overall while its legacy space segment shrinks. Neither the WeWork rule nor N11 applies. We register it as an observation item, not a hypothesis — the definition is too immature — and reclassify the moment organic/acquisition split disclosures appear.
4. The N11 boundary: infra pivots don't show the margin signature. Volume one's three platform transitions (AppLovin, Palantir, MongoDB) each carried a GM +15pp-class margin conversion. Volume two's infra pivots (CORZ, WULF, APLD, IREN, CIFR — mining to datacenter) show none yet. That is not a refutation of N11 but a scope limit: N11's "platform transition" means software's margin-structure conversion; the crux of a capex-heavy infra pivot is not margins but contract-to-revenue conversion speed. CIFR (60× contracted-to-recognized) → WULF (in-segment explosion) → CORZ (consolidated reversal) is the spectrum of that speed.
5. Explosions are events, not states. The six cooling companies (TSLA, COIN, SYM, HIMS, ABCL, DBX) share one trait: their first explosions were clearly proven in recognized revenue. Explosions happen, then cool. And the score knee of the cooling group descends — TSLA 67 (2021) → 57 (2026), COIN 67 (2021) → 58 (2026) — symmetric with volume one's rising knees during ignition windows.
6. Volume two is the AI-capex cluster's second wave. 9 of the 12 ignition companies are AI infrastructure or power (NBIS, CBRS, IREN, CRWV, CORZ, APLD, WULF, SEI, TLN) — consistent with volume one's "2023–25 AI capex cluster." If volume one held the cluster's first wave (NVDA, ANET, APP, PLTR), volume two is the second wave that met the same trigger: the first wave proved itself in margins, the second is still proving.
7. Pre-ignition is judged by contract enforceability × credibility. ASTS ($1.3B backlog), LEU ($3.7B, 7.8× TTM), CIFR ($11.4B, 60× recognized), FRMI ($6.5B, $0 recognized). FRMI shows the widest gap: contract enforceability is SEC-confirmed (a-grade) while demand credibility is d-grade (anchor tenant termination, pending lawsuit, founder dismissal). The judging order is (a) enforceability → (b) revenue conversion timing → (c) demand credibility — and when (c) breaks, (a) becomes meaningless.
What comes next
The four pre-ignition verdicts (ASTS's 2027 commercial start, LEU's expansion funding, CIFR's revenue conversion, FRMI's credibility) and the margin proof of the twelve ignition companies get refreshed on each quarterly rescore. IONQ and RDW reclassify when organic/acquisition disclosures appear. A merged atlas of both volumes (v3) waits until two quarterly rescores accumulate.
The boundary
Every measurement here is retrospective, and no forward predictive power has been established — as H1 and H3 already taught us, business ignitions and price timing are different problems. Pre-ignition companies in particular tend to have expectations priced in early. This atlas is not a buy signal for anything; the verdicts are retrospective descriptions.
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; excluded or unlabeled items carry their reasons in the source study. This atlas is volume two of the explosion-atlas track (volume one: “An Atlas of Ignition Points,” research article 17) and reports only technical statistics, never signals.