Thesis
LOWexiteddecaying
CoreWeave is the leading AI cloud infrastructure provider — NVIDIA GPU cloud, H200/B200 clusters, enterprise AI workloads. Capital-efficient GPU-as-a-service model competing with AWS/Azure/GCP in the AI training/inference layer. IPO'd March 2025, still underfollowed relative to hyperscaler peers.
Narrative: Jul 16 SEC thesis review: CRWV at approximately $77.25, current cycle -13.4%, 12.3% portfolio allocation, plus $72.27 previously realized loss. Demand is real: Q1 revenue grew 112% to $2.078B and unsatisfied RPO reached $98.8B. But the original 'capital-efficient GPU cloud' framing is disproven by the balance sheet: $25.1B principal debt at Mar 31, $7.5B current debt, $427M quarterly int
Next catalyst: Q2 earnings expected in August — revenue conversion from $98.8B RPO, gross margin, financing costs, capex, customer concentration, and any reduction in insider selling.
Max entry: 105.0 · Entry limit: 95.0 · Profit target: 15.0 · Timeframe: months
Research mentions — timeline (8)
newest first · undated last
Sep 30Camillo macro-timing event — Iced Coffee Hour ECU ends with a hard-dated segment "Is This 2001 Again?" (his capex-supercycle 2000/2001 analog call, same day as MU's print). Content unreviewed from this pod; treat as an attention/consensus event, not a signal. Late Oct: AMZN Q3 (AWS growth >30% watch), GOOGL Q3 (dead-thesis revisit condition: FCF/cloud monetization proof), CRWV Q3 (exit verdict stands).
2026-09-23Aug 25 DO NOW batch appears UNEXECUTED on broker truth (CRWV still held full; HOOD/NBIS/AMZN/VG at pre-trim size). Wash-sale stay-out expired → TLH intent still executable. Surfaced on the review board, not actioned.
Sep 11Re-entry, verified vs our own EDGAR pulls: CNBC — fund active again via OPTIONS on AMD/CRWV/BE/SNDK (capped downside, no PB margin dependency). Our Aug 28 SharonAI Schedule 13D (8.07M sh, ~22.5%, amended Sep 17 w/ Form 4) is part of this post-collapse re-accumulation — the SHAZ flag below is now sharper: he is actively rebuilding, not passively parked. Jul 2 Form 4 on SHAZ (EDGAR FTS) predates the collapse.
Aug 28Situational Awareness (Aschenbrenner, EDGAR primary): Q2 13F largest positions: SANDK $5.67B (2.50M sh), MU $5.57B (4.83M sh, ~$1,154 avg), NBIS $1.23B (4.46M sh — held, trimmed from Q1), BE $1.90B common + call, TSM $1.27B, CRWV $745M (kept), APLD $469M, IREN $433M, RIOT $468M, CORE $666M, STM $584M. TE: trimmed 10M -> 4.9M shares ($46.5M) — POSITION KEPT, half trimmed (resolves the standing next_moves open item: neither full validation nor death). NEW: active-intent Schedule 13D filed on SharonAI Holdings (SHAZ) — 8.07M shares (~22.5% of 35.8M shares out), amended Sep 17 with a Form 4. Research flag added below.
Aug 1413F caveat: the Q2 13F ($20.24B reportable, values Jun 30) predates the Jul 30 fire sale — MU/Sandisk/NBIS/CRWV/TE etc. holdings shown there were SOLD to Citadel. Do not read those as current positions. Next true read on Aschenbrenner's book: Q3 13F (~Nov 14) + his option filings.
Jul 16CRWV $87.59 — EXIT verdict stood (): leverage/concentration/insider-selling overwhelmed the $98.8B RPO; hard stop validated at $84 Jul 2. Dead for portfolio; -43% off high.
Altimeter (Gerstner, $9.8B): NVDA 19.2% top holding. NEW: Cerebras 16.2% ($1.60B), MU, Qualcomm, Applied Materials, LRCX ($171M), SpaceX ($303.5M). EXITED AVGO and AXON. Kept: CRWV 6.2%, ARM $582M, META $763M, TSM $683M, AMZN $578M, HOOD/SNOW core. Signal: rotation out of AXON/AVGO into compute-hardware picks-and-shovels. AVGO + AXON watchlist entries lost their initiating guru.
Portfolio lessons (no action, doctrine confirmation): (1) forced deleveraging manufactured part of the July-August AI drawdown; CRWV/Iron/SNDK rebounded 27-41% once Citadel absorbed the block — price damage from financing mechanics is not thesis death. (2) The highest-conviction AI-leverage operator converted to defined-risk options after nearly dying on margin — direct validation of Raj's defined-risk LEAP doctrine and Costello's leverage warnings from the Camillo debate transcript. (3) Correlation risk inside the AI-infra sleeve is real: his shorts did not hedge, everything fell together — diversification across AI-infra tickers is not risk reduction.
Read-only insight page · assembled Oct 19, 2026 04:21 PM ET · not a trade recommendation; execution only in the agentic account after Raj agrees to a specific trade.