House View

9 of 9 live theses

Our live investment theses — each one tracked against explicit kill-signals through the research pipeline, and re-dated as the evidence changes. If a kill-signal trips, we say so.

Scoreboard: daily (2026-11-20) · Kill-Signal Sentinel: weekly (2026-11-15) · Beating benchmark: 5/9 scored
Market tape check — where the AI surplus is landing 2026-11-20
Surplus migrating downstream75% OF LAST 20 SESSIONS
Physical
-17.7%
fabs · litho · memory
Compute infra
-16.2%
GPUs · networking
Application
-4.8%
AI-native apps
Token price
+26.8%
median, per 1M tokens
SPREAD HISTORY →
One-month moves. The tell we watch: if the application layer keeps outrunning the physical layer while token prices fall, the market is saying the efficiency surplus is escaping the chokepoints — the fiber-optic scenario our token-cost thesis is built to catch early.

Application vs Physical — the surplus spread

Smoothed one-month return spread, application basket minus physical basket, in percentage points · 2026-04-30 → 2026-11-20
17.5pp +27pp -28pp Apr 30Jul 2Sep 1Nov 20
Above amber line: application outrunning physical (surplus escaping downstream) Below green line: chokepoints being paid Dashed: even
The top 3 live theses are open in full below — exact basket tickers, entry logic, and the kill-signals that would change our mind.
Unlock all 9 active institutional theses, daily kill-signal alerts, and the 70/30 Portfolio Builder with Tradestie Premium.
Unlock all theses free for 14 days → Explore 70/30 Portfolio Builder →
Late-Cycle Rotation: Overweight Healthcare, Financials & Energy While Tech Corrects
As technology and semiconductor momentum decelerates (XLK and SMH flow scores dropping below 32), institutional capital is rotating aggressively into…
HIGH CONVICTION bullish Published 2026-10-15 3 kill-signals sector-rotationlate-cycleinstitutional-flow
▼
Since published (2026-10-15): basket -0.8% vs SPY +9.8pp vs SPY +9.8pp last 20 trading days: -1.7%
Avoid basket (SMH, IGV): -22.4% (L/S spread +21.6pp) — lagging long basket as intended
Per-ticker performance (5)
TickerEntry dateEntry pxLast pxReturn
LLY2026-10-151251.611347.11+7.6%
REGN2026-10-15796.34920.69+15.6%
JPM2026-10-15335.53310.53-7.5%
GS2026-10-15891.28790.69-11.3%
XOM2026-10-15174.68159.99-8.4%

The Thesis

As technology and semiconductor momentum decelerates (XLK and SMH flow scores dropping below 32), institutional capital is rotating aggressively into Healthcare (XLV, Flow Score 91), Financials (XLF, Score 100, ADX 36.4), and Energy (XLE, Score 100).

Rather than fighting the Tech drawdown (VXN-VIX spread elevated), our quantitative rotation model overweights cash-generative leaders in Biopharma (LLY, REGN), Money-Center Banks (JPM, GS), and Integrated Energy (XOM) where 1-month relative strength vs. SPY is +3.8% to +6.6% with positive Chaikin Money Flow (CMF) breadth.

Positioning & Sizing

  • Long Basket: Equal-weight LLY, REGN, JPM, GS, XOM (or via 70% Stock / 30% Low-IV Call overlays in the Portfolio Strategy Builder).
  • Avoid / Underweight: High-multiple software (IGV) and extended semiconductor beta (SMH) until sector CMF breadth crosses back above 50%.
KILL-SIGNALS — WHAT WOULD CHANGE OUR MIND
  • XLV and XLF 20-day relative strength vs SPY both turn negative (< -1.5%) while XLK Flow Score reclaims > 65
    Currently XLV +6.57% rel 1M, XLF +3.77% rel 1M; XLK at 31.3
  • High-Yield Credit Spreads widen > 450 bps, signaling systemic credit stress for Financials (XLF)
    Credit spreads remain contained; bank net-interest margins resilient
  • WTI crude breaks below $62/bbl for 10 consecutive sessions, invalidating Energy cash-flow support
    Energy free-cash-flow yields remain >8% at current strip
Sector Peak Rollover: Asymmetric 2x Put Spread + 1x OTM Call on Overbought Stocks UPDATED
Shorting overbought stocks (RSI = 68) outright or with naked puts fails when a sector stages a blow-off squeeze (44.7% win rate, -7.2% avg ROI on…
HIGH CONVICTION bearish Published 2026-11-02 2 kill-signals options-strategysector-rolloverasymmetric-short
▼
Since published (2026-11-02): basket -0.7% vs SPY +7.6pp vs SPY +7.6pp last 20 trading days: -0.7%
Per-ticker performance (5)
TickerEntry dateEntry pxLast pxReturn
MCD2026-11-02265.64273.24+2.9%
PEP2026-11-02140.42140.54+0.1%
KO2026-11-0295.8396.99+1.2%
CRH2026-11-0296.5988.05-8.8%
CL2026-11-0297.4398.71+1.3%

The Empirical Edge (1,244-Trade Backtest)

Shorting overbought stocks (RSI >= 68) outright or with naked puts fails when a sector stages a blow-off squeeze (44.7% win rate, -7.2% avg ROI on pure put spreads). However, pairing 2x Near-the-Money Bear Put Spreads (~0.98x / 0.88x Spot) with 1x Out-of-the-Money Call (~1.10x Spot) at 35-45 DTE inside Peak Rollover Sectors (Sector MFI >= 56 with negative 5-day flow change) lifted historical win rates to 75.0% with +24.7% average ROI (+42.1% median ROI).

Why Sector Rollover Matters

  1. No Sideways Theta Bleed: When an entire sector ETF (XLY, XLP, XLB) reaches peak money flow (MFI 56-60) and 5-day institutional inflows roll over (-9 to -17 flow points), stocks in that sector rarely stay flat—they either air-pocket -8% to -12% (paying up to +250% to +500% on the 2x Put Spread) or stage a final blow-off spike (hedged by the uncapped 1x Call).
  2. Low Implied Volatility at the Top: Because the sector is still near its highs, Put IV remains cheap (21%-25% on names like MCD, PEP, KO) right before dispersion expands.
KILL-SIGNALS — WHAT WOULD CHANGE OUR MIND
  • Parent sector 5-day Money Flow Score rebounds > +10 points with CMF breadth > 75% (re-entering Markup)
    Currently XLY 5d flow is -17.3, XLB is -15.9, XLP is -9.3
  • VIX spikes above 32 before entry, inflating ATM Put IV > 55% and compressing Put Spread risk/reward below 1.5x
    Current candidate Put IV is 21.7% - 25.7% (ideal entry zone)
The Custom ASIC + AMD Host-CPU Alliance: Why Google TPUv8 & OpenAI Jalapeno Displace Single-Vendor NVDA Racks UPDATED
Hyperscaler diversification away from single-vendor Nvidia (NVDA) racks has reached a structural inflection: Alphabet (GOOGL) confirmed AMD Zen 5…
HIGH CONVICTION Long / Short Pair Published 2026-11-05 3 kill-signals ai-siliconcustom-asicsamd-epyc
▼
Since published (2026-11-05): basket -3.0% vs SOXX +3.5pp vs SPY +2.9pp last 20 trading days: -3.0%
Avoid basket (NVDA, MU, SMCI): -20.1% (L/S spread +17.1pp) — lagging long basket as intended
Per-ticker performance (5)
TickerEntry dateEntry pxLast pxReturn
AMD2026-11-05404.40424.39+4.9%
AVGO2026-11-05291.53304.41+4.4%
GOOGL2026-11-05282.57265.29-6.1%
MRVL2026-11-05133.55122.28-8.4%
TSM2026-11-05353.10318.71-9.7%

Hyperscaler diversification away from single-vendor Nvidia (NVDA) racks has reached a structural inflection: Alphabet (GOOGL) confirmed AMD Zen 5 EPYC "Turin" as the host CPU for its next-generation TPUv8 (TPU 8tx / 8i Max) systems, while OpenAI confirmed EPYC Turin as the host node for its custom Jalapeno ASIC (co-developed with AVGO).

1. Why AMD + AVGO + GOOGL Win Together (Instead of Competing)

  • Google's $135B Third-Party TPU Expansion: Google's TPUv8 delivers ~30% lower TCO than Nvidia GPUs at comparable inference throughput, driving projected third-party TPU sales from $12B in 2026 to $135B by 2028 across Apple (AAPL), Meta (META), Oracle (ORCL), Anthropic, and OpenAI.
  • Dual-Pronged AMD Server TAM ($25B $\rightarrow$ $220B by 2030): Whenever a hyperscaler shifts workloads from Nvidia's proprietary Grace-Blackwell / Vera-Rubin racks (which lock customers into Nvidia's internal ARM Grace/Vera CPU) toward custom ASICs (Google TPU 8tx or OpenAI Jalapeno), Nvidia loses both the accelerator and the host CPU, whereas AMD captures the high-margin EPYC host CPU socket on top of its own ramping Helios rack-scale + Instinct MI455/MI500 GPU deployments (Q3 Data Center revenue +97% YoY to $8.5B; 2027 revenue guided +80% YoY to ~$94.8B, supporting a $634 base-case / $1T+ market cap valuation).

2. Why NVDA, MU, and SMCI Face Near-Term Relative Headwinds

  • Governance & Singapore (Megaspeed) Export Scrutiny: Nvidia's November 2026 board investigation and U.S. intelligence/BIS scrutiny over Southeast Asian datacenter intermediaries threaten export license velocity on ~15%–20% of merchant GPU cluster shipments.
  • Collateral Compression on Micron (MU): Every deferred Nvidia rack or workload migrated to TCO-optimized inference ASICs (TPU 8tx) reduces near-term HBM3e/HBM4 bit intensity per token, pressuring high-fixed-cost memory multiples first.

3. System Portfolio & Scanner Rules

  • Long / Overweight Basket: AMD, AVGO, GOOGL, MRVL, TSM
KILL-SIGNALS — WHAT WOULD CHANGE OUR MIND
  • Hyperscalers (Google TPUv8 / OpenAI Jalapeno) shift custom ASIC host-node CPUs away from x86 EPYC Turin/Venice toward internal custom ARM host processors
    Active monitoring via Sentinel (since 2026-11-05)
  • Avoid basket (NVDA, MU, SMCI) outperforms Long ASIC/Host-CPU basket (AMD, AVGO, GOOGL, MRVL) by >8 percentage points over a trailing 20-trading-day window
    Relative performance spread check (AVGO/MRVL/ANET vs NVDA/MU/SMCI) over 60 trading days
  • AMD quarterly Data Center revenue growth decelerates below +50% YoY or gross margins compress >200 bps below guidance on component inflation
    Tracked via quarterly earnings & supply-chain checks
Own What Algorithms Can't Deflate (token-cost stack)
HIGH CONVICTION Published 2026-08-30 2 kill-signals ai_economicstoken_costsemiconductors
🔒
Since published (2026-08-30): basket -26.8% vs SOXX +0.7pp vs SPY -16.8pp last 20 trading days: -17.9%
Per-ticker performance (3)
TickerEntry dateEntry pxLast pxReturn
ASML2026-08-311696.011266.22-25.3%
TSM2026-08-31415.32318.71-23.3%
BESI2026-08-31226.89154.61-31.9%
Hyperscalers ranked: who captures the surplus when intelligence gets cheap
MEDIUM CONVICTION Published 2026-08-31 5 kill-signals hyperscalerscloudai_economics
🔒
Since published (2026-08-31): basket -9.6% vs QQQ +5.5pp vs SPY +0.5pp last 20 trading days: -10.2%
Avoid basket (MSFT): -4.5% (L/S spread -5.1pp) — outperforming long basket / market (headwind)
Per-ticker performance (3)
TickerEntry dateEntry pxLast pxReturn
GOOGL2026-08-31339.35265.29-21.8%
AMZN2026-08-31259.77246.82-5.0%
META2026-08-31572.34560.84-2.0%
3D stacking / hybrid bonding: the packaging supercycle picks
HIGH CONVICTION bullish Published 2026-07-22 1 kill-signal packaginghybrid_bondingcfet
🔒
Since published (2026-07-22): basket -36.7% vs SOXX -3.3pp vs SPY -28.9pp last 20 trading days: -15.3%
Per-ticker performance (5)
TickerEntry dateEntry pxLast pxReturn
BESI2026-07-22283.80154.61-45.5%
SNPS2026-07-22377.61396.17+4.9%
UCTT2026-07-22101.4637.00-63.5%
AMKR2026-07-2266.9730.24-54.9%
TSM2026-07-22421.21318.71-24.3%
Chinese LLM parity: model layer commoditizes, silicon independence is the real question
HIGH CONVICTION Published 2026-08-31 1 kill-signal chinallmopen_source
🔒
Since published (2026-08-31): basket -19.8% vs QQQ -4.7pp vs SPY -9.7pp last 20 trading days: -15.6%
Per-ticker performance (2)
TickerEntry dateEntry pxLast pxReturn
AVGO2026-08-31370.34304.41-17.8%
GOOGL2026-08-31339.35265.29-21.8%
The Invariant Stack: Own the Layers Every Future Must Pass Through
HIGH CONVICTION Published 2026-08-30 4 kill-signals ai_infrastructurepowergrid
🔒
Since published (2026-08-30): basket -27.5% vs SPY -17.4pp vs SPY -17.4pp last 20 trading days: -15.7%
9 of 10 names priced (missing: 000660)
Per-ticker performance (9)
TickerEntry dateEntry pxLast pxReturn
ETN2026-08-31401.88311.33-22.5%
POWL2026-08-31179.73148.18-17.6%
HUBB2026-08-31453.00403.66-10.9%
GEV2026-08-31898.53699.33-22.2%
CCJ2026-08-3198.7658.00-41.3%
LEU2026-08-31171.4081.43-52.5%
ASML2026-08-311696.011266.22-25.3%
TSM2026-08-31415.32318.71-23.3%
BESI2026-08-31226.89154.61-31.9%
Cheap tokens: agents eat labor budgets, seat-based SaaS is the short
MEDIUM CONVICTION Published 2026-08-30 1 kill-signal softwaresaasagents
🔒
Since published (2026-08-30): basket -17.8% vs QQQ -2.7pp vs SPY -7.7pp last 20 trading days: -5.4%
Avoid basket (CRM, NOW, ACN, INFY): -13.6% (L/S spread -4.2pp) — outperforming long basket / market (headwind)
Per-ticker performance (1)
TickerEntry dateEntry pxLast pxReturn
AVGO2026-08-31370.34304.41-17.8%