Flagship · Bulletin
Wednesday, July 1, 2026
Regime
NEUTRAL
Cash
10.77%
Positions
17
Tickets
2
Macro rationale
Live inputs confirm NEUTRAL with a defensive lean: HY OAS 2.75 < 90d SMA 2.91 (credit tight) and 2s10s +0.36 (not inverted) argue against risk-off, but the extension escalator is ACTIVE (QQQ +14.3% >200d) and the Warsh Fed remains hawkish (CME Sep-hike odds 72.8%, dot median 3.8%), so cash sits toward the band top (12.4% vs 14% neutral target), not the floor. This is an event-driven patch, not a regime change: the midweek trigger fired on MU -10.12% (verified $1,032) and >15pt score drift in NVDA/ASML/AVGO. My read — consistent across four runs and reinforced by MU's $41.46B Q3 / $50B Q4 guide / ~$100B contracted SCAs — is that the compute drift is Warsh-Fed discount-rate multiple compression on duration, NOT a flow reversal. Underlying AI-capex demand ($725B 2026 hyperscaler line) is accelerating, so the named funders (EUV monopoly, 70% ASIC share, CUDA moat) are intact. Trimming intact franchises into rate-driven compression would be the exact flow-chasing over-trim this book exists to avoid — so the compute names are HOLDs. The two scoped defense actions are risk right-sizing, not thesis exits.
Thesis
Event-driven patch, scope-locked to [ASML,AVGO,KTOS,MU,NVDA,TSM,ANET,FN,HEI,PSN]; all other names held at delta 0 per the trigger. The midweek gate fired on MU -10.12% (verified $1,032) and >15pt score drift in NVDA/ASML/AVGO. My core read — consistent across four runs and confirmed by MU's own blowout ($41.46B Q3, $50B Q4 guide, ~$100B contracted SCAs) — is that the compute drift is Warsh-Fed discount-rate multiple compression on duration, NOT a flow reversal. The named funders are all intact: EUV monopoly (ASML, GPFG holds, Terafab signal), 70% ASIC share (AVGO, OpenAI Jalapeño launch), CUDA moat + $725B hyperscaler capex (NVDA), foundry monopoly (TSM, Tesla AI5). MU's -10% was rotation with no company news. So all seven compute names are HOLDs — trimming intact franchises into rate-driven compression is the flow-chasing over-trim this book exists to avoid. The two actions are defense risk right-sizing, not thesis exits: KTOS trimmed -2pp (to 8%) into today's +7.2% strength — an outsized passive-drift overweight on an unconfirmed revenue inflection, thesis (MACH-TB 2.0, UK £5B drone) still intact; HEI trimmed -2pp (to 9.88%) as a partial profit-trim honoring the 54x>50x P/E invalidation flag and reducing concentration near the 12% cap. PSN held (locked, unlock=none — close refused). The ~4pp freed lifts cash to 12.4%, toward the 14% neutral target with the extension escalator active. Book stays at 17 (scope forbids new opens).
Reflection
Compute score drift (NVDA/ASML/AVGO) is Warsh-Fed discount-rate compression, not demand break — MU's $100B-SCA blowout confirms the AI-capex flow is intact; hold the franchises.
Fourth straight run where the midweek trigger tempts a reflexive compute trim and the disciplined answer is hold. MU -10% on zero company news, into a $50B guide, is the cleanest evidence yet that the drift is duration/discount-rate, not flow. What I'm still sitting with: the book is stuck at 17 vs the 18-20 band and every event patch scope forbids the open that would fix it — that gap must close at the next full Sunday deep. Also KTOS/HEI both drifted to ~10-12% via appreciation on speculative/rich theses; I keep right-sizing them 2pp at a time rather than making a clean sizing decision, whic
Positions (17)
- MUMicron Technologycomputehold4.29%79flow steadyconf
Funder
US HBM3e/HBM4 supplier in a capacity-disciplined memory cycle, sold out through 2026-27. Trigger: blowout Q3 $41.46B vs $36B est, $50B Q4 guide, 16 multi-year Strategic Customer Agreements worth ~$100B contracted revenue; underpinned by the ~$725B 2026 hyperscaler capex line and Anthropic supply pact.
Unwind
SK Hynix $29B capacity raise floods memory and breaks HBM pricing discipline; contracted SCAs cancelled or renegotiated lower; DRAM cycle rolls with FCF turning negative.
Catalyst
Q4 print vs the $50B guide; HBM4 qualification milestones with hyperscaler customers.
- TSMTaiwan Semiconductor Manufacturingcomputehold7.43%83flow steadyconf
Funder
Leading-edge foundry monopoly (<3nm), >50% gross margin, prodigious FCF, disciplined capex — routes the world's AI silicon for decades. Goldman's $920B-$1.4T 2027 AI-capex forecast names TSMC explicitly; CHIPS $6.6B Arizona disbursement locked; Tesla AI5 manufacturing win.
Unwind
Gross-margin compression toward 40s, competitor (Samsung/Intel) leading-edge yield parity, or decelerating advanced-node bookings against stretched valuation.
Catalyst
Next monthly sales print and Q earnings advanced-node mix / capex guide.
- NVDANVIDIAcomputehold3.51%72flow softeningconf
Funder
CUDA/accelerator software moat, ~75% gross margin, fortress balance sheet. ~$725B 2026 hyperscaler capex anchored on Blackwell/Rubin (MSFT $190B, AMZN $200B, GOOGL $180-190B, META $125-145B); Stargate $500B JV. Score drift is rate-driven multiple compression, not a demand break — MU's blowout confirms the AI-infrastructure flow is intact.
Unwind
Hyperscaler ASIC in-sourcing (GOOG TPU, META MTIA) durably erodes training/inference TAM share; a hard hyperscaler capex cut; margin break below 60%.
Catalyst
Next earnings data-center revenue and Rubin ramp commentary; hyperscaler capex guides.
- ASMLASML Holdingcomputehold3.52%63flow softeningconf
Funder
EUV lithography monopoly — the single most irreplaceable tool in advanced semis, ~50% gross margin, multi-year backlog, no commercial-scale alternative. Norway GPFG holds, Coatue/Third Point added Q1; TSMC/Samsung 2025-27 EUV orderbook; Terafab JV signal ($55-119B fab capex, Jun-30) flags fresh long-cycle demand.
Unwind
EUV book-to-bill breaks below 1 on order deferrals; litho order cancellations from a fab-capex retrenchment; conviction stays sub-70 into a second quarter with bookings deterioration.
Catalyst
Q bookings/book-to-bill print; TSMC/Samsung capex confirmation.
- AVGOBroadcomcomputehold2.72%62flow softeningconf
Funder
Custom-silicon ASIC leader (~70% share) + VMware software, huge FCF, $7B+ annual buyback, serial dividend grower. GOOG TPU multi-gen, META MTIA, Anthropic 3.5GW ASIC contracts; new inference-focused Jalapeño chip for OpenAI (Jun-30) expands the hyperscaler relationship set; Q3 AI revenue +200%.
Unwind
Hyperscaler full in-sourcing of custom silicon, custom-chip revenue stall, or a break in the $30B+ AI bookings run-rate; score sustained below 60.
Catalyst
Next earnings AI-segment revenue and ASIC bookings; OpenAI/Jalapeño ramp.
- ANETArista Networkscomputehold2.34%67flow steadyconf
Funder
Durable hyperscaler-networking franchise, share-gainer since 2004, ~65% gross margin, net cash, single-image EOS software moat, Ullal-led capital discipline. Microsoft & Meta named as largest customers funding 35% YoY revenue growth; Morgan Stanley PT raised to $190, Overweight.
Unwind
Nvidia Spectrum-X displaces Arista in AI back-end fabrics; Microsoft/Meta concentration creep reverses into in-house switching; gross margin breaks below 60%.
Catalyst
Q2 bookings for Microsoft/Meta concentration read; 800G Etherlink adoption.
- FNcomputehold1.66%63flow softeningconf
Funder
Fabrinet — optical transceiver/packaging manufacturer interconnecting hyperscaler AI clusters; capital-light, debt-free, extracts hyperscaler capex without silicon R&D risk. AWS dominant customer drove ~35.9% YoY growth; 800G/1.6T demand from NVDA/AMD GPU clusters. Q3 rev $1.21B and EPS $3.72 both beat.
Unwind
Datacom sequential revenue decline; hyperscaler in-sources optical packaging; 800G/1.6T pricing power evaporates and margins compress.
Catalyst
Next quarter Datacom sequential trajectory post-earnings overhang.
- CCJCamecoenergyhold7.51%65flow steadyconf
Funder
Tier-1 Saskatchewan uranium franchise + Westinghouse 49% JV with Brookfield; commodity tail-risk hedge slot. DOE $17.5B AP1000 loan program (CCJ owns 49% Westinghouse); utility contract book $60+/lb vs term market ~$93/lb; Kazatomprom 8Mlb output cut sustains structural deficit; AI-datacenter nuclear PPA rush.
Unwind
Reactor build cancellations, uranium term-price collapse below the contract book, or a cost-structure breach in Saskatchewan operations.
Catalyst
AP1000 loan disbursement milestones; uranium term-contracting print.
- GEVGE Vernovaenergyhold8.93%63flow steadyconf
Funder
Gas-turbine + grid franchise with pricing power — ~$163B backlog sold through 2028, 100GW turbine backlog, 10-20% price realization. Chevron-Microsoft 20-yr 2.7GW Project Kilby gas deal names GE Vernova directly; Norway GPFG added post-spin; $17.5B DOE nuclear loan supports the nuclear-services leg.
Unwind
Turbine order-book break or backlog cancellations; AI-load power demand slows; margin execution slips on the grid segment; conviction stays sub-65 into a fresh order print.
Catalyst
Next order-book disclosure; Project Kilby revenue-recognition timing.
- SCCOSouthern Copperenergyhold6.95%63flow steadyconf
Funder
Lowest-cost copper producer, vast reserves, fat margins, large dividend; commodity tail-risk equity hedge (§5.3/10.2). Copper at all-time highs >$6.50/lb (vs $5 entry trigger), supply deficit ~320k tonnes; AI/data-center ~30% of new copper demand by 2030 (BloombergNEF); BHP confirms copper margin leadership.
Unwind
Copper falls back toward the cost curve on a demand-collapse/recession; Peru/Mexico permitting or tax shock breaks the low-cost position.
Catalyst
Copper spot trajectory; Tia Maria/Los Chancas project milestones.
- MPMP Materialsenergyhold6.14%flow steadyconf
Funder
Only scaled US rare-earth + magnet producer, DoD-backstopped moat vs China's ~98% magnet control. DoD $400M Series A preferred + $150M loan + 10-yr NdPr price floor ($110/kg) + 10-yr magnet offtake for the 10X facility; China MOFCOM control-list addition (Jun-22) paradoxically validates strategic value; Q1 record production.
Unwind
FCF stays negative with the magnet-segment ramp stalling; DoD offtake/price-floor commitments unwound; NdPr pricing collapses despite the floor.
Catalyst
FCF verification at next print; 10X magnet facility ramp; MOFCOM cliff (Nov-2026).
- NEMNewmont Corporationenergyhold5.02%89flow softeningconf
Funder
Tail-risk equity hedge (§5.3/10.2): largest gold producer, tier-1 long-life assets, strong post-Newcrest FCF and dividend, improving cost discipline. Central-bank gold accumulation; structural hedge against a hawkish-Warsh policy-error/stagflation tail; C-suite promotions signal operational continuity.
Unwind
Gold sustains a deeper bear-market drawdown as the Fed put stays out-of-the-money and Hormuz risk fully deflates; cost blowout compresses miner margins as gold rolls over.
Catalyst
Gold spot vs the Jan peak; Fed July FOMC (Jul 28-29) path.
- KTOSKratos Defensedefensetrim10.01%-2.05ppflow steadyconf
Funder
Tactical drones (XQ-58 Valkyrie), hypersonics, turbine engines — DAWG-adjacent defense-tech. DoD MACH-TB 2.0 $1.45B IDIQ ceiling, Project Helios $68.3M, USMC Valkyrie mods; UK £5B drone-investment program and ARK $22M accumulation (Jun-30); FY27 DAWG line ~$54.6B. Verified +7.2% to $53 today on the UK/ARK flow.
Unwind
Revenue-inflection/margin-firming trigger fails to confirm at next print; DAWG/drone appropriations cut in reconciliation; passive-drift overweight not supported by fundamentals.
Catalyst
Next earnings for revenue-scale/margin confirmation; FY27 DAWG appropriations markup.
- HEIHEICOdefensetrim12.00%-2.00pp68flow steadyconf
Funder
Defense/aero aftermarket compounder — Mendelson family owners since 1957, 20%+ FCF CAGR, disciplined serial acquirer ($1B+/yr). Q2 beat (FSG $929M vs $864M, ETG $460M vs $396M, +10.7%); NDAA mandatory funding + commercial-aerospace MRO. Textbook 20-30yr franchise — trim is concentration/valuation right-sizing, not a thesis exit.
Unwind
Forward P/E sustained above the 54x/50x invalidation threshold with organic growth decelerating below the ~6% floor; a value-destroying acquisition; family capital-allocation discipline breaks.
Catalyst
Next quarterly organic-growth print vs 6% floor; acquisition-pace tape.
- PSNdefensehold2.52%50flow softeningconf
Funder
Parsons — C5ISR/cyber government-services prime, sticky multi-year IDIQ backlog. MDA SHIELD $151B-ceiling IDIQ, DTRA cyber $137M, USAF MATOC $15B, Navy WEXMAC $10B — all prime awards per Parsons IR; $1.1-1.4B Blatnik Bridge win (6/16) + $500M cyber award. Held at reduced weight inside conviction lock (unlock=none).
Unwind
Book-to-bill sustained below 1 with the confidential-contract softness spreading; MDA SHIELD task-order conversions fail to materialize; score stays sub-60 past the lock.
Catalyst
Next earnings book-to-bill and MDA SHIELD task-order conversion.
- LLYEli Lillybiologyhold2.50%70flow steadyconf
Funder
Premier pharma compounder — Mounjaro/Zepbound GLP-1 franchise (60%+ share) + retatrutide Ph3 extends a multi-decade runway, deep margins, strong capital return. Norway GPFG and large AM holders; EMA favorable opinion for Jaypirca (Jun-26 all-time high). Profitable, insulated from the rate vise.
Unwind
Tirzepatide loses >10pt US GLP-1 share; oral-GLP-1 competitor disruption; P/E overstretch with pipeline setback; IRA Part-D pricing shock.
Catalyst
GLP-1 share prints; retatrutide Ph3 readouts; IRA negotiation updates.
- VRTXVertex Pharmaceuticalsbiologyhold2.18%76flow steadyconf
Funder
CF monopoly (Trikafta) cash engine, self-funded, high margins; Casgevy gene therapy + suzetrigine (Journavx) non-opioid pain optionality. Norway GPFG/large AM holders; Casgevy pediatric Ph3 (Jun-11) hit all primary endpoints in sickle cell/thalassemia — structural commercial uptake on plan.
Unwind
CF franchise erosion (payer/patent), Casgevy commercial uptake stalls, suzetrigine launch disappoints; valuation flag with pipeline miss.
Catalyst
Suzetrigine launch traction; Casgevy reimbursement expansion.
Warnings
- tilt_truncated: biology tilt -5pp truncated at 0% cap
- §2 cap re-applied after pillar tilt: HEI
- position_count: 17 < 18 after mechanical closes (§2 floor) — book under-populated; strategist must open replacements next run
- ledger stale (ledger over-bought (cash -38.38% of NAV)) — rebalance diffed against intent
Conviction-lock actions
Cost breakdown
- $0.0952
B1unknown
10 calls · in 19.4k · out 2.5k
- $0.0440
B2unknown
4 calls · in 4.7k · out 2.0k
- $1.0106
Cunknown
2 calls · in 106.0k · out 16.0k · cache-read 11.9k · cache-write 11.9k
Per-call cost computed from cost.js pricing constants (Opus 4.7, Sonnet 4.6, Haiku 4.5). Cache-read tokens billed at 10% of input. See Cost & ROI for the rolling 30-day ledger.