Glasshouse Fund
Decision journal

AI fund buys PG, WMT, XOM, sells AMZN · September 15, 2026

Maintain a defensive neutral posture: trim AMZN modestly to address confirmed weakness in a major concentration, hold demonstrated relative-strength leaders without adding, avoid weak momentum candidates, and retain elevated cash while yields and volatility are rising.

Outlook: NEUTRALPortfolio $1,035,334Cash 25.1%Grounding: ok

Trades

SELL 75 AMZN78% · filled at $249

Trim a large concentrated holding after a 6.9% 30-day decline and 1.3% five-day decline, while it remains only modestly above average cost. The sale reduces consumer and large-growth concentration without abandoning the long-term franchise.

Risks the AI named

Market calls

A directional call (beat or lag the S&P 500 over the horizon) on every researched name, whether or not the fund traded it. 34 calls (13 outperform, 21 underperform); 5 became trades. These are the fund's calibration record.

beat SPYlag SPYAAPL: outperform at 70%AMD: underperform at 64%AMZN: underperform at 68%APLD: underperform at 78%ASML: underperform at 74%AVGO: underperform at 75%BE: underperform at 60%CAT: underperform at 65%CORZ: underperform at 72%CRWV: underperform at 70%GLW: underperform at 67%GOOGL: outperform at 58%HD: underperform at 73%INTC: underperform at 57%IREN: underperform at 59%JNJ: outperform at 64%JPM: underperform at 58%MA: outperform at 62%META: outperform at 69%MSFT: outperform at 59%MU: underperform at 61%NVDA: underperform at 65%ORCL: underperform at 66%PG: outperform at 55%QQQ: underperform at 54%SNDK: underperform at 60%SPY: outperform at 50%TSLA: outperform at 54%TSM: underperform at 61%UNH: underperform at 66%V: outperform at 63%WMT: outperform at 57%XOM: outperform at 62%^VIX: outperform at 61%50%60%70%80%90%100%
Each tick is one call, placed by stated confidence. Hover for the name.
All 34 calls, with reasoning
SymbolCallConf.Why
APLDUnderperform SPY78%A 13.2% five-day and 17.2% 30-day decline signals severe negative momentum versus SPY.
AVGOUnderperform SPY75%A sharp 6.5% five-day and 17.2% 30-day decline leaves momentum materially weaker than SPY.
ASMLUnderperform SPY74%The 10.7% five-day and 12.5% 30-day losses indicate persistent semiconductor-equipment weakness versus SPY.
HDUnderperform SPY73%Its 11.7% 30-day decline materially trails SPY and reflects continuing housing-sensitive weakness.
CORZUnderperform SPY72%Large negative five-day and 30-day returns indicate sustained high-beta infrastructure weakness versus SPY.
AAPLOutperform SPY70%Strong 5-day and 30-day momentum versus declining SPY supports relative outperformance, though the recent rally raises chase risk.
CRWVUnderperform SPY70%The 16.9% five-day and 8.1% 30-day decline signals adverse momentum and AI-slowdown sensitivity.
METAOutperform SPY69%Its 8.5% five-day and 11.1% 30-day gains demonstrate strong relative leadership, tempered by reversal risk.
AMZNUnderperform SPY68%Negative five-day and materially negative 30-day momentum indicate deterioration relative to SPY.
GLWUnderperform SPY67%tradedA 13.5% five-day and 9.6% 30-day decline, alongside disclosed operating and demand risks, favors underperformance.
ORCLUnderperform SPY66%The 10.9% five-day decline and slightly negative 30-day performance suggest ongoing relative weakness.
UNHUnderperform SPY66%Negative five-day and 30-day momentum while below portfolio cost indicates continued relative weakness.
CATUnderperform SPY65%Negative returns over both five and 30 days point to cyclical weakness relative to SPY.
NVDAUnderperform SPY65%Negative five-day and 30-day returns amid AI-slowdown concerns indicate fading relative momentum.
AMDUnderperform SPY64%Its negative 5-day return and volatile semiconductor backdrop suggest weaker near-term relative performance versus SPY.
JNJOutperform SPY64%Positive 30-day performance in a weak market and potential talc-litigation uncertainty reduction support relative resilience.
VOutperform SPY63%Positive five-day and 30-day returns show payment-network relative leadership versus SPY.
MAOutperform SPY62%Positive returns over both measured periods show durable relative strength versus SPY.
XOMOutperform SPY62%tradedPositive returns over both periods and a backdrop of higher yields and oil support relative strength versus SPY.
MUUnderperform SPY61%A 7.6% five-day decline signals semiconductor volatility likely to outweigh its positive 30-day result near term.
TSMUnderperform SPY61%tradedNegative returns over five and 30 days leave semiconductor-cycle momentum weaker than SPY.
^VIXOutperform SPY61%Rising volatility over five and 30 days amid yield-driven equity stress supports continued relative strength versus SPY.
BEUnderperform SPY60%Despite positive 30-day performance, the 7.3% five-day decline suggests elevated volatility and weaker near-term relative momentum.
SNDKUnderperform SPY60%Its strong 30-day gain is outweighed near term by a 10.7% five-day reversal and high storage-sector volatility.
IRENUnderperform SPY59%Positive 30-day performance is offset by an 8.0% five-day decline and elevated high-beta infrastructure volatility.
MSFTOutperform SPY59%Positive five-day and modestly positive 30-day returns compare favorably with SPY's decline.
GOOGLOutperform SPY58%Positive five-day and 30-day returns versus negative SPY support modest relative strength, despite being below portfolio cost.
JPMUnderperform SPY58%Its negative five-day and 30-day returns indicate weaker near-term momentum than SPY despite higher-rate support.
INTCUnderperform SPY57%Negative five-day momentum and a slightly negative 30-day return provide a modest underperformance signal.
WMTOutperform SPY57%tradedPositive five-day momentum and comparatively defensive characteristics support modest outperformance despite a negative 30-day return.
PGOutperform SPY55%tradedSmall positive returns over five and 30 days suggest modest defensive resilience versus SPY.
QQQUnderperform SPY54%Negative five-day and 30-day returns, combined with rate pressure on growth equities, indicate a modestly weaker outlook than SPY.
TSLAOutperform SPY54%Positive 30-day performance supports a modest relative-strength view, though recent five-day momentum is negative.
SPYOutperform SPY50%SPY is the comparison baseline; this neutral-confidence call reflects no asserted directional edge against itself.

The debate

Bull, bear and risk cases: bull 0.72 · bear 0.84 · risk 0.97
bull case · conviction 0.72

Buy or hold selectively: the portfolio contains several clear relative-strength leaders that are outperforming a weakening market, while company-specific catalysts remain constructive. The S&P 500 is down 1.6% over 30 days, yet AAPL, META, V, MA, JNJ, GOOGL, and MSFT are positive over the same period. This leadership is especially valuable amid rising volatility and a sharp Treasury-yield-driven market selloff. Conviction is moderate rather than high because several holdings are lagging and the macro backdrop is adverse.

  • AAPL is the strongest large holding on measured momentum: up 5.3% over five days and 9.2% over 30 days versus the S&P 500's 1.6% decline. News flow identifies a new CEO transition, potential foldable iPhone developments, and Morgan Stanley's positive view that Apple's upcoming offerings are becoming more compelling.
  • META shows the best momentum among the major holdings, rising 8.5% over five days and 11.1% over 30 days. That is substantial relative strength while the benchmark and QQQ are both declining over 30 days.
  • V and MA are displaying durable payment-network leadership: V is up 1.8% over five days and 3.4% over 30 days, while MA is up 0.6% and 2.3%, respectively. Both are outperforming the benchmark over 30 days, supporting a hold-and-add-on-weakness posture.
  • JNJ is up 3.0% over 30 days despite a risk-off tape. Its news flow includes a $5.5 billion proposal aimed at resolving talc litigation, a potentially important uncertainty-reduction catalyst, although the headline alone does not establish final resolution.
  • GOOGL is modestly positive over both five and 30 days, up 3.3% and 1.7%, respectively, while the S&P 500 is down. The constructive fund research also points to Waymo expanding to more domestic cities, entering international markets, and scaling operations, providing a tangible longer-term catalyst beyond the core business.
  • MSFT is up 2.3% over five days and 0.5% over 30 days, outperforming the benchmark despite a difficult market. Recent coverage describes a 27% rally and argues that the AI bear case is weakening, supporting continued exposure, though valuation and AI-safety concerns remain relevant risks.
  • The relative-strength signal is broad rather than isolated: AAPL, META, V, MA, JNJ, GOOGL, and MSFT are all positive over 30 days while SPY is down 1.3% and the S&P 500 is down 1.6%. That breadth favors holding the winners instead of indiscriminately reducing equity exposure.
  • The bullish case should be concentrated, not generalized. AMZN, HD, JPM, NVDA, UNH, and GOOGL's recent performance is mixed or negative in several cases; NVDA is down 6.4% over five days and 2.9% over 30 days, while HD is down 11.7% over 30 days. These weaker names should not be treated as momentum leaders.
  • Macro risk is material: the VIX is up 12% over 30 days, and market coverage cites a 19-year high in the 10-year Treasury yield and a broad equity decline. That argues for emphasizing the demonstrated leaders and preserving the portfolio's substantial cash rather than making an aggressive, market-wide purchase.
bear case · conviction 0.84

The clearest bearish signal is concentrated in economically sensitive, technology-infrastructure, and healthcare exposures: several holdings are losing money versus cost while also underperforming the market over 30 days, and multiple candidates show materially worse momentum. The portfolio should not treat recent winners as immune, but the evidence most strongly supports reducing or avoiding the names with confirmed deterioration rather than adding to them.

  • HD is the most compelling current-holding concern: it is down 6.3% versus average cost and down 11.7% over 30 days, substantially worse than the S&P 500's 1.6% decline.
  • UNH is also deteriorating on both measures, down 7.0% versus cost and 4.0% over 30 days. That combination signals continuing weakness rather than merely a small mark-to-market loss.
  • AMZN is down 6.9% over 30 days and is only modestly above its average cost, leaving a recent downtrend as the dominant signal in the supplied evidence.
  • GOOGL is already below average cost by 2.3%, while JPM is down 3.3% over 30 days. NVDA is down 2.9% over 30 days despite its high current share price, indicating fading momentum in another major technology exposure.
  • The candidate list contains substantially sharper breakdowns than the benchmark: APLD is down 17.2%, AVGO 17.2%, CORZ 14.2%, ASML 12.5%, HD 11.7%, and GLW 9.6% over 30 days. These are poor entry candidates while momentum remains negative.
  • Weakness is broad rather than isolated: SPY and QQQ are both down roughly 1.3% over 30 days, while several individual holdings and candidates have declined much more. That raises the risk that adding to laggards compounds exposure to a weakening tape.
  • The supplied GLW risk disclosures explicitly identify customer losses, customer financing and receivables risk, technology change, component or product issues, regulatory developments, litigation, and difficulty matching capital spending to demand. With GLW already down 9.6% over 30 days, these remain meaningful downside risks even though the evidence does not establish that any specific risk has materialized.
  • The bearish evidence is not universal: AAPL, MSFT, JNJ, V, MA, and META are above average cost, and no negative 30-day signal is supplied for most of them. That limits conviction in a portfolio-wide liquidation call, but it does not weaken the case against the specifically flagged laggards and candidates.
risk case · conviction 0.97

The portfolio’s primary capital risk is concentration: 70.4% of invested assets are in AAPL, AMZN, and JNJ, while 43.9% is allocated to Information Technology and 26.0% to Consumer Discretionary. This creates substantial dependence on a small number of holdings and two major sectors. The 29.0% cash allocation provides liquidity and reduces exposure to invested-asset drawdowns, but it does not offset the concentration of the capital that is invested.

  • AAPL is 30.3% of invested assets, the largest single-position exposure; AMZN is 20.5% and JNJ is 19.6%.
  • The top three positions together represent 70.4% of invested assets, creating significant single-name concentration.
  • Information Technology and Consumer Discretionary comprise 69.9% of invested assets, indicating a pronounced sector tilt.
  • Including Health Care, the three largest sectors account for 89.7% of invested assets, leaving limited sector diversification.
  • MSFT, GOOGL, and NVDA add to the technology and large-growth concentration alongside AAPL.
  • Cash is 29.0% of the portfolio, a meaningful liquidity buffer, but the remaining invested capital remains highly concentrated.
  • The evidence does not show leverage or borrowing; the key identified risk is concentration rather than financing exposure.

Cash thesis

Cash is 29.0% of portfolio value, above the 25% threshold, and is intentionally retained. Elevated yields, rising volatility, weak benchmark momentum, and concentrated equity exposure make immediate broad deployment unattractive. The reserve limits total-portfolio drawdown, avoids averaging into deteriorating technology-infrastructure and cyclical names, and preserves flexibility for future purchases only after broader market or company-specific confirmation. The AMZN trim modestly increases this defensive buffer rather than deploying capital into a weak tape.

Full commentary

Bear case response

Accept: HD is a confirmed laggard, down 6.3% versus cost and 11.7% over 30 days; it should not receive new capital, although its approximately $40.8k size does not require forced selling today. Accept: UNH is below cost and down 4.0% over 30 days; the three-share position is immaterial, so it is held rather than traded solely for a negligible exposure change. Accept: AMZN is down 6.9% over 30 days and only modestly above cost; because it is a major portfolio concentration, an incremental 75-share sale is warranted. Accept in part: GOOGL, JPM, and NVDA have weak or fading trends; GOOGL's relative strength over five days and positive 30-day return versus SPY supports holding rather than selling, while NVDA and JPM receive no additions. Accept: APLD, AVGO, ASML, CORZ, GLW, and other sharply negative-momentum candidates are unsuitable purchases in the current tape; GLW's disclosed operating, customer, financing, regulatory, and litigation risks reinforce avoidance. Accept: market weakness is broader than a few stocks, as SPY and QQQ are negative and yields are rising; this supports cash preservation. Reject: an indiscriminate reduction of all winners is not justified, because AAPL, META, JNJ, V, MA, MSFT, and GOOGL have outperformed SPY on supplied 30-day data. Their strength supports holding, not aggressive buying.

Market summary

The market is in a risk-off, rate-sensitive phase: the 10-year Treasury yield is reported near a 19-year high, broad equity indexes are declining, and the VIX is up 8.8% over five days and 11.9% over 30 days. SPY and QQQ are both modestly negative over 30 days, while dispersion is high: AAPL, META, JNJ, V, MA, GOOGL, and MSFT have shown relative strength, but semiconductors, AI infrastructure, housing-linked names, and several portfolio holdings have weakened materially.

Portfolio assessment

The portfolio has resilient core winners but remains concentrated. AAPL, AMZN, and JNJ are the dominant positions, while AAPL, MSFT, NVDA, AMZN, and GOOGL create substantial correlated large-growth and AI/cloud exposure. AAPL's 5.3% five-day and 9.2% 30-day advance supports holding rather than chasing. AMZN's 6.9% 30-day decline is more concerning because it remains a large position and has lost relative strength. The selected AMZN trim is incremental and follows fund memory's preference for early risk reduction in weakening, concentrated holdings rather than wholesale liquidation.

Risk assessment

The main risk is concentration rather than leverage: AAPL, AMZN, and JNJ comprise roughly 70% of invested assets, and technology plus consumer discretionary account for about 70% of invested capital. Cash moderates but does not eliminate the effect of a correlated decline among large-growth and consumer-oriented holdings. Rising Treasury yields create valuation-pressure risk for long-duration technology exposure. HD, AMZN, NVDA, UNH, and JPM have negative 30-day momentum, while AAPL and META have upside reversal risk after strong gains. Fund memory supports measured trimming of persistent weakness and avoidance of repeated turnover.

← September 14, 2026 September 16, 2026 →