AGI core
96.5% equities / 0% gold · GLD / 3.5% T-bills · SHV
BASELINE · 26 POSITIONSConviction. Meet the downside.
One thesis. Three allocations. Explore the price of protection before choosing your risk.
Core + gold
86.5% equities / 10% gold · GLD / 3.5% T-bills · SHV
Research candidate · not a mandate change80 / 10 / 10
80% equities / 10% gold · GLD / 10% T-bills · SHV
Research candidate · not a mandate changeLoading dated research data…
Growth of account value · USD
AGI coreCore + gold80 / 10 / 10
Research model v3 · not investment advice
HINDSIGHT BIAS: today’s names and weights replayed in the past. This is an exposure test, not a fund track record. Unavailable names stay in SHV until 20 prior observations. Signals use the previous close; fills occur at the next close with ≤5% trailing dollar ADV per asset/session. Fractional shares; simplified settlement. Adjusted returns reinvest distributions. No taxes or FX.
No leverage, shorts or options. Rates are modeled, not broker quotes: $0.005/share ($1 minimum capped at 1%), assumed venue allowance $0.003/share, 5 bp half-spread, square-root impact, SEC/FINRA sell fees at April 2026 rates, also applied retrospectively. Management fee is an expense assumption, not a published tariff. Inflation is your constant assumption, not historical CPI. Cash flows are unitized so deposits cannot erase drawdowns. Buy & hold keeps later contributions in cash; other rules invest them at the next rebalance.
Evidence starts after the decision.
A dated policy, frozen allocations and hash-linked paper records. We do not manufacture a track record from a backtest.
Virtual orders only. Versioned and hash-linked, not externally notarized. No real accounts connected. Missed sessions are recorded as gaps, not backfilled trades.
Inspect the source records ↗What if the thesis hits a wall?
A static, simultaneous shock. Not a probability or a forecast. Gold can fall too.
The thesis has competition.
Seven investable benchmarks. Five concentration rules. One standard of evidence. Challenge the strategy before trusting it.
Loading dated evidence…
AGI candidateETF comparator
Current holdings are replayed with hindsight. ETF lines use adjusted exchange prices, not official fund NAV. Matching equity/gold/T-bill budgets does not guarantee equal risk. No ranking across unequal histories.
Concentration
These are target caps, not continuous stop rules. Between rebalances, market moves can breach them. “Top seven” fixes the original seven names. All 20 combinations are shown; the highest backtest is not a recommendation.
Retrospective results do not establish out-of-sample alpha.
FUNDAMENTAL RESEARCH · 2026-09-06
25 portfolio companies + 1 candidate. No blind spots by design.
A dated research map, not completed valuations. The 25-company portfolio baseline is unchanged; IBM is monitored at 0% outside it. Public facts do not establish an edge: every hypothesis still needs normalized financials, price-implied expectations and independent forward tests.
26Company hypotheses
25Financial-report anchors
0Approved valuations
SHV is the reserve ETF, not a 26th operating company.
GOOGLExisting target 10%Hypothesis · not a trading signal
Q2 2026 · Published 2026-07-22 · Source checked 2026-09-06
- 01 / What the issuer reported
- Cloud revenue grew 82%; other income included a $98.0bn net gain, mainly unrealized equity gains.
- 02 / Our hypothesis — not established alpha
- AI can expand Search and Cloud cash earnings, but investment revaluations must not masquerade as recurring operating growth.
- 03 / What to measure next
- Search revenue and traffic-acquisition costs; Cloud operating profit; capex, depreciation and diluted shares at the next report.
- 04 / What would challenge the thesis
- Reassess if Search economics deteriorate while AI cost rises, or Cloud growth fails to improve normalized cash earnings per share.
- 05 / Appropriate valuation method
- Segment DCF; separate investment assets and their gains. Never annualize this quarter’s headline EPS.
- 06 / Still missing
- Missing: normalized earnings bridge, capex replacement needs and price-implied growth.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
MSFTExisting target 9%Hypothesis · not a trading signal
Q4 FY2026 · Published 2026-07-29 · Source checked 2026-09-06
- 01 / What the issuer reported
- Quarterly capex including leases was $41bn; cash property/equipment spending was $35.8bn. Commercial RPO was $678bn.
- 02 / Our hypothesis — not established alpha
- The edge must come from paid AI adoption exceeding replacement-compute costs, not from counting backlog as cash.
- 03 / What to measure next
- Azure growth, cloud margin, RPO concentration/duration, cash capex versus leases and cash earnings per diluted share.
- 04 / What would challenge the thesis
- Reassess if cloud margins and cash conversion weaken across two comparable reports despite strong bookings.
- 05 / Appropriate valuation method
- DCF with cash capex, finance leases and realistic hardware replacement; normalize investment gains separately.
- 06 / Still missing
- Missing: AI-cohort returns and independent customer-credit assumptions.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
AMZNExisting target 8%Hypothesis · not a trading signal
Q2 2026 / FCF TTM · Published 2026-07-30 · Source checked 2026-09-06
- 01 / What the issuer reported
- AWS sales grew 37%; trailing-twelve-month free cash flow was negative $7.6bn under the company’s definition.
- 02 / Our hypothesis — not established alpha
- Custom silicon and AWS utilization could improve investment returns; retail and advertising must be valued separately.
- 03 / What to measure next
- AWS operating margin, capex/lease commitments, retail margins and rolling cash flow; exclude non-operating investment gains.
- 04 / What would challenge the thesis
- Reassess if capacity spending keeps outrunning cash generation without evidence of better utilization or unit returns.
- 05 / Appropriate valuation method
- Sum of segment DCFs; reconcile leases and maintenance versus growth investment.
- 06 / Still missing
- Missing: defensible capex allocation and a normalized per-share cash-flow forecast.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
NVDAExisting target 8%Hypothesis · not a trading signal
Q2 FY2027 · Published 2026-08-26 · Source checked 2026-09-06
- 01 / What the issuer reported
- Revenue was $96.2bn; Data Center revenue $89.0bn; GAAP gross margin 75.0%.
- 02 / Our hypothesis — not established alpha
- System/software integration may preserve economic rents as inference expands; technical leadership alone does not justify any entry price.
- 03 / What to measure next
- Gross margin, receivables relative to sales, inventory commitments, customer concentration and real deployment economics.
- 04 / What would challenge the thesis
- Reassess if margins erode while receivables/inventory outgrow sales, or customers demonstrate superior ASIC economics at scale.
- 05 / Appropriate valuation method
- Reverse DCF with margin fade, working capital and dilution; no permanent peak-margin assumption.
- 06 / Still missing
- Missing: current price-implied cash growth versus an independently built demand scenario.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
METAExisting target 7%Hypothesis · not a trading signal
Q2 2026 · Published 2026-07-29 · Source checked 2026-09-06
- 01 / What the issuer reported
- Revenue grew 28%; operating margin was 31% versus 43%; quarterly free cash flow was $0.784bn.
- 02 / Our hypothesis — not established alpha
- AI recommendation and ad conversion can monetize before AGI, but incremental gross profit must pay for infrastructure.
- 03 / What to measure next
- Ad price/impressions, expenses excluding clearly identified one-offs, capex, depreciation and cash flow.
- 04 / What would challenge the thesis
- Reassess if recurring cost growth exceeds ad gross-profit growth for two comparable reports after one-off adjustments.
- 05 / Appropriate valuation method
- DCF separating core advertising, infrastructure replacement and speculative projects.
- 06 / Still missing
- Missing: one-off cost bridge and incremental AI investment returns; one quarter cannot establish a trend.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
AVGOExisting target 7%Hypothesis · not a trading signal
Q3 FY2026 · Published 2026-09-02 · Source checked 2026-09-06
- 01 / What the issuer reported
- AI semiconductor revenue was $16.7bn; company free cash flow was $13.7bn, or 46% of revenue.
- 02 / Our hypothesis — not established alpha
- Custom accelerators and networking may gain wallet share even as GPU demand grows; customer bargaining power can absorb that gain.
- 03 / What to measure next
- AI sales converted to cash, customer concentration, software renewals and net debt.
- 04 / What would challenge the thesis
- Reassess if cash margins fall as AI sales rise, or a major design program is cancelled or insourced.
- 05 / Appropriate valuation method
- Separate semiconductor and software cash flows; deduct debt and account for share compensation.
- 06 / Still missing
- Missing: program-level economics and a price-implied customer-retention scenario.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
TSMExisting target 7%Hypothesis · not a trading signal
Q2 2026 · Published 2026-07-16 · Source checked 2026-09-06
- 01 / What the issuer reported
- US-dollar revenue was $40.20bn; gross margin 67.7%; quarterly EPS was US$4.31 per ADR.
- 02 / Our hypothesis — not established alpha
- Leading-node yield and packaging can capture value across competing chip designers; geographic disruption remains a separate tail risk.
- 03 / What to measure next
- Node mix, utilization, packaging capacity, overseas-fab margin impact, capex and monthly sales with seasonal comparisons.
- 04 / What would challenge the thesis
- Reassess after sustained yield/ramp failure or lower utilization with rising capital intensity; analyze disruption separately.
- 05 / Appropriate valuation method
- Cycle-aware DCF in reporting currency, then consistent ADR/FX conversion; no arbitrary catastrophe probability.
- 06 / Still missing
- Missing: capacity/yield scenarios, FX bridge and explicit supply-disruption stress.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
ASMLExisting target 5%Hypothesis · not a trading signal
Q2 2026 · Published 2026-07-15 · Source checked 2026-09-06
- 01 / What the issuer reported
- Sales were €9.326bn; installed-base management sales €2.762bn; gross margin 54%.
- 02 / Our hypothesis — not established alpha
- More lithography intensity and installed-base service may stabilize economics beyond quarterly equipment bookings.
- 03 / What to measure next
- EUV/High-NA adoption, customer acceptance, service sales, gross margin and shipment restrictions.
- 04 / What would challenge the thesis
- Reassess if customers defer adoption because cost per good wafer fails to improve, or restrictions materially impair demand.
- 05 / Appropriate valuation method
- EUR DCF separating systems and service; consistent listed-share currency conversion.
- 06 / Still missing
- Missing: adoption economics and export-restriction scenarios; backlog is not guaranteed revenue.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
ETNExisting target 4%Hypothesis · not a trading signal
Q2 2026 · Release date not verified · Source checked 2026-09-06
- 01 / What the issuer reported
- Sales grew 21%, including 14% organic growth; electrical backlog grew 43% year over year.
- 02 / Our hypothesis — not established alpha
- Electrical bottlenecks may sustain pricing, but acquisitions must not be confused with organic demand.
- 03 / What to measure next
- Organic orders, book-to-bill, backlog conversion, margins, working capital and acquisition returns.
- 04 / What would challenge the thesis
- Reassess if order growth slows while delivery capacity expands and price/cash conversion deteriorates.
- 05 / Appropriate valuation method
- Segment DCF with acquisition capital included in return-on-capital calculations.
- 06 / Still missing
- Missing: organic/acquisition bridge, transaction-adjusted debt and exact release-date verification.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
ORCLExisting target 3%Hypothesis · not a trading signal
FY2026 / Q4-end RPO · Published 2026-06-10 · Source checked 2026-09-06
- 01 / What the issuer reported
- RPO reached $638bn; full-year free cash flow was negative $23.7bn. Customer-prepaid/provided hardware in large AI contracts totaled $75bn.
- 02 / Our hypothesis — not established alpha
- Customer funding may reduce financing pressure, but backlog value depends on payment, capacity delivery and residual obligations.
- 03 / What to measure next
- RPO maturity/counterparties, cash receipts, customer hardware, debt, leases and OCI operating economics.
- 04 / What would challenge the thesis
- Reassess after material deferrals, adverse contract revisions or financing needs that dilute per-share value.
- 05 / Appropriate valuation method
- Contract-aware DCF with funding scenarios; RPO cannot be added to equity value at face value.
- 06 / Still missing
- Missing: contract-level cash schedules and credit exposure, not just aggregate RPO.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
ANETExisting target 3%Hypothesis · not a trading signal
Q2 2026 · Published 2026-08-04 · Source checked 2026-09-06
- 01 / What the issuer reported
- Revenue was $3.036bn, up 37.7% year over year.
- 02 / Our hypothesis — not established alpha
- Ethernet may gain AI-network spending share; open standards do not automatically guarantee Arista captures the margin.
- 03 / What to measure next
- Disclosed AI-network revenue, large-customer concentration, gross margin, deferred revenue and cash conversion.
- 04 / What would challenge the thesis
- Reassess if AI-network deployments shift to rivals or white boxes while Arista’s margins or customer retention weaken.
- 05 / Appropriate valuation method
- DCF with explicit hardware-margin competition and recurring software/service contribution.
- 06 / Still missing
- Missing: comparable AI-fabric share data; headline growth is not proof of share gains.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
VRTExisting target 3%Hypothesis · not a trading signal
Q2 2026 · Published 2026-07-29 · Source checked 2026-09-06
- 01 / What the issuer reported
- Sales grew 24%, including 18% organic growth; adjusted operating margin was 22.6%.
- 02 / Our hypothesis — not established alpha
- Higher power density may increase cooling and power content per rack faster than data-center floor area grows.
- 03 / What to measure next
- Organic orders and sales, liquid-cooling adoption, service mix, working capital and warranty expenses.
- 04 / What would challenge the thesis
- Reassess if orders fail to convert or competitive pricing/warranty costs erase the gain in content per rack.
- 05 / Appropriate valuation method
- DCF with normalized working capital, maintenance capex and a margin-downside case.
- 06 / Still missing
- Missing: verified content per rack and margin attribution; quarter cash flow may be timing-sensitive.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
CEGExisting target 2.5%Hypothesis · not a trading signal
Q2 2026 · Published 2026-08-06 · Source checked 2026-09-06
- 01 / What the issuer reported
- The company reported an additional 920 MW of long-term power purchase agreements.
- 02 / Our hypothesis — not established alpha
- Firm power scarcity can create valuable contracts, but megawatts alone say nothing about equity cash returns.
- 03 / What to measure next
- Contract duration/pricing when disclosed, fuel/hedges, outages, maintenance, debt and integration costs.
- 04 / What would challenge the thesis
- Reassess if contract economics deteriorate after fuel, refurbishment, regulation or counterparty costs.
- 05 / Appropriate valuation method
- Plant/contract cash-flow scenarios plus net debt; separate merchant-price exposure from contracted revenue.
- 06 / Still missing
- Missing: net cash economics of the new contracts; MW announcements are not valuation inputs by themselves.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
GEVExisting target 2.5%Hypothesis · not a trading signal
Q2 2026 · Published 2026-07-22 · Source checked 2026-09-06
- 01 / What the issuer reported
- Backlog was $176bn; the 116 GW gas-power figure includes both equipment backlog and slot reservations.
- 02 / Our hypothesis — not established alpha
- Scarce turbine and grid capacity may improve future contract margins; reservations must not be treated as firm orders.
- 03 / What to measure next
- Reservation conversion, backlog margins, service mix, customer advances and wind warranty/cash costs.
- 04 / What would challenge the thesis
- Reassess if reservations lapse, execution costs consume pricing gains or cash flow depends on growing advances.
- 05 / Appropriate valuation method
- Segment DCF; normalize advances and working capital before extrapolating quarterly free cash flow.
- 06 / Still missing
- Missing: firm-order/slot split, delivery schedule and recurring versus timing-driven cash flow.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
AMDExisting target 2%Hypothesis · not a trading signal
Q2 2026 · Published 2026-08-04 · Source checked 2026-09-06
- 01 / What the issuer reported
- Data Center revenue was $6.7bn, up 107% year over year and 58% of company revenue.
- 02 / Our hypothesis — not established alpha
- A second accelerator supplier creates value only if production adoption and software economics translate into durable profit.
- 03 / What to measure next
- Separate CPU and accelerator growth; customer production usage, gross profit, inventory and software adoption.
- 04 / What would challenge the thesis
- Reassess if announced deployments fail to scale or price discounts prevent accelerator growth from earning an adequate return.
- 05 / Appropriate valuation method
- Segment DCF with GPU share/margin scenarios and separate CPU economics.
- 06 / Still missing
- Missing: accelerator-only cash economics; aggregate Data Center growth includes CPUs.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
KLACExisting target 2%Hypothesis · not a trading signal
Q4 / FY2026 · Published 2026-07-28 · Source checked 2026-09-06
- 01 / What the issuer reported
- Full-year free cash flow was $3.77bn. Per-share figures reflect the June 2026 ten-for-one stock split.
- 02 / Our hypothesis — not established alpha
- Process control can capture more spend per wafer as defects become more costly; this needs proof beyond AI-related demand.
- 03 / What to measure next
- Process-control intensity versus fab spending, service mix, margin, cash conversion and split-adjusted shares.
- 04 / What would challenge the thesis
- Reassess if inspection intensity or competitive share declines, or cash growth persistently lags reported earnings.
- 05 / Appropriate valuation method
- Mid-cycle DCF/earnings; align both EPS and price to the same split basis.
- 06 / Still missing
- Missing: through-cycle intensity evidence and harmonized share history.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
MRVLExisting target 2%Hypothesis · not a trading signal
Q2 FY2027 · Published 2026-08-27 · Source checked 2026-09-06
- 01 / What the issuer reported
- Data Center revenue was $2.172bn, up 46% year over year; total revenue was $2.739bn.
- 02 / Our hypothesis — not established alpha
- Custom silicon and optical links may gain content per system; design wins only matter when they become profitable shipments.
- 03 / What to measure next
- Ramp revenue, cash flow, wafer commitments, acquisition costs and fully diluted shares including preferred equity.
- 04 / What would challenge the thesis
- Reassess if program delays or dilution absorb operating gains, or committed supply outruns customer demand.
- 05 / Appropriate valuation method
- DCF to common equity after commitments, debt and dilution; not headline adjusted EPS alone.
- 06 / Still missing
- Missing: program conversion rates and a fully diluted capital-structure bridge.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
PWRExisting target 2%Hypothesis · not a trading signal
Q2 2026 · Published 2026-07-30 · Source checked 2026-09-06
- 01 / What the issuer reported
- Remaining performance obligations were $33.6bn versus total backlog of $53.4bn; these are different measures.
- 02 / Our hypothesis — not established alpha
- Grid energization and skilled execution may be scarcer than announced data-center plans; execution margins decide value.
- 03 / What to measure next
- RPO/backlog conversion, organic revenue, contract margins, labor capacity, receivables and cash flow.
- 04 / What would challenge the thesis
- Reassess if project overruns or labor inflation erase margins, or receivables rise faster than completed work.
- 05 / Appropriate valuation method
- Contractor DCF with normalized working capital and acquisition-adjusted capital returns.
- 06 / Still missing
- Missing: consistent organic growth and fixed-price versus reimbursable contract exposure.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
AMATExisting target 1.5%Hypothesis · not a trading signal
Q3 FY2026 · Published 2026-08-13 · Source checked 2026-09-06
- 01 / What the issuer reported
- Revenue was $9.12bn, up 25%; company-reported free cash flow grew 14%.
- 02 / Our hypothesis — not established alpha
- More materials steps and advanced packaging may expand spend per chip even when wafer growth slows.
- 03 / What to measure next
- Foundry/DRAM mix, services, gross margin, cash conversion and customer capex excluding acquisition effects.
- 04 / What would challenge the thesis
- Reassess if growth depends on a short capacity surge rather than intensity/share gains, or cash conversion keeps weakening.
- 05 / Appropriate valuation method
- Mid-cycle earnings and DCF, separating service durability from equipment cyclicality.
- 06 / Still missing
- Missing: multi-cycle cash conversion and quantified process-intensity gains.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
MUExisting target 1.5%Hypothesis · not a trading signal
Q3 FY2026 · Published 2026-06-24 · Source checked 2026-09-06
- 01 / What the issuer reported
- Revenue was $41.46bn versus $9.30bn a year earlier; the company reported $18.3bn adjusted free cash flow.
- 02 / Our hypothesis — not established alpha
- HBM mix and supply discipline could improve through-cycle returns; extrapolating a memory boom is the central valuation trap.
- 03 / What to measure next
- HBM contract economics, bit supply, pricing, inventory, capex and cash margins across cycles.
- 04 / What would challenge the thesis
- Reassess if supply expands faster than demand while pricing and inventory turn adverse; test a downcycle explicitly.
- 05 / Appropriate valuation method
- Through-cycle DCF and normalized earnings, not a low P/E computed from peak EPS.
- 06 / Still missing
- Missing: filing reconciliation of exceptional results, supply assumptions and contract protections.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
CDNSExisting target 1.5%Hypothesis · not a trading signal
Q2 2026 · Published 2026-07-27 · Source checked 2026-09-06
- 01 / What the issuer reported
- Revenue was $1.584bn; backlog $8.1bn; RPO expected within twelve months was $4.2bn.
- 02 / Our hypothesis — not established alpha
- Design complexity and more custom chips can increase software value; AI productivity may also pressure seat-based pricing.
- 03 / What to measure next
- Organic recurring growth, renewal economics, RPO conversion, cash flow and stock-based compensation.
- 04 / What would challenge the thesis
- Reassess if AI usage grows without monetization, renewal pricing weakens or dilution offsets cash growth.
- 05 / Appropriate valuation method
- Recurring-software DCF with stock compensation and explicit multiple-compression stress.
- 06 / Still missing
- Missing: pricing/retention evidence and current price-implied duration of high growth.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
SNPSExisting target 1.5%Hypothesis · not a trading signal
Q3 FY2026 · Published 2026-08-26 · Source checked 2026-09-06
- 01 / What the issuer reported
- Revenue was $2.477bn; GAAP EPS $2.84 versus non-GAAP EPS $3.91 for the quarter.
- 02 / Our hypothesis — not established alpha
- Verification, IP and broader engineering tools may deepen the customer relationship, but purchased growth must earn its capital cost.
- 03 / What to measure next
- Organic versus acquired growth, IP licensing, cash conversion, debt, integration expenses and diluted shares.
- 04 / What would challenge the thesis
- Reassess if integration/dilution prevents recurring cash earnings per share from improving, or IP demand deteriorates.
- 05 / Appropriate valuation method
- DCF with acquisition funding and normalized adjustments; reconcile GAAP to non-GAAP rather than selecting the larger EPS.
- 06 / Still missing
- Missing: pro-forma organic base and justified treatment of each earnings adjustment.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
EQIXExisting target 1.5%Hypothesis · not a trading signal
Q2 2026 · Published 2026-07-29 · Source checked 2026-09-06
- 01 / What the issuer reported
- AFFO was $11.78 per share; growth benefited from one-time xScale fees. Net interconnections increased by 9,700.
- 02 / Our hypothesis — not established alpha
- Interconnection density may produce durable pricing beyond commodity space, provided returns exceed construction and financing costs.
- 03 / What to measure next
- Recurring AFFO per share, interconnections, power availability, development yields, debt maturities and refinancing costs.
- 04 / What would challenge the thesis
- Reassess if recurring AFFO stagnates despite capex, or development returns fall below the cost of funding.
- 05 / Appropriate valuation method
- REIT AFFO/NAV and property cash-flow analysis; the simple EPS/P-E worksheet is not the primary valuation model.
- 06 / Still missing
- Missing: AFFO excluding one-time fees and project-level debt/capital commitments.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
LRCXExisting target 1%Hypothesis · not a trading signal
Quarter ended 2026-06-28 · Published 2026-07-29 · Source checked 2026-09-06
- 01 / What the issuer reported
- Quarterly revenue was $6.722bn; non-GAAP operating margin was 38.4%.
- 02 / Our hypothesis — not established alpha
- More etch/deposition steps may increase value per wafer, but memory spending remains cyclical.
- 03 / What to measure next
- DRAM/NAND/foundry mix, process intensity, service revenue, margins and customer acceptance timing.
- 04 / What would challenge the thesis
- Reassess if node transitions fail to increase tool intensity or a memory downturn exposes excess capacity.
- 05 / Appropriate valuation method
- Mid-cycle DCF with a memory downcycle and separate services assumptions.
- 06 / Still missing
- Missing: multi-cycle earnings normalization and comparable process-step intensity data.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
SPCXExisting target 1%Identity only · financial review incomplete
IPO / instrument identity only · Published 2026-06-15 · Source checked 2026-09-06
- 01 / What the issuer reported
- The issuer states that Class A shares began trading under SPCX on June 12, 2026. This source confirms identity, not operating valuation.
- 02 / Our hypothesis — not established alpha
- Space/AI optionality is investable only when the listed entity’s ownership, operating cash flows and funding obligations are understood.
- 03 / What to measure next
- Read the prospectus and periodic filing: segment cash flow, satellite replacement, related parties, dilution and shareholder rights.
- 04 / What would challenge the thesis
- Do not advance the valuation if entity scope or financial reconciliation is unclear; reject a thesis supported only by a space-compute narrative.
- 05 / Appropriate valuation method
- Segment/scenario valuation after filings review. Do not invent EPS or splice the former ETF’s SPCX price history.
- 06 / Still missing
- Financial coverage incomplete: identity verified, operating statements and price-implied expectations not reviewed in this dossier.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
IBMWatchlist candidate · outside portfolio · 0%Watchlist candidate · outside portfolio · 0%
Research trigger / Q2 2026 context · Published 2026-08-13 · Source checked 2026-09-06
- 01 / What the issuer reported
- IBM and OpenAI announced a joint enterprise go-to-market motion: GPT-5.6, Codex and ChatGPT Work are being embedded into IBM Consulting Advantage, with a dedicated OpenAI practice and forward-deployed teams.
- 02 / Our hypothesis — not established alpha
- IBM may become a trusted integration and governance tollbooth between frontier models and regulated legacy workflows. This is a distribution hypothesis, not evidence that IBM owns a frontier model or has created AGI.
- 03 / What to measure next
- Convert AI “book of business” into recurring Software/Red Hat/watsonx revenue, Consulting productivity and segment margin; track paid deployments, cash conversion, diluted shares, net debt and the utilization economics of third-party AI infrastructure.
- 04 / What would challenge the thesis
- Challenge the thesis if partnership announcements multiply while comparable Software/Consulting margins and cash per diluted share fail to improve, or if model and compute suppliers capture most of the economics.
- 05 / Appropriate valuation method
- Segment sum-of-parts DCF. Separate recurring software economics, implementation services, infrastructure-cycle exposure, acquisitions and net debt; do not capitalize bookings or partnership counts as cash.
- 06 / Still missing
- Missing: reproducible bridge from AI bookings to segment revenue, margin and free cash flow per share; workload-level retention, utilization and third-party model/compute costs; price-implied expectations.
No measured consensus gap or approved price target. A familiar good business can still be an expensive stock.
Issuer-reported figures, not an independent accounting audit. Review after the next report or material filing. These review triggers are proposed, not automatic sell rules. Weights and paper mandates are unchanged.
Separate a business from its price.
25 research worksheets. No verified fundamental valuation yet. Enter figures from a dated company filing; assumptions remain yours and never place an order.
Terminal EPS = entry EPS × [(1 + profit growth) ÷ (1 + share-count growth)]^years. Terminal price = EPS × exit P/E. These are scenario prices, not price targets. No dividends, taxes or fees; unsuitable for negative-earnings businesses. No invented probabilities or expected return.
Complete worksheet ≠ verified data or approval. Review the filing, ADR/share units and normalization independently.
Rules first. Evidence later.
A separate, hash-linked research mandate freezes 54 virtual experiments. The original portfolio and v3 journal are untouched. No real accounts or orders.
Recorded after the mandate, not reconstructed from a backtest. Missed common sessions are gaps: stale orders are cancelled, no retrospective fills. Hashes detect changes but are not independent notarization. Engine changes require a new mandate; target caps are not production risk limits.
Raw records ↗Download source evidence · MODEL 4.0.0
ETF operating expenses are embedded in prices and are not deducted again. AGI fee is your assumption; ETF wrapper fee is zero. Both sides pay modeled execution costs. Default: $0.005/share (min $1, cap 1%), assumed venue fee $0.003/share, half-spread 5 bps, square-root impact; no more than 5% of previous 20-observation dollar ADV per asset/session. EUR TEQ adds assumed FX 2 bps; SEC/FINRA assumptions apply only to US listings. Not a broker tariff. Taxes and settlement are excluded.
Only common observed dates. Returns span omitted dates; missing quotes are not zero returns. EUR TEQ is translated with same-date EUR/USD; closing times differ. Pre-IPO AGI positions wait in SHV. Young ETFs have a 20-observation liquidity warm-up.
JSON · Download source evidence ↗