GS 31 min read

The Goldman Sachs Group, Inc. (GS) — Investment Tree v1

Stage 7 final essay. Bilingual companion: tree_v1_zh.md. Date: 2026-07-20 · Anchor price: $1,067† (2026-07-19; intraday ATH ~$1,152 reached 2026-07-15, day after Q2 2026 earnings) · Forward P/E: ~15.6x† (2027E) · P/TBV: ~3.4x† · Dividend yield: ~1.9%† Archetype: Dealmaking-franchise fee momentum racing a still-dominant, capital-intensive trading engine — the third bulge-bracket bank scaffolded in this corpus (after JPM, MS), and the one where the fee-vs-trading revenue-quality decomposition has to happen within a single reported segment (GBM) rather than across reported segments.

SOURCE QUALITY: Mixed Tier A/B/C. FY2025 (10-K, filed 2026-02-25) and Q1 2026 (10-Q, filed 2026-05-01) figures were fetched directly from SEC EDGAR (Tier A, CIK 0000886982) — see evidence_2026-07-11.jsonl (22 rows, 20 load-bearing). Q2 2026 figures (reported 2026-07-14, five days before this scaffold) were obtained via WebSearch against Goldman Sachs' own press release and secondary aggregators (StockTitan, Yahoo Finance, Investing.com, Benzinga) — Tier B/C, not yet independently confirmed against the Q2 2026 10-Q, which had not filed as of this tree's construction. Items flagged † throughout require R2 verification before this tree is treated as final for a real capital-allocation decision — most urgently the Q2 2026 Standardized CET1 ratio (Leaf 2.1), the single highest-priority open item.

0. Company Fundamentals — what Goldman Sachs is and how it earns

Figures FY2025 (calendar 2025) / Q2 2026† unless noted. All Tier A unless marked.

What it is & how it earns. The Goldman Sachs Group is a global investment bank and financial-services holding company operating through three segments: Global Banking & Markets (GBM) — investment-banking advisory/underwriting plus FICC and Equities trading/market-making/financing, 71.1% of FY2025 net revenues — Asset & Wealth Management (AWM) — fee-based Assets Under Supervision (AUS) plus private banking/lending, 28.6% of FY2025 net revenues — and Platform Solutions — a wind-down consumer/transaction-banking remnant post-Apple Card exit, now immaterial (0.3% of FY2025 revenue). FY2025 net revenue $58.28B (+9% YoY), net earnings $17.18B, diluted EPS $51.32 (+26.6%), firmwide ROE 15.0% / ROTE 16.0%. Q2 2026 (most recent quarter, reported 2026-07-14†) stepped up sharply to a record: net revenues $20.34B (+39% YoY), net earnings $6.63B, diluted EPS $20.98, annualized ROE 23.5% / ROTE 25.5% — running roughly 9-10 points above the top of management's own 14-16% ROE / 15-17% ROTE through-cycle target range.

Cash-flow anatomy. As a regulated bank holding company, GS's "capital generation" framing substitutes for a traditional FCF waterfall — excess capital above regulatory minimums is what funds capital return:

Unlike JPM's tree (where the CET1 buffer sat at ~280bps above requirement) or a typical G-SIB durability finding, GS's own capital cushion above its 11.4% requirement has compressed from >300bps (Dec 2025) to roughly 110bps (Mar 2026) — the thinnest buffer of any bulge-bracket bank scaffolded in this corpus — even as capital return continued at pace.

Balance sheet & capital allocation. 2026 CCAR (announced 2026-06-24) held the Stress Capital Buffer flat at 3.4% through Sep 30, 2027, keeping the Standardized CET1 requirement unchanged at 11.4% — a favorable regulatory-requirement signal. But this arrived in the same window as the 180bps single-quarter CET1 decline — i.e., the requirement held steady while GS's own buffer above it compressed materially. This two-sided tension (regulatory tailwind vs. company-specific capital consumption) is the single cleanest fact pattern this tree found that resists a simple bull/bear binary.

What drives it. taxonomy.md's custom axis re-cuts GBM's own disclosed sub-lines by revenue-generation mechanism and capital intensity: Advisory & Underwriting (16.0% of FY2025 revenue, low capital intensity, fee-based) — Trading & Market-Making/Intermediation (33.7%, high capital intensity, the single largest category by dollars) — Financing (19.6%, high capital intensity but more annuitized in character) — plus AWM (28.6%, low-medium capital intensity) and a Platform Solutions/Other residual (2.1%). The tree resolves whether Advisory & Underwriting's persistently faster growth rate (+20.8% FY2025 → +48.4% Q1 2026 → +55%† Q2 2026, each period outpacing total GBM growth) represents a durable, franchise-fee-driven mix shift the market hasn't yet separately priced — or a two-quarter dealmaking-window artifact riding on top of a still-dominant, capital-intensive trading engine that has not shrunk in absolute importance.

I. One-sentence verdict

GS's ~$1,067†/~3.4x tangible-book/25.5%† Q2 2026 ROTE combination is being priced by both bulls and bears as one undifferentiated "AI-capex cycle" story, when the taxonomy-level decomposition shows a genuinely accelerating, capital-light Advisory & Underwriting franchise (IB fees +55%† Q2 2026, outpacing total GBM growth for a third consecutive period) racing against a still-dominant, capital-intensive Trading & Market-Making engine (53.3% of revenue combined with Financing) that is simultaneously consuming an accelerating share of a materially thinner capital buffer (Standardized CET1 down 180bps in one quarter, to the thinnest cushion of any G-SIB tree in this corpus) — and the tree's own math finds current pricing offers a mildly negative expected value (-5.9%) and a near-parity-but-unfavorable 0.94x risk/reward asymmetry, making this a WATCH position at 0% today, with a contingent 1-2% starter appropriate only if Q3 2026 earnings confirm the fee-lead persists a third consecutive quarter AND the Q2 2026 10-Q shows the CET1 buffer stabilizing rather than continuing to compress.

II. Company snapshot

The Goldman Sachs Group, Inc. is a global investment bank and financial-services holding company organized into three reportable segments — Global Banking & Markets (GBM, investment-banking advisory/underwriting + FICC/Equities trading and financing; 71.1% of FY2025 net revenues, ROE 16.4%) — Asset & Wealth Management (AWM, $3.606T-$4.04T† AUS, ROE 12.5%) — and Platform Solutions (wind-down consumer-banking remnant post-Apple-Card-exit, now immaterial). FY2025 net revenue $58.28B, net earnings $17.18B, diluted EPS $51.32.

Q2 2026 (reported July 14, 2026) was the print that set up this tree's central question: record net revenues $20.34B (+39% YoY), net earnings $6.63B, diluted EPS $20.98 (~45% consensus EPS surprise†), annualized ROE 23.5% / ROTE 25.5%, efficiency ratio 57.4%. GBM record $15.52B (Equities + FICC + IB fees all up sharply, IB fees specifically +55%†); AWM record $4.60B (record $4.04T† AUS, record $59B† Q2 third-party alternatives fundraising); Platform Solutions $221M with a small net loss. Layered on top: a five-year-high investment-banking backlog, a record $1 trillion in H1 2026 announced M&A deal volume with GS maintaining the #1 global M&A-advisor position, a dividend raised to $5.00/share (+11% QoQ), continued large-scale buybacks ($4.00B† in Q2 alone), and — the tension this tree is built to test — a 180bps single-quarter Standardized CET1 decline (14.3%→12.5%, Dec 2025→Mar 2026) that has not yet been confirmed to have reversed, stabilized, or continued compressing as of this scaffold's construction date.

III. The five facts that drive everything

1. Advisory & Underwriting (IB fees) has outgrown total GBM revenue in every period observed — and the gap is widening, not narrowing: +20.8% vs. +18.2% (FY2025) → +48.4% vs. +18.6% (Q1 2026) → +55%† vs. record-but-unbroken-out (Q2 2026). ⚠️B (real, three-period pattern; the fourth confirming period, Q3 2026, has not yet happened)
2. GBM's own revenue-level concentration is rising, not falling — the opposite direction from a "diversifying, less-cyclical franchise" narrative — GBM was 71.1% of FY2025 revenue and an estimated ~76%† of the Q2 2026 record. The two capital-intensive categories (Trading + Financing) remain 53.3% of FY2025 revenue — a dollar-weighted majority that has not shrunk even as the fee sub-line's growth rate accelerates. ⚠️A — a real tension neither bulls nor bears in consensus.md explicitly reconcile.
3. Standardized CET1 fell 180bps in a single quarter (14.3%→12.5%, Dec 2025→Mar 2026) despite positive earnings — the thinnest capital buffer (~110bps above the 11.4% requirement) of any G-SIB tree in this corpus. ⊗ (Q2 2026 figure not yet confirmed — the single highest-priority open item)
4. GS trades at ~15.6x 2027E EPS / ~3.4x tangible book — a premium over the ~13.3x/~2.6x peer average — while the 21-analyst average price target ($935.74) sits ~14% below the current ~$1,067 spot price, an "orphan premium" the sell-side's own aggregate model has not yet caught up to justify even after the Q2 beat. ✗A — the valuation math is genuinely unfavorable by this tree's own reverse-engineering, though less severely so than the analogous finding in this corpus's JPM tree.
5. Management's own framing (CEO Solomon, Q2 2026 earnings call) explicitly concedes cyclicality risk underneath the bull case — the AI investment cycle is "relative early innings" but "won't be a straight line," with "bumps and recalibrations" possible on a 6-18 month view. ✅B — direct, management-sourced evidence the current run rate is not being guided as a new steady state.

IV. The H-0 thesis

H-0 (one sentence): GS's record-earnings re-rating is being priced as one undifferentiated "AI-capex cycle" story, but the growth is really two structurally different engines — a durably outgrowing advisory-fee franchise and a capital-hungry trading book — and the market hasn't separated them yet.

Mispricing taxonomy: Structural blindness × Lifecycle stage (per mispricing.md, primary). GS's own segment reporting bundles Advisory & Underwriting fee revenue (durable, franchise-relationship-based, capital-light) with FICC/Equities trading and financing revenue (cyclical, balance-sheet-intensive) into a single undifferentiated GBM line. No field in the standard sell-side model — nor in this session's own consensus.md/assumptions.md sourcing — separately tracks which of the two is actually driving the reported growth, so the market defaults to characterizing the whole 71-76% GBM concentration as uniformly one lifecycle stage or the other.

Secondary mechanism: Cognitive bias × Category. This corpus's existing ai_capex_cycle_overlay.md taxonomy risks being imported onto GS at face value via availability/narrative bias — Solomon's own AI-capex-cycle language makes it easy to categorize GS alongside direct infrastructure sellers (NVDA, TSM, AVGO) without first testing whether GS's actual exposure mechanism (advisory/financing fees on AI-buildout capital raising, not AI-infrastructure sales) scales the same way.

3 falsification conditions (per h0_thesis.md):

• FF1 Q3/Q4 2026 shows IB-fee growth decelerating to match or fall below total-GBM growth (rather than continuing to outpace it) — the "durable, increasingly fee-driven mix" claim would be falsified for that quarter, and the FY2025-Q2 2026 pattern would read as a temporary IB-fee upswing (consistent with the record $1T H1 2026 M&A volume), not a structural mix shift.
• FF2 Standardized CET1 stabilizes or recovers toward the >300bps-above-requirement buffer seen at Dec 2025 (rather than continuing to compress) — this would weaken the "capital-hungry trading engine still expanding" half of the thesis.
• FF3 GS (or a major sell-side house) explicitly publishes a fee-vs-trading revenue decomposition and the market's multiple does NOT move in response — this would indicate the structural-blindness mechanism is not actually load-bearing, i.e., the market may already be pricing this distinction implicitly.

V. Tree — five branches

H-0: Market prices GS's record 2026 earnings as one undifferentiated "AI-capex
     cycle" story; taxonomy decomposition shows a durably outgrowing advisory-fee
     franchise racing a still-dominant, capital-intensive trading engine
│
├── L1A — GBM Revenue-Quality Decomposition  ⚠️B PARTIALLY SUPPORTED (highest priority)
│   ├── 1.1 Advisory & Underwriting lead persists a 3rd/4th consecutive period  ⚠️B three-period pattern real, 4th pending
│   ├── 1.2 Individual Trading/Financing sub-line growth rates confirmed        ⊗ evidence gap — R2 owed
│   └── 1.3 Record $1T H1 2026 M&A volume sustaining into H2 2026               ⊗ catalyst pending
│
├── L1B — Capital/RWA Trajectory  ⊗ UNRESOLVED (nearest-term data point)
│   ├── 2.1 Q2 2026 Standardized CET1 ratio confirmed                          ⊗ catalyst pending — highest priority
│   ├── 2.2 Q2 2026 buyback deceleration correlates with tighter buffer         ⚠️B directionally correlated, causation uncertain
│   └── 2.3 CCAR SCB headroom vs. company-buffer already-consumed reconciled    ⚠️B genuinely two-sided finding
│
├── L1C — AI-Capex Financing Exposure Specificity  ⊗ WEAKEST DATA AVAILABILITY (by design)
│   ├── 3.1 Fraction of IB fees attributable to AI-infrastructure deals         ⊗ evidence gap — may be unresolvable
│   ├── 3.2 2021-2023 M&A/IPO boom-bust base rate                              ⚠️C training-knowledge interpolation
│   └── 3.3 IB backlog composition skews AI/tech-infrastructure, per Solomon    ⚠️B qualitative, promotional-bias-prone
│
├── L1D — Valuation-vs-Consensus Gap Resolution  ⊗ UNFAVORABLE (illustrative)
│   ├── 4.1 Fraction of 21-analyst targets refreshed post-Q2 vs. stale         ⊗ evidence gap — resolves orphan-premium question
│   ├── 4.2 Reverse-valuation: 3.4x TBV requires low-to-mid-20s% ROTE           ⚠️C illustrative, directionally H-0-consistent
│   └── 4.3 GS/sell-side publishes fee-vs-trading decomposition, multiple moves ⊗ catalyst pending — primary resolution trigger
│
└── L1E — Peer-Relative Durability Cross-Check  ⊗ INCONCLUSIVE (genuine cross-check value)
    ├── 5.1 JPM's own CIB pattern is trading-led, not fee-led — divergent        ⚠️C informative but imprecise comparison
    ├── 5.2 MS's ISG advisory-fee sub-line vs. total ISG growth                 ⊗ evidence gap — MS tree lacks this granularity
    └── 5.3 GS's magnitude vs. JPM/MS equivalent fee-growth rates               ⊗ contingent on 5.1/5.2 resolving first

Total: 0 ✅ / 7 ⚠️ / 0 ✗ / 8 ⊗ across 15 leaves (5 branches × 3 leaves)
H-0 verdict: PARTIALLY SUPPORTED, ~50% confidence — directionally suggestive, evidence-thin, close to its own resolution frontier

VI. Key findings

Finding 1 — The advisory-fee lead is real and three-for-three, but the pattern is young and rests on an unusually strong dealmaking window

Advisory & Underwriting's growth rate has outpaced total GBM growth in every period this session obtained evidence for: FY2025 (+20.8% vs. +18.2%), Q1 2026 (+48.4% vs. +18.6%), and Q2 2026 (+55%† vs. record-but-unbroken-out). This is genuinely more evidence than a single-quarter observation, and it is the most load-bearing finding in the tree. But the window itself (record $1T H1 2026 M&A volume) is, by GS's own characterization, an unusually strong dealmaking period — and Leaf 3.2's historical base rate (GS's own 2021-2023 cycle, where industry-wide IB fee pools declined an estimated 40-50%+ from peak) is a direct reminder that fee-revenue surges of this magnitude have proven cyclical, not permanent, in GS's own recent history. Neither bulls nor bears in consensus.md explicitly control for this — the entire debate proceeds from the blended, undifferentiated GBM number.

Finding 2 — GBM's revenue-level concentration is rising even as its growth-rate composition may be shifting toward quality — these are two different claims and should not be conflated

GBM was 71.1% of FY2025 revenue and an estimated ~76%† of the Q2 2026 record — a level trend moving in the opposite direction from a "diversifying, less-cyclical franchise" narrative. At the same time, the growth-rate composition within GBM may be shifting toward the more durable, fee-based sub-line (Finding 1). Both can be true simultaneously: GS could be becoming more exposed to GBM as a whole while the quality of that GBM exposure is improving at the margin. This is the single most important nuance this tree's taxonomy-level decomposition surfaces, and it is exactly the distinction the market's undifferentiated "71-76% cyclical GBM exposure" framing (the bear case's X2 in consensus.md) fails to make.

Finding 3 — The capital-buffer compression is the tree's most acute, nearest-term-resolving tension, and it remains genuinely unconfirmed

Unlike JPM's tree (which could cite a stable 280bps CET1 buffer and a CCAR-validated $11.4B cumulative-income stress-test figure), GS's own Standardized CET1 buffer above its 11.4% requirement has compressed from >300bps to roughly 110bps in a single quarter — the thinnest of any G-SIB tree in this corpus — and the Q2 2026 trajectory (Leaf 2.1) remains unconfirmed as of this scaffold's construction date. This is not, on its own, evidence of distress: GS's 2026 CCAR outcome (SCB held flat at 3.4%) is a genuine regulatory tailwind, and the Q1 2026 RWA growth is plausibly explained by tariff-driven market-making balance-sheet deployment (a one-off), consistent with the same explanation JPM's own tree gave for its Q1 2026 credit-quality anomaly. But it is the single most concrete, near-term-testable (Q2 2026 10-Q, expected ~2026-08-01) open question in the entire tree, and this tree deliberately declines to force a premature verdict on it (Leaf 2.1: ⊗).

Finding 4 — The valuation math, even on illustrative Tier-C inputs, finds a mildly unfavorable entry — a less severe version of the pattern this corpus's JPM tree found

A simplified reverse-valuation (Leaf 4.2) suggests GS's current ~3.4x tangible-book multiple is more consistent with requiring something close to the current cyclical-peak ROTE (25.5%†) than the 15-17% through-cycle target management itself has stated. implied_prob.md's full reverse-engineering finds a mildly negative expected value (-5.9%) and a near-parity-but-unfavorable 0.94x asymmetry ratio at the $1,067 anchor — directionally the same signal as JPM's own 0.78x finding for a structurally analogous setup, though less severe (GS's gap-to-parity entry price is only ~0.6% below the current anchor, vs. JPM's ~2.1%). The orphan-premium arithmetic (price ~14% above the sell-side's own $935.74 average target) is, if anything, a more pronounced starting-point gap than JPM's own tree found.

Finding 5 — Structural franchise strength is real and should not be conflated with the cyclical-composition-and-capital-buffer question

This is not a "GS is secretly weak" thesis. GS's #1 global M&A-advisor position, five-year-high IB backlog, record AWM alternatives fundraising ($59B† Q2 2026), and accelerating dividend growth are genuine, durable competitive facts — confirmed by durability_test.md's 21/25 score with zero fatal flags. The tree's finding is narrower and more precise: these structural strengths justify some premium over BAC/peers, but the market's current pricing appears to be extrapolating the current cyclical print (both the fee-mix acceleration and the record ROTE) as the durable baseline, rather than isolating what the structural evidence actually, cautiously supports.

VII. Valuation analysis

(See peers.md, scenarios.md, and implied_prob.md for full breakdowns.)

Peer-group framing (bulge-bracket investment banks):

Peer average ~13.3x P/E / ~2.6x tangible book. GS at ~15.6x/~3.4x is a full multiple-point-plus premium — directionally consistent with (though somewhat larger than) MS's own premium, and materially above BAC's.

Sub-segment (fee-based peer) framing (per taxonomy.md's peer-group-implications hypothesis):

Read: Unlike MS (where Wealth Management's ~41% revenue share genuinely supports a partial SOTP-based premium argument, per reports/MS/leaves.md Leaf 5.3) or JPM (where AWM's disproportionate ROE provides partial SOTP support), GS's fee-based sub-segment (Advisory & Underwriting, 16.0% of revenue) and AWM (28.6%) are individually and combined too small, relative to GBM's capital-intensive core (53.3% Trading + Financing), to fully explain the current premium via a sum-of-the-parts reframe alone. GS's valuation case rests almost entirely on the within-GBM fee-vs-trading decomposition (L1A) actually being durable — there is no meaningful SOTP escape hatch the way there is at MS or (partially) at JPM.

Reverse-valuation read (Leaf 4.2, Tier C/illustrative): at a 9-11% cost-of-equity assumption, the current ~3.4x tangible-book multiple appears to require a durable ROTE close to the current cyclical-peak print (low-to-mid 20s%), not the 15-17% through-cycle target this tree's own decomposition suggests may be more representative of a normalized structural floor. This is the single most direct (if least evidence-strong) confirmation of H-0's valuation-mechanism claim.

VIII. Scenario architecture

(See scenarios.md for full analytical breakdown and implied_prob.md for the reverse-engineering math.)

Probability-weighted expected return: -5.9% (my assigned probabilities, per implied_prob.md).

Asymmetry: 0.94:1 — MILDLY UNFAVORABLE, near parity. Unlike JPM's tree (0.78x, a more decisively unfavorable finding), GS's asymmetry sits close enough to 1.0x that a shallow ~0.6% price decline would already reach neutral risk/reward. This reflects a genuinely less severe mispricing signature than JPM's — the Bull case (20%) is trimmed below the 25% default for the same "narrow conjunction of favorable outcomes" reasoning as JPM's tree, while Bear (30%) is raised above default specifically because GS's price already sits ~14% above the sell-side's own average target, a more pronounced orphan-premium starting point than JPM's own tree found.

Market-implied probabilities (reverse-engineered from current price): Bull ~37% / Base 50% / Bear ~13% — the market is assigning roughly 3x less probability to the Bear scenario than this tree's own analysis, the same directional mispricing signature this corpus's JPM tree found for a structurally analogous setup, though somewhat less extreme (JPM's market-implied Bear was ~10% vs. this tree's own 25%, a 2.5x-wider gap than GS's ~13% vs. 30%, a smaller 2.3x gap).

IX. What this means for position sizing

Position-sizing recommendation

• Current price ($1,067): 0% — WATCH ONLY. Mildly negative expected value (-5.9%) combined with a near-parity 0.94x asymmetry ratio means the math argues modestly against initiating at all, though the margin is thin enough that this is a closer call than JPM's own 0.78x finding — a small favorable data surprise (e.g., T2/RF2 firing favorably) could meaningfully shift this read.
• Contingent starter (1-2%): appropriate only if BOTH (a) Q2 2026 Form 10-Q (expected ~2026-08-01) confirms Standardized CET1 stabilizing or rebuilding (not continuing to compress — Leaf 2.1) AND (b) Q3 2026 earnings (~mid-October 2026) confirm the Advisory & Underwriting growth-rate lead persists a third consecutive quarter (Leaf 1.1) AND (c) price has moved toward or below the ~$1,024-1,061 range identified in implied_prob.md Step 5.
• Scale to 2-3%: only on confirmation of T4 (GS or a major sell-side house explicitly publishes a fee-vs-trading revenue decomposition, per mispricing.md's primary resolution trigger, with the multiple re-rating in response) AND price remaining at or below the entry band above.
• Do not initiate, regardless of price, if RF1 fires (Q3 2026 earnings show the Advisory & Underwriting growth-rate lead reversing or converging to parity with total GBM growth) until the subsequent price reaction is observed — a confirmed H-0-consistent print combined with a price that has not yet adjusted would represent an even more unfavorable entry than today.
• Hard cap 3-4% even at full confirmation — cyclical-macro-sensitive archetype, not compounder-under-review; per K.3 position-sizing convention, GS's structural quality (21/25 durability) supports a higher cap than its current cycle-position-and-capital-buffer risk should be sized to.

Concentration-risk note (per K.3.4)

GS's load-bearing macro factor is the capital-markets-and-dealmaking cycle, not the AI-capex cycle — per mispricing.md's secondary mechanism, GS should NOT be imported into the ai-capex-high/ai-capex-mid-s-curve bands at face value despite Solomon's AI-capex-cycle language; this tree tags GS uncorrelated, alongside JPM, COST, F, WMT, TSLA. This means GS does NOT compound correlated exposure with the portfolio's NVDA/TSM/AMD/AJNMY/ASML/AVGO/MU AI-infrastructure sleeve. It DOES share a distinct capital-markets-and-credit-cycle factor with any other financials names in the library — specifically JPM and MS, both already scaffolded in this corpus with structurally analogous cyclical-macro-sensitive/capital-markets-cycle H-0 theses, and V (payments network) and COIN (crypto-cycle) to a lesser, more indirect degree. If JPM and/or MS positions are already held, GS would add a third bulge-bracket-bank cyclical-financial factor exposure; per K.3.4, Ming should explicitly acknowledge this overlap in the decision-journal entry — a severe capital-markets-cycle downturn would pressure GS, JPM, and MS simultaneously via closely related (though not identical) mechanisms.

X. Triggers and red flags

(Full detail with 应对 playbooks in triggers_redflags.md.)

Triggers (confirm the market's current framing — i.e., falsify H-0):

• T1 (Q3 2026 earnings, ~mid-October 2026): Advisory & Underwriting's growth-rate lead over total GBM persists for a third consecutive period
• T2 (Q2 2026 10-Q, expected ~2026-08-01, NEAREST-TERM): Standardized CET1 rebuilds toward ≥13.5%
• T3 (H2 2026, resolves ~2027-01-15): Announced M&A deal volume sustains at or near the H1 2026 $1T pace
• T4 (timing unconfirmed): GS or a major sell-side house publishes an explicit GBM fee-vs-trading decomposition, and the multiple re-rates upward
• T5 (resolves ~2026-08-31): Full 21-analyst target distribution shows a genuine majority refreshed above $1,067

Red flags (confirm H-0):

• RF1 (Q3 2026 earnings, HIGHEST PRIORITY): Advisory & Underwriting's growth-rate lead decelerates to at or below total-GBM growth
• RF2 (Q2/Q3 2026 10-Q, NEAREST-TERM): Standardized CET1 continues compressing toward the 11.4% requirement
• RF3 (H2 2026): Announced M&A deal volume falls materially short of the H1 2026 pace
• RF4 (any quarter): A GSIB peer (JPM/MS/BAC) confirms a sector-wide capital-markets-cycle deceleration
• RF5 (ongoing): Sell-side broadly follows Oppenheimer's "the party is over" downgrade framing
• RF6 (ongoing): A material senior-PM/banker departure to a named non-bank competitor is disclosed

XI. Long-term holdability verdict

Per durability_test.md: aggregate score 21/25 (Medium-High durability, edge of High band) — identical band position to this corpus's JPM and MS trees. 0 fatal flags fired (Q3 capital allocation 4/5 ✅A, Q4 disruption survival 4/5 ✅B, balance-sheet survivability CET1 12.5% vs. 11.4% required — a ~110bps buffer, the thinnest of any G-SIB tree in this corpus, flagged explicitly rather than averaged away).

Top 3 reasons supporting 5-10 year hold:

1. Investment banking, capital-markets intermediation, and asset/wealth management at GS's G-SIB scale and #1 M&A-advisor ranking are among the most durable, regulation-protected business models available
2. Capital allocation discipline is genuinely strong (accelerating dividend growth, large sustained buyback capacity, real reinvestment in the alternatives platform and advisory-franchise depth)
3. The Advisory & Underwriting fee-growth-rate lead, if it proves durable across Q3/Q4 2026, would represent a genuine structural mix shift toward the more capital-light, higher-quality side of GS's business — a real, if not yet confirmed, upgrade path

Top 3 reasons against (entry-timing and capital-buffer concerns, not quality concerns):

1. A material share of the recent ROTE improvement is attributable to a genuinely unresolved mix of capital-markets-cycle-sensitive trading revenue and a still-young (three-quarter) advisory-fee-mix-widening pattern — if the fee lead reverses, the durable ROTE level may sit meaningfully below the current 25.5%† print
2. GS's Standardized CET1 buffer above requirement (~110bps) is the thinnest in this corpus's G-SIB cohort and has compressed 180bps in a single quarter with the Q2 2026 trajectory unconfirmed
3. Non-bank/alternative-asset-manager talent competition and private-credit disintermediation are real, structural, gradual competitive threats named directly in GS's own risk-factor disclosure

Required catalysts for upgrade to high-conviction hold: Q3/Q4 2026 earnings confirming the Advisory & Underwriting lead persists a third and fourth consecutive period; Q2/Q3 2026 CET1 data showing the buffer rebuilding rather than continuing to compress; an explicit GBM fee-vs-trading decomposition published by GS or sell-side, with the multiple re-rating in response.

Required disconfirms for downgrade: any of FF1-FF3 firing per h0_thesis.md in the H-0-consistent (normalization) direction — most critical: Q3 2026 earnings showing the fee-lead reversing, combined with continued CET1 compression.

Recommended position management: 0% at current price; re-evaluate for a 1-2% starter only if the Q2 2026 10-Q confirms CET1 stabilizing AND Q3 2026 earnings confirm the fee-lead persists AND price moves toward $1,024-1,061.

XII. Investment Scorecard (per MANUAL_en.md Part K.6 + K.10)

15-question scorecard (analytical-tree Q-list, Format B per K.3.5)

Verdict tally (M1 evidence-tier suffixes per K.3.6): 12 ✅ · 2 ⚠️ · 1 ✗ — Q1✅A, Q2✅A, Q3✅B, Q4✅B, Q5✗A (valuation genuinely unfavorable by this tree's own math, though mildly so), Q6✅A, Q7✅A, Q8✅B, Q9⚠️A (thinnest G-SIB capital buffer in this corpus), Q10⚠️B (multi-front competition), Q11✅B, Q12✅B, Q13✅A, Q14✅B, Q15✅A.

K.3.5 Weighted-score derivation

Applying the 4-tier weighting from MANUAL §K.3.5 (verdict values: ✅ = 1.0, ⚠️ = 0.5, ✗ = 0.0), Format B tier mapping (matching this corpus's JPM/MS/COST/GOOGL trees):

84.6% (rounded to 85%) sits at the very top of the 65-85% moderate-buy-with-sizing band — just below the ≥85% high-conviction threshold. This is a meaningfully different placement from JPM's 91% and reflects two real, distinguishing factors: Q9's downgrade to ⚠️A (GS's ~110bps CET1 buffer is genuinely thinner than JPM's ~280bps, a Critical-tier, 5x-weighted difference) and Q5's full zero-weighting persisting on the Load-bearing tier (the valuation problem, though milder than JPM's, is still real). Unlike JPM's case (where Q5's zero-weighting was the primary drag on an otherwise-clean scorecard), GS's score reflects two independent, genuine concerns — capital-buffer thinness AND valuation — compounding rather than one dominant issue.

Critical distinction — do NOT read 85% as a buy signal. Per MANUAL_en.md's own framing: xii_score measures structural quality (what the scorecard says about business + balance sheet + moat + valuation-as-one-input-among-fifteen); h0 measures thesis confidence (50% here — partially supported, with 8 of 15 leaves still ⊗). The gap between GS's 85% structural score and its 50% H-0 confidence, combined with the Load-bearing Q5 ✗ and the Critical-tier Q9 ⚠️, is precisely why this tree's final verdict is WATCH/0% rather than a high-conviction buy: the franchise is genuinely strong; the entry price and the capital-buffer trajectory both currently argue for patience.

Scorecard summary

Final verdict: WATCH — 0% at current price; contingent 1-2% starter on Q2 2026 CET1 confirmation + Q3 2026 fee-lead confirmation + price pullback

For Ming specifically:

• ❌ Do NOT initiate at $1,067 — mildly negative expected value and near-parity-but-unfavorable asymmetry per this tree's own reverse-engineering
• ⏳ Watch the Q2 2026 Form 10-Q (expected ~2026-08-01, ~12 days from this tree's construction) — the nearest-term, single cleanest test of the CET1-trajectory half of the thesis
• ⏳ Watch Q3 2026 earnings (~mid-October 2026) — the cleanest test of whether the Advisory & Underwriting fee-lead persists a third consecutive period
• ⚠️ If BOTH confirm favorably (Base-scenario partial-normalization-with-durable-fee-lead path) AND price moves toward $1,024-1,061: consider a 1-2% starter
• ⚠️ If the Bull scenario fires (fee-lead extends AND CET1 rebuilds AND multiple sustains): the stock is closer to fairly priced, not cheap — do not chase; re-evaluate the entire tree rather than assume "confirmation = buy"
• 🔻 If RF1+RF2 both fire (Bear scenario): wait for the subsequent price reaction before considering entry — a confirmed bear thesis without a price adjustment is a worse entry, not better
• 📅 Re-test on the Q2 2026 10-Q (~2026-08-01), Q3 2026 earnings (~mid-October 2026), and any GS investor-day or major sell-side fee-decomposition note
• 🎯 Hard cap 3-4% even at full confirmation — cyclical-macro-sensitive archetype sizing discipline, not compounder-under-review sizing; coordinate with any existing JPM/MS positions per the K.3.4 correlated-exposure note above

The 2-minute pitch:

"Goldman Sachs trades at ~$1,067 — a record quarter (ROTE 25.5% annualized, five-year-high dealmaking backlog, record $1 trillion in H1 2026 announced M&A volume) that both bulls and bears are reading as one undifferentiated 'AI-capex cycle' story. But dig into GS's own segment disclosure and there's a sharper question hiding inside: the fee-based advisory business (M&A advisory, underwriting) has grown faster than the trading business in every quarter for a year — +55% year-over-year fees in the latest quarter alone — which would normally mean GS is becoming a more durable, less risky bank. Except the capital-intensive trading engine hasn't shrunk at all as a share of revenue — it's actually grown as a share, even as its own growth rate lags the fee side — and in the same window, GS's own capital cushion above its regulatory minimum got cut by more than half in a single quarter. Neither the bulls citing the AI-cycle story nor the bears calling this a cyclical peak are actually separating these two questions. My own math on this one says the current price is a close call — modestly negative expected return, not the clean 'don't touch it' signal I found on a similar setup at JPMorgan, but not attractive enough to buy either. Watch the capital ratio in the 10-Q due in the next couple of weeks, and watch whether the fee-growth lead holds up for a third straight quarter in October — if both confirm and the price drifts down toward $1,024-1,061, that's where this becomes a starter position, not before."

Risk types most relevant (per MANUAL_en.md Part K.4):

• Valuation risk (premium-to-peer-average multiple at a cyclical-earnings peak; this tree's own math finds mildly negative EV at current price)
• Balance-sheet risk (thinnest G-SIB capital buffer in this corpus, actively compressing, Q2 2026 trajectory unconfirmed)
• Cyclical risk (capital-markets-and-dealmaking-cycle exposure is the entire L1A/L1C analytical core of this tree)
• Competition risk (multi-front: MS/JPM/BAC on GBM, advisory boutiques on the fee sub-line specifically, non-bank/alternative managers on talent)
• Correlated-factor risk (shares capital-markets-cycle exposure with any other financials names in the library — JPM, MS, V, COIN — per K.3.4)

"When NOT to buy" anti-pattern check (per MANUAL_en.md Part K.5):

• ❌ NOT buying because of a single news headline (this tree explicitly weighs a record beat-and-raise quarter against a same-window capital-buffer compression and an orphan-premium valuation gap — multiple converging signals, not one)
• ✅ NOT chasing a fast-rising stock — GS's move to an intraday all-time high the session after Q2 earnings is a real momentum signal worth naming, and this tree explicitly treats it as a caution flag (the orphan-premium finding), not a buy signal
• ❌ NOT an "AI stock" theme attachment — this tree explicitly classifies GS as cycle_exposure: uncorrelated and names the specific mechanism (Solomon's AI-capex language risks importing an availability-bias-driven category error) this tree is designed to resist
• ❌ Not a "looks cheap because the price is low" trap — the analysis is multiple-based (fwd P/E, P/TBV), not share-price-based
• ❌ Not brand-familiarity bias — the analysis is grounded in segment-level revenue-mechanism decomposition (taxonomy.md), not "Goldman Sachs is a great bank" vibes
• ✅ This tree's discipline is specifically resisting the opposite anti-pattern — the temptation to reflexively credit a genuinely strong franchise (85% structural score, #1 M&A advisor, record backlog) without separately checking whether the current price and capital-buffer trajectory offer favorable entry terms. The 85%-quality / mildly-negative-EV combination is exactly the scenario K.5 exists to guard against, applied with the same discipline this corpus's JPM tree applied to itself.

Net: 0 anti-pattern flags on the buy-side hype dimensions; 1 self-imposed discipline flag (resisting quality-bias-driven entry at a mildly unfavorable price with an unresolved capital-buffer question) — the tree's own verdict is WATCH, not buy, specifically because it applies this discipline to itself.

XIII. What's NOT in this tree (deferred / documented but not built)

• Stage 2 supplementary: historical-analogue.md (the 2021-2023 GS-specific IB-fee-cycle analogue is discussed inline in leaves.md Leaf 3.2 and peers.md Peer 6, but not built as a standalone file)
• Stage 4 supplementary: premortem-steelman.md (the bear case is steelmanned inline in consensus.md and Section VI Finding 1, but not built as a standalone file)
• Sources catalog: sources.md — full bibliography of aggregator sources (StockTitan, Yahoo Finance, Investing.com, Benzinga, and the analyst-target sources cited in consensus.md) — deferred to next refresh
• R2 primary-filing verification — the Q2 2026 Form 10-Q had not yet filed as of this scaffold's construction date (2026-07-19; earnings reported 2026-07-14); the R2 verification items listed in dashboard.md (most urgently the Q2 2026 Standardized CET1 ratio) are the highest-priority follow-up before this tree should be treated as fully load-bearing for a real capital-allocation decision
• Stage 6 living: update_{YYYY-MM-DD}.md files — first scheduled immediately after the Q2 2026 Form 10-Q files (~2026-08-01)
• External research packets: ChatGPT review packet + Codex review packet — not built this session

Last updated 2026-07-20 (Routine C, Stage 3-7 build). Source quality: mixed Tier A/B/C — FY2025/Q1 2026 primary-filing-verified, Q2 2026 WebSearch-sourced pending 10-Q confirmation. K.3.6 evidence-strength suffix convention applied to all leaf verdicts. Next refresh: immediately after the Q2 2026 Form 10-Q filing (~2026-08-01) — the single highest-priority refresh trigger in this tree, given it resolves Leaf 2.1's CET1 trajectory directly.

"Don't read news; update your tree." — 90s.PM.Investing