NYSE:SE Sea Limited

Lifecycle & sector. Sea is a high-growth business (TTM revenue +43.1% to US$27.72bn across the four filed quarters to 30 Jun 2026; Q2 2026 revenue +48.1% YoY) that is profitable but thinly so — TTM net margin 5.9%, operating margin 8.0%. Its GICS home is Consumer Discretionary (internet retail, the same shelf as Amazon), which is where Shopee's 71.8% of Q2 2026 revenue sits. But the group is a genuine conglomerate and the consolidated multiple hides it, so every metric below is scored segment by segment and the segment mix is what drives the guardrail and the growth cap in §4.

Segment scorecard — Q2 2026 (quarter ended 30 June 2026)

Segment revenue, adjusted EBITDA and the growth rates are as disclosed in Sea's Q2 2026 results (6-K filed 11 Aug 2026); prior-year comparatives are derived from the stated YoY growth rates. Segments sum to US$7.74bn against group revenue of US$7.788bn — the ~US$51m residual is other/unallocated. Segment adj EBITDA sums to US$973.2m against a group US$917.2m, i.e. roughly −US$56m of unallocated corporate cost; percentage-of-EBITDA shares above are of the US$973.2m segment total.

Universal sub-signals

Competitive moat scorecard

Moat score = 66/100 (average of the five). Two of these sub-scores — Switching Costs and Cost Advantage — are set by the competitive read below rather than assessed in the abstract.

ROIC, capital allocation and alignment

Quality = 74 (−2 vs 76 on 7 Aug). The rise in evidenced pricing power and the now-confirmed fortress balance sheet are offset by a real deterioration this report is not going to bury: group adjusted-EBITDA margin fell 4.0pp YoY, all three segments compressed at once, Monee's provisions outgrew its loan book, and per-share earnings growth has decayed to +1.4% while revenue grew 48.1%. Quality stays High because this is still a business winning share, expanding take rate and sitting on net cash — the problem is not the business, it is that its growth is no longer arriving at the earnings line the share price is a multiple of. Quality remains comfortably in the High band (≥65) — which matters mechanically, because at Quality below 65 the Decision Matrix would read Expensive-valuation as SELL rather than HOLD. It does not; but the margin of safety on that boundary is 9 points, and it narrowed this quarter.

The anchor — warranted multiple from rates, disciplined growth, and the segment mix

Actual multiple, and the score

Every basis lands in the Expensive band. The recorded ratio is 2.25× warranted on the clean forward multiple of 50.7×, which independently exceeds the 22.5× guardrail line by 125%. Note what 2.25× means: it is above 2.0×, which is the first limb of Do-Not-Buy Trigger 2's absolute arm — so that test is now breached by the Valuation pillar's own headline ratio, not merely by a supplementary figure. See §2. Score placement: the Expensive band is 0–39, and 2.25× sits deep inside it. Within that band the relative lenses are genuinely supportive (below), and embedded optionality earns the permitted +4 tilt, so the score is set near the top of the band at 32 rather than the high-20s it would otherwise take. Valuation = 32 (−6 vs 38 on 7 Aug).

Why the ratio moved so much (1.42 → 2.25) — very little of it is price. The share price rising 9.9% since 7 August accounts for about 0.13 of the move. The rest is methodology corrections this report owes the reader: the 7 August calibration recorded nonop_pct_of_net_income: 0 and clean_pe equal to trailing_pe, i.e. the step-7b earnings-quality decomposition returned nothing; and this report's own first draft then recorded the reported forward multiple under a "clean" label. It should not have — non-operating income is 19.1% of TTM net income. That run also compared an unrestated consensus multiple to a single-sector guardrail. Both are corrected here, and the correction moves the ratio against the stock.

Earnings quality — step 7b decomposition (mandatory, and it bites here)

An honest caveat in Sea's favour: most of that US$312m is recurring interest on a real US$9.10bn cash pile, not a mark-to-market gain that could reverse. The internally consistent way to credit it is to net the cash off the price rather than add the income back — do that and the clean forward multiple falls only from 50.7× to about 46× ((US$121.94 − US$11.25 of net cash per share) ÷ US$2.40), because net cash is 9.2% of the market capitalisation. It does not change the band, and 46× is still more than double the warranted 22.5×. The 13.6% haircut is applied because the framework requires scoring off operating earnings; the reader should know it is a conservative treatment of a good-quality income stream.

Sum-of-the-parts — what the three businesses are worth separately

So a run-rate sum of parts values Sea at roughly US$80–86 against a market price of US$121.94 — the market is paying a 42–52% premium to what the three businesses are worth on today's earning power. That is not a claim the stock is worth $85; it is a statement of how much future growth is already in the price. It corroborates the anchor independently, from a completely different direction.

Relative cross-checks — these rank within the band, they do not lift it out of it

FCF yield (universal anchor): TTM free cash flow of ~US$3.26bn on an enterprise value of ~US$67.7bn = 4.8% — "fair, typical of a quality growth company", and better than the earnings multiple suggests. Treat with the caveat from §3: a group containing a fast-growing lender reports flattering free cash flow.

Implied growth (narrative, not scored): at US$121.94 against a warranted 22.5× on GAAP FY2026 earnings, the price embeds roughly double the earnings stream the disciplined 15% growth assumption supports. Put the other way: to justify 50.7× on a 9.13% discount rate, near-term growth would have to run at about 22–24% for a decade rather than five years. That is not impossible for this business — it is simply the bet the buyer is making, and it should be named.

Analyst panel and independent cross-reference

The tape has turned, and it turned on the Q2 print. On 11 August 2026 the stock closed at US$131.51, up 14.6% from US$114.80 the prior session, on 15.2m shares — about 3.2× the 20-day average. It has faded for three sessions since: US$128.11 (12 Aug), US$123.32 (13 Aug), US$121.94 (14 Aug), giving back roughly 60% of the pop on declining volume, and leaving the shares 6.2% above their pre-earnings level and 7.6% below the post-print high of US$131.94. All dates and closes here are taken from the dated technical-indicator series, not from the raw price-bar labels — see §15.

Risk-reward assessment

Timing = 60 → "Improving" (≥55). Up 8 points from 52 on 7 Aug, and the upgrade is earned by three concrete changes: the earnings blackout has cleared, the Q2 gap of 11 August extended an uptrend on 3.2× volume, and relative strength is now top-band on both windows. To be precise about sequence, since it bears on how fresh the setup is: the 200-day reclaim happened earlier — price closed at US$111.15 against a 200-day average of US$110.27 on 3 August, before the last report — so what 11 August added was confirmation and volume, not the reclaim itself. It is held to 60 rather than higher by the wide stop, the hostile sector macro, and the fact that the move is three sessions old and fading. Note the mechanics: a strong Timing score cannot rescue this signal. With Valuation in the Expensive band, the Decision Matrix returns HOLD at any Timing score — good timing on an expensive name is precisely the trap the framework's valuation anchor exists to catch.

Upcoming events (next 30 days)

Recent surprises (last 7 days)

The calendar is quiet for this name and that is the point. The binary event is behind the stock, not in front of it, and nothing company-specific falls inside 30 days — which is why the catalyst-clustering score is 80 and no event-based size reduction applies. The macro items above matter to Sea through two channels and it is worth separating them. The US consumer prints of 14 August are a read-across, not a direct hit: Sea sells to the Southeast Asian consumer, so weak American demand is not evidence about its order book, and marking the name down for it would be wrong. But the long end of the US curve rising — the 30-year auction at 5.216% against 5.058% — is a direct headwind, because a high-beta emerging-market ADR on a 50.7× clean forward multiple is a long-duration asset priced against a US discount rate. In short: US macro reaches Sea through its multiple, and Southeast Asian macro reaches it through its earnings. Only one of those is in the US data.

This is the textbook higher-timeframe-bullish / lower-timeframe-pullback configuration, and normally that is a high-probability buy setup. Monthly, weekly and daily are all in uptrend and all three register a resistance breakout; the weekly MACD histogram at +5.03 on 1.8× volume is a genuine bullish cross. Price at US$121.94 sits above the 20-day (US$111.55), the 50-day (US$102.21) and the 200-day (US$108.40). Two qualifications keep it honest. First, the 50-day is still below the 200-day, so the longer-term structure has not confirmed — this is a reclaim, not a golden cross. Second, the hourly and 15-minute charts have broken down and the daily MACD histogram is already narrowing (+1.55 to +1.20 over the last two sessions), consistent with a three-session fade from the US$131.94 post-print high. Support convergence is meaningful lower down: the pre-earnings shelf at US$113–116, then a cluster at US$97.59–100.77. The level to watch on the upside is US$131.94; a close above it on volume would say the fade is over.

125 daily sessions to the Friday 14 August 2026 close of US$121.94, with the 50-day average (ending 102.21) and key levels. The vertical move near the right edge is the 11 August Q2 reaction (+14.6% to US$131.51 on 15.2m shares), followed by three fading sessions. The gap down in early March is the FY2025 result. Series drawn from the OHLCV sequence; every date cited in this report is taken from the dated indicator series, not from the price-bar labels — see §15.

Probability-weighted 12-month value = 0.18 × 168 + 0.52 × 128 + 0.30 × 84 = US$122.00 — within 0.1% of the US$121.94 close. Weights sum to 100% with the base case still most probable. Read that: on this report's own scenario set there is no expected-value edge at all. That corroborates the Do-Not-Buy in §2 — it is not its foundation. The trigger is a rule that fires on a multiple test; a near-zero expected value is independent evidence pointing the same way. Basis note: the 33×/38×/30× exit multiples used to build these three targets are on the reported GAAP forward basis (comparable with today's 43.8× reported), not the 50.7× clean basis the valuation anchor scores on — the two differ by the step-7b clean adjustment and are kept separate deliberately. Read the hierarchy of the five value numbers in this report deliberately, because they answer different questions. Our fair value is US$96 — what the framework says the business is worth on rates and disciplined growth. The base path is US$128 — where the price probably goes in twelve months, which is higher because it models a partial de-rate (43.8× → ~33×) offset by 39% EPS growth, not a move all the way to fair value. Beneath both, the run-rate sum of parts is US$80–86 and the pure anchor is US$54 on the clean basis (US$63 on the reported basis); above them, the Street is at US$146.80. The gap between our US$96 fair value and the US$128 base path is not an inconsistency — it is precisely the risk the buyer takes on: it is the market's continued willingness to pay above warranted value, and it is the first thing to disappear in a de-rating.

Forecast:

ENTRY — Fundamental group: price below fair value ~US$96.
→ FORECAST: unlikely within 6 months without a de-rating catalyst. US$96 is 21% below the current price and below both the 200-day average (US$108.40) and the pre-earnings shelf. Reaching it needs either a genuine disappointment — most plausibly Monee's NPL stepping up at the 10 November print — or a broad emerging-market risk-off event.
→ BASIS: price is trending up, not down, and the fair-value figure itself rises with delivered earnings: on FY2027 GAAP EPS of ~US$3.87 the same blended method gives roughly US$118–125 a year out, so the gap closes mostly by the fair value rising, not the price falling. That is the single most useful thing in this forecast.
→ CONFIDENCE: Moderate.

ENTRY — Technical group: already MET.
→ FORECAST: met now; at risk of un-meeting within 2–4 weeks. The daily MACD histogram has narrowed for two straight sessions (+1.55 → +1.20) and the hourly chart has broken support. A close back below the US$113–116 shelf would flip this group to unmet.
→ BASIS: three-session fade from US$131.94 on declining volume; ATR US$5.05, so the shelf is ~1.4 ATR away.
→ CONFIDENCE: Moderate — the higher-timeframe structure is intact, which argues the fade is corrective.

ENTRY — Catalyst group: MET, and decaying.
→ FORECAST: lapses by mid-September; next opportunity 10 November 2026. This condition is backward-looking and factually satisfied, but the information has been in the price since 11 August and roughly 60% of the reaction has already reversed.
→ BASIS: catalyst-dependent, not time-projectable.
→ CONFIDENCE: High on the fact; Low on it still being tradeable.

EXIT — Stop-Loss: two closes below US$107.
→ FORECAST: unlikely in 4–6 weeks. US$107 is 12.2% and 2.9 ATR below the close, beneath the 20-day average and the breakout shelf.
→ RISK TRIGGER: a broad emerging-market drawdown, or a rupiah move through 19,000.
→ CONFIDENCE: Moderate-High.

EXIT — Profit-Target: US$150 median target with RSI > 70.
→ FORECAST: possible but not likely within 3 months. US$150 is 23% up and requires clearing US$131.94 first. On the four-week rate of ascent it is reachable inside a quarter, but that rate is decaying.
→ CONFIDENCE: Low.

EXIT — Thesis Invalidation: the one to actually diarise.
→ FORECAST: first real test 10 November 2026. The condition is Monee's 90-day NPL at or above 1.5%, or a third consecutive quarter of provisions growing faster than the loan book.
→ BASIS: provisions +71.5% against book +62.5% in Q2 2026 — the ratio is already moving the wrong way, and the 2026 vintages season into the 90-day bucket over the next two quarters.
→ CONFIDENCE: Moderate that the ratio deteriorates further; Low that the NPL itself breaches 1.5% by November.

Forecast: No exit trigger is live, so exit_action is Hold. Read that precisely, because the two audiences get different answers: for an existing holder it means hold — nothing has broken, no stop is hit and the business is performing. For a non-holder the signal is DO NOT BUY, not Hold: a Do-Not-Buy trigger fired in §2 and governs new capital, while these exit rules govern an existing position. The Do-Not-Buy is about valuation, not deterioration. The Thesis-Invalidation group is deliberately de-conjuncted at 2-of-N and now carries an explicit credit condition and an explicit competitive condition, both propagated from §3. Neither is live today; the credit one is the closer of the two.

Position sizing is not computed — no portfolio allocation or role was specified for this analysis, and the framework does not invent one. What can be given is the context that would feed it.

A Do-Not-Buy trigger fired in §2, so there is no position to size. Everything below is context for why, and for what a holder rather than a buyer faces. Note the framework keeps two questions apart: a DNB governs new capital, while §12's exit rules govern an existing position — and no exit rule is live (no stop hit, no thesis condition fired, profit target 23% away), which is why exit_action remains Hold. A holder is not being told to sell; a buyer is being told not to start.

The §12 Conviction Ladder reads Full-Size (2 of 3 entry paths open), which is mechanically correct and substantively void. The ladder measures how many independent reasons to enter exist, which sets position size; it is not permission to enter. Two paths are genuinely open — the tape turned decisively (Technical) and the earnings event confirmed it (Catalyst). The path that is shut is the Fundamental one, price versus value, and that is exactly the one the signal keys off. So the honest summary is: the tape and the catalyst say go, the price says no, and the price wins. Under a DNB the ladder confers no permission at all. Note also that the Catalyst path has already been paid for — the +14.6% gap was three sessions ago and about 60% of it has reversed — so in practice this is closer to a one-path setup than the count suggests.

Volatility context. ATR is US$5.05, or 4.1% of price per day — roughly four times a typical large-cap. Beta is 1.51, so any position acts like about 1.5× its size in portfolio-risk terms: a 4% holding carries the market risk of a 6% one. The 52-week range is US$77.05–US$199.30, a 61% drawdown peak to trough within the last year, and the shares sit at 36.7% of that range. Catalyst clustering is 80 (calm), so no event-based reduction applies. Anyone sizing this name should size it for a 60% drawdown, because it has just delivered one.

{
  "ticker": "SE",
  "date": "2026-08-16",
  "version": "v6",
  "brand": "Sea",
  "exchange": "NYSE",
  "exchange_ticker": "NYSE:SE",
  "isin": "US81141R1005",
  "api_ticker": "SE",
  "company": "Sea Limited",
  "currency": "USD",
  "sector": "Consumer Discretionary",
  "sub_industry": "SE-Asia Internet (EM)",
  "lifecycle_stage": "high-growth",
  "user_context": {
    "horizon": null,
    "allocation_pct": null,
    "portfolio_role": null
  },
  "user_horizon": null,
  "user_allocation_pct": null,
  "portfolio_role": null,
  "price_at_rating": 121.94,
  "price_asof": "2026-08-14 close (report dated 2026-08-16, a weekend; Friday's close is the latest print)",
  "signal_short": "DO NOT BUY",
  "signal_medium": "DO NOT BUY",
  "signal_long": "DO NOT BUY",
  "primary_signal": "DO NOT BUY",
  "composite_short": 55.8,
  "composite_medium": 55.1,
  "composite_long": 59.3,
  "quality_score": 74,
  "valuation_score": 32,
  "timing_score": 60,
  "driver_score": 59,
  "overall_confidence": 60,
  "quality_confidence": 68,
  "valuation_confidence": 62,
  "timing_confidence": 60,
  "driver_confidence": 55,
  "quality_detail": {
    "industry_benchmark_name": "Rule of 40 (Q2 2026: revenue growth + adj EBITDA margin)",
    "industry_benchmark_value": 59.9,
    "industry_benchmark_score": 88,
    "moat_score": 66,
    "moat_pricing_power": 72,
    "moat_network_effects": 85,
    "moat_switching_costs": 42,
    "moat_cost_advantage": 62,
    "moat_intangibles": 70,
    "roic_percentile_vs_peers": 75,
    "capital_allocation": 62,
    "management_skin_in_game": 85,
    "roic_pct": 24.0,
    "roic_pct_undeducted": 9.7,
    "roe_ttm_pct": 12.6
  },
  "valuation_detail": {
    "fcf_yield": 4.8,
    "consensus_growth_rate": 24.7,
    "implied_growth_rate": 22.5,
    "historical_valuation_decile": 3,
    "sotp_per_diluted_share": 80.0,
    "sotp_per_basic_share": 85.6,
    "anchor_value_per_share": 54.0,
    "anchor_value_per_share_reported_basis": 62.6,
    "blended_fair_value": 96.0,
    "blended_fair_value_clean_basis": 88.0,
    "fair_value_basis_note": "anchor_value_per_share is 22.5 x clean forward EPS 2.40 = 54.0, matching the basis warranted_ratio uses. On the reported GAAP basis (22.5 x 2.78) it is 62.6. fair_value_est carries 96.0, the REPORTED-basis blend, deliberately: it is the more generous of the two (clean-basis blend 88.0), and a DNB that fires while using the charitable fair value is firmer than one needing the punitive one."
  },
  "relative_strength_vs_spy": 12.8,
  "relative_strength_vs_sector": 14.8,
  "changes_since_last_report_summary": "Versus the 2026-08-07 report: quality_score fell 76 -> 74; valuation_score fell 38 -> 32; timing_score rose 52 -> 60; driver_score fell 66 -> 59 (label Tailwind -> Neutral); economic_alignment_conviction fell 48 -> 42; warranted_ratio rose from 1.42 to 2.25; entry_groups_met rose 0 -> 2 and entry_conviction moved Wait -> Full-Size as the Technical and Catalyst paths opened on the Q2 breakout while the Fundamental path stayed shut; short_entry_confirmed moved false -> true; nonop_pct_of_net_income rose 0 -> 19.1 correcting a step-7b decomposition that wrongly returned zero; stop_loss rose from 83 to 107; fair_value_est fell from 121 to 96. All three horizon signals FLIPPED from HOLD to DO NOT BUY as do_not_buy_triggers went from empty to Trigger 2 arm (a) and hard_gate_state from caution to donotbuy; total_debt_usd_bn was restated from 1.22 to 2.22.",
  "timing_detail": {
    "mtf_confluence": 67,
    "risk_reward_score": 45,
    "relative_strength_score": 92,
    "relative_strength_vs_spy": 12.8,
    "relative_strength_vs_sector": 14.8,
    "catalyst_clustering_score": 80,
    "dynamic_macro_weight": 0.15,
    "atr": 5.05,
    "atr_pct_of_price": 4.1,
    "pct_of_52wk_range": 36.7
  },
  "economic_alignment_stance": "Contrarian",
  "economic_alignment_conviction": 42,
  "economic_alignment_pressure": "Headwind",
  "economic_alignment_source": "sector-map (EBITDA-weighted XLY SU/U/N + XLF N/U/N + XLC U/N/N)",
  "macro_report_date": "2026-08-12",
  "val_multiple_basis": "clean forward P/E (FY2026, GAAP-restated consensus, step-7b clean earnings base)",
  "warranted_multiple": 22.5,
  "actual_multiple": 50.7,
  "warranted_ratio": 2.25,
  "risk_free_10y": 4.63,
  "risk_free_source": "FRED DGS10 at 2026-08-13 (macro report prose carries 4.70%; sensitivity stated in section 4 \u2014 ratio 2.25 either way)",
  "equity_risk_premium": 4.5,
  "risk_add_on": 0.0,
  "discount_rate_r": 9.13,
  "g_near": 15.0,
  "g_term": 3.0,
  "warranted_multiple_raw": 27.7,
  "sector_guardrail_line": 22.5,
  "sector_guardrail_basis": "SOTP: Shopee 24x (Cons Disc) / Monee 16x (deposit-funded lender) / Garena 26x (Comm Svcs), weighted by Q2 2026 segment adj-EBITDA share 26.2/29.6/44.2. Single-sector Cons-Disc line 24x gives ratio 2.11 on the clean multiple (50.7/24) - same Expensive band AND still above the 2.0x DNB limb, so arm (a) fires on the SKILL published guardrail too, not only on this bespoke SOTP line.",
  "val_band": "expensive",
  "currency_frame": "single USD frame - Sea reports AND trades in USD; no FX rate applied or recorded anywhere in this report",
  "eps_trailing": 2.59,
  "trailing_pe": 47.08,
  "eps_forward_gaap_restated": 2.78,
  "forward_pe": 43.8,
  "forward_pe_unrestated_consensus": 35.0,
  "consensus_to_gaap_realisation_factor": 1.253,
  "nonop_pct_of_net_income": 19.1,
  "clean_eps_trailing": 2.24,
  "clean_pe": 54.5,
  "clean_forward_pe": 50.7,
  "clean_peg": 2.05,
  "peg_fwd": 1.15,
  "fcf_yield": 4.8,
  "enterprise_value_usd_bn": 67.7,
  "market_cap_usd_bn": 74.57,
  "shares_basic_m": 611.5,
  "shares_diluted_m": 654.5,
  "total_debt_usd_bn": 2.22,
  "total_debt_basis": "30 Jun 2026 balance sheet, ALL THREE tranches: current borrowings 316.2m + convertible notes 996.3m + non-current borrowings 908.2m = 2,220.7m. An earlier draft omitted the non-current tranche and understated debt by 908.2m; corrected post-audit. NOT the lease-inclusive provider totalDebt field, but note FMP debtToMarketCapTTM implied ~2.2bn and that read was correct.",
  "net_cash_usd_bn": 6.88,
  "net_cash_pct_of_market_cap": 9.2,
  "cash_and_st_investments_usd_bn": 9.1,
  "segment_q2_2026": {
    "shopee": {
      "revenue_usd_m": 5590,
      "revenue_yoy_pct": 48.2,
      "adj_ebitda_usd_m": 255.4,
      "adj_ebitda_yoy_pct": 12.2,
      "adj_ebitda_margin_pct": 4.6,
      "adj_ebitda_margin_py_pct": 6.0,
      "gmv_usd_bn": 38.3,
      "gmv_yoy_pct": 28.4,
      "gross_orders_bn": 4.2,
      "core_marketplace_rev_yoy_pct": 65.6,
      "take_rate_pct": 11.2,
      "pct_of_segment_ebitda": 26.2
    },
    "monee": {
      "revenue_gross_usd_m": 1400,
      "revenue_gross_yoy_pct": 58.9,
      "provision_usd_m": 555.2,
      "provision_yoy_pct": 71.5,
      "revenue_net_of_provision_usd_m": 845,
      "revenue_net_yoy_pct": 51.1,
      "adj_ebitda_usd_m": 288.0,
      "adj_ebitda_yoy_pct": 12.8,
      "adj_ebitda_margin_pct": 20.6,
      "adj_ebitda_margin_py_pct": 29.0,
      "loan_book_usd_bn": 11.1,
      "loan_book_yoy_pct": 62.5,
      "npl_90day_pct": 1.0,
      "provision_annualised_pct_of_book": 20.0,
      "pct_of_segment_ebitda": 29.6
    },
    "garena": {
      "revenue_usd_m": 746.6,
      "revenue_yoy_pct": 33.5,
      "bookings_usd_m": 763.5,
      "bookings_yoy_pct": 15.5,
      "adj_ebitda_usd_m": 429.8,
      "adj_ebitda_yoy_pct": 16.7,
      "adj_ebitda_margin_pct": 57.6,
      "adj_ebitda_margin_py_pct": 65.9,
      "deferred_rev_change_usd_m": 16.9,
      "deferred_rev_change_yoy_pct": -83.4,
      "pct_of_segment_ebitda": 44.2
    },
    "group": {
      "revenue_usd_m": 7787.8,
      "revenue_yoy_pct": 48.1,
      "adj_ebitda_usd_m": 917.2,
      "adj_ebitda_yoy_pct": 10.6,
      "adj_ebitda_margin_pct": 11.8,
      "adj_ebitda_margin_py_pct": 15.8,
      "net_income_usd_m": 440.8,
      "net_income_continuing_usd_m": 458.1,
      "eps_diluted": 0.7
    }
  },
  "revenue_growth_ttm_pct": 43.1,
  "revenue_growth_fy2025_pct": 36.4,
  "revenue_growth_fy2025_basis": "FY2025 22,938.5m (sum of four filed quarters) vs FY2024 16,819.9m = +36.4%. An audit query proposed +48.3% on an FY2025 base of 24,938.5m, which is ~2,005m above the filed sum and arithmetically impossible given quarterly YoY rates of +29.6/+38.2/+38.3/+38.4% (a full-year rate cannot fall outside the range of its quarterly rates). The FY2024 cross-reconciliation agrees to within a rounding step, not exactly: implied Q1 2024 3,735.5 vs filed 3,734.329. Not adopted - see section 15.",
  "eps_diluted_yoy_by_quarter": {
    "Q3 2025": 126.9,
    "Q4 2025": 61.5,
    "Q1 2026": 6.3,
    "Q2 2026": 1.4
  },
  "eps_diluted_ttm_yoy_pct": 31.5,
  "conversion_note": "Q2 2026 cascade: revenue +48.1% -> adj EBITDA +10.6% -> net income +6.4% -> diluted EPS +1.4%. Diluted share count +10.5% YoY on convertible dilution. This is the evidence that defeats the DNB Trigger 2 arm (a) growth carve-out.",
  "eps_fy2026_vs_fy2025_pct": 9.4,
  "competitive_share_trajectory": "stable",
  "competitive_threat_level": "elevated",
  "competitive_evidence": "CY2025 (last measurable period): Shopee ~59% of SEA platform GMV, TikTok Shop ~31% (from ~23%), Lazada ~9% (from ~16%). NO period-matched Q2 2026 rival GMV exists publicly (ByteDance private; Alibaba does not disclose Lazada GMV quarterly) so no Q2 2026 share claim is made. Q2 2026 company evidence: Shopee held volume (GMV +28.4%, orders +27.5%) while segment adj-EBITDA margin fell 6.0%->4.6% - competition shows up in MARGIN, not volume.",
  "driver_commodity_trend": null,
  "tail_risk_adjudication": {
    "register_source": "MacroDriver-state-20260812.json - FOUR tails, two live",
    "1_sp500_concentration_ai_unwind": {
      "status": "armed, trigger receding",
      "ruling": "does not apply - not in the cohort. Not an S&P 500 constituent (foreign private issuer ADR), no AI-capex or AI-monetisation leverage, non-operating income is treasury interest on a real 9.10bn cash pile not AI mark-ups."
    },
    "2_private_credit_crack": {
      "status": "building - NOT armed",
      "ruling": "does not apply on two independent grounds. Status: inheritance requires an ARMED tail. Cohort: it describes US institutional private credit (HYG below both MAs, BCRED gated in June); Monee is on-balance-sheet SEA consumer/SME lending funded by SeaBank deposits, no leveraged loans or fund vehicles."
    },
    "3_hormuz_closure_escalation": {
      "status": "LIVE",
      "ruling": "materially relevant to SE and named as the dominant business risk in sections 5 and 11 - but does NOT fire arm (b). (i) Its own trigger is not met: Brent 88.58 vs a trigger of >100 sustained; a live-but-not-triggering tail is treated as a loud bear leg plus a Valuation-Ceiling cap, not a DNB. (ii) It reaches SE only through the consumer/credit chain ALREADY scored in the section 5 Driver (marked down 66 -> 59, Tailwind -> Neutral partly for this) and already the primary section 11 bear trigger, so counting it again would double-count the same mechanism. Not an index-level multiple-compression cohort for a company with no direct energy P&L line."
    },
    "4_fed_hikes_september": {
      "status": "LIVE",
      "ruling": "does not fire arm (b) - already inside the anchor. ~44% market-implied is a macro path scenario, not an index-level de-rating cohort, and would fire a DNB on every rate-sensitive name. The mechanism is captured in r = 9.13% on a 4.63% 10-Y, with a 4.70% sensitivity run in section 4 (raw warranted 27.37x vs 27.70x; the guardrail binds either way, ratio 2.25x unchanged)."
    },
    "net": "Arm (b) does not fire on any of the four. The DO NOT BUY rests entirely on arm (a)."
  },
  "hard_gate_state": "donotbuy",
  "gates_triggered": [
    "Valuation-Ceiling (Gate 3) - Expensive band: clean forward P/E 50.7x vs warranted 22.5x = ratio 2.25 >= 1.40, AND actual 50.7x >= the 22.5x SOTP sector guardrail line. Would cap all three horizons at HOLD, but is superseded by the Do-Not-Buy below. Price is NOT above the highest analyst target (156) so that arm does not fire.",
    "DNB Trigger 2 absolute arm (a) - Valuation Extreme: clean forward multiple 50.7x is >= 2.0x warranted (45.0x) AND >= 1.5x the sector guardrail (33.75x). The growth carve-out written into the trigger does not save it: the growth that could excuse breaching an earnings multiple is growth in the stream that multiple capitalises, and on that stream Sea is NOT EXCEPTIONAL - clean EPS +7% and GAAP EPS +9.4%, both BELOW the already-haircut 15% g_near that produced the 22.5x benchmark."
  ],
  "gates_caution": [
    "Accounting/earnings-quality (Gate 4) - non-operating income 19.1% of TTM net income: above the 15% recompute threshold, below the ~30% gate backstop; clean figures carry the score",
    "Regulatory (Gate 5) - chronic EM digital-lending / e-commerce regulation, nothing dated or binary",
    "Monee credit-cycle concentration - 29.6% of segment adj EBITDA off an 11.1bn loan book with provisions +71.5% vs book +62.5%",
    "FX translation - SEA/BRL revenue into USD reporting; rupiah at record lows through 18,000/USD",
    "Competitive intensity - TikTok Shop 23%->31% of SEA platform GMV in CY2025; threat is to margin not volume"
  ],
  "do_not_buy_triggers": [
    "Trigger 2 arm (a) - Valuation Extreme (absolute, deep-expensive): clean forward P/E 50.7x = 2.25x warranted 22.5x (>= 2.0x limb) and >= 1.5x the 22.5x sector guardrail (33.75x). The trigger conditions its carve-out on there being no exceptional, proven, durable growth, and supplies no noun - growth of WHAT. The framework answers three times that it is the stream the multiple capitalises: the warranted multiple is computed to capitalise that stream (fair multiple ~ 1/(r-g)), g_near is built from a consensus EPS CAGR, and the only other use of the same word-pair in the SKILL attaches it to that same earnings-growth input. So the growth that could excuse breaching an earnings multiple is EARNINGS growth - and on that stream Sea is NOT EXCEPTIONAL: clean EPS 2.24 -> 2.40 = +7%, reported GAAP EPS FY2025 2.54 -> FY2026 2.78 = +9.4%, both BELOW the already-haircut 15% g_near that produced the 22.5x benchmark, and decaying. A name cannot claim an exception from a multiple test on the very growth rate the test already credited it with, at a level beneath what it was credited. Revenue growth IS exceptional, proven and durable (+48.1% YoY Q2 2026, +43.1% TTM, +36.4% FY2025) but revenue is not the stream a P/E capitalises. Corroboration, not the argument: diluted EPS YoY fell +126.9% (Q3 2025) -> +61.5% (Q4 2025) -> +6.3% (Q1 2026) -> +1.4% (Q2 2026) while revenue growth accelerated, with group adj EBITDA margin 15.8% -> 11.8% and all three segment margins compressing. Deliberately NOT relied on: the un-haircut 24.7% consensus CAGR (the anchor already haircut it 25% and capped it at 15%; re-using it here would double-count growth against the discipline that produced the warranted multiple) and the TTM EPS gain of +31.5% (which sits entirely in the two oldest quarters, lapping a depressed 2024 base). Overrides the Gate 3 HOLD cap; sets all three horizons to DO NOT BUY."
  ],
  "dnb_trigger2_test": "FIRED. Arm (a) - Valuation Extreme, absolute/deep-expensive. NUMERIC TEST, both limbs satisfied on the actual CLEAN multiple of 50.7x: >= 2.0x warranted (45.0x) AND >= 1.5x the sector guardrail (33.75x). On the SKILL table single-sector Consumer-Discretionary line of 24x the ratio is 2.11x, so limb 1 fires on the framework published guardrail too, not only on this report bespoke 22.5x SOTP line. THE EXCEPTION - growth of WHAT: arm (a) supplies no noun. The framework answers three times: (i) the trigger measures against the warranted multiple, computed to capitalise the stream the multiple is OF (fair multiple ~ 1/(r-g)); (ii) g_near is min(0.75 x consensus forward growth, sector cap) and the consensus series used is an EPS CAGR; (iii) the only other use of the same word-pair in the SKILL is a proven, durable >20 percent grower may use up to 20 percent, attached to that same earnings-growth input. So the growth that can excuse breaching an earnings multiple is EARNINGS growth. On that stream SE is NOT EXCEPTIONAL: clean EPS 2.24 -> 2.40 = +7 percent, reported GAAP EPS FY2025 2.54 -> FY2026 2.78 = +9.4 percent - both BELOW the already-haircut 15 percent g_near that produced the 22.5x benchmark, and decaying. A name cannot claim exemption from a multiple test on the very growth rate the test already credited it with, at a level beneath what it was credited. Revenue growth IS exceptional, proven and durable (+48.1 percent YoY Q2 2026, +43.1 percent TTM, +36.4 percent FY2025) but revenue is not the stream a P/E capitalises. CORROBORATION (not the argument): diluted EPS YoY +126.9 -> +61.5 -> +6.3 -> +1.4 percent across four filed quarters while revenue growth accelerated. COUNTER-EVIDENCE DISCLOSED AND SURVIVED: the effective tax rate stepped 25.8 -> 35.4 percent so pre-tax income grew +27.0 percent and operating income +28.4 percent; and Q2 2025 diluted shares of 592.02m were exactly equal to basic (anomalous vs 634.6m and 635.6m either side) so the +10.5 percent diluted share count is partly a comparator artefact. It survives because Q1 2026 is share-count-clean (+0.24 percent) and still only +6.3 percent, the tax step-up has persisted two quarters so it is a level shift not a one-off, and the test runs on clean earnings per share at ~7 percent against g_near 15 percent. NOT RELIED ON: the un-haircut 24.7 percent consensus CAGR (the anchor already haircut it 25 percent and capped it at 15 percent; re-using it to disapply the trigger would double-count growth against the discipline that produced the benchmark) and the TTM EPS gain of +31.5 percent (which sits entirely in the two oldest quarters, lapping a depressed 2024 base). ARM (b) DOES NOT FIRE: see tail_risk_adjudication - all FOUR tails named and ruled on individually; SE is not in the AI-concentration cohort, the private-credit tail is building not armed and describes US institutional credit, the LIVE Hormuz tail is the dominant business risk but its own Brent>100 trigger is unmet and it reaches SE only via the consumer/credit chain already scored in the Driver, and the LIVE September-Fed tail is a sub-50-percent macro path already inside r=9.13 percent. CLEARING PRICES: ~108 clears the 2.0x limb only (11 percent below spot, essentially the 200-DMA of 108.40); ~81 clears the binding 1.5x guardrail limb; or clean forward EPS would have to reach 3.61 (+50 percent) at today price.",
  "entry_groups_met": 2,
  "entry_conviction": "Full-Size",
  "exit_groups_live": 0,
  "exit_action": "Hold",
  "short_entry_confirmed": true,
  "short_hold_reason": null,
  "short_cap_reason": null,
  "short_hold_reason_note": "Removed deliberately: the field is scoped to a short HOLD or the sanctioned half-size starter, and signal_short is now DO NOT BUY. On the base matrix alone the short would have been HOLD for reason expensive (High Quality 74 + Expensive Valuation 32), then Gate 3 would have capped it at HOLD; the Do-Not-Buy overrides both.",
  "short_hold_note": "The short technical-confirmation cap did NOT fire - the base matrix returned HOLD directly (High Quality 74 + Expensive Valuation 32 -> HOLD, great business wrong price), so there was no BUY to cap. short_entry_confirmed is true because the Technical entry group IS met. The short quality-starter override is explicitly barred: it requires Valuation >= 40 and not Expensive. Final override: Do-Not-Buy Trigger 2 arm (a) supersedes the whole chain on all three horizons.",
  "fair_value_est": 96.0,
  "stop_loss": 107.0,
  "target_price": 128.0,
  "scenario_base_target": 128,
  "scenario_bull_target": 168,
  "scenario_bear_target": 84,
  "scenario_probabilities": {
    "bull": 18,
    "base": 52,
    "bear": 30
  },
  "scenario_weighted_value": 122.0,
  "scenario_weight_note": "Verified against the COMMITTED 7 Aug report section 11, whose prior register was Bull 162 (22%) / Base 125 (50%) / Bear 84 (28%). Versus that register: bull CUT 22 -> 18, base broadly HELD 50 -> 52, bear RAISED 28 -> 30, on the conversion evidence (the base case needs the per-share earnings re-acceleration consensus expects; the last two filed quarters delivered +6.3% and +1.4%). An earlier draft of this calibration wrongly stated the prior register as 22/55/23 and therefore described the base weight as cut, which was directionally wrong. Probability-weighted value 122.00 vs a 121.94 price = no expected-value edge, which corroborates the DNB rather than founding it.",
  "scenario_probabilities_prior": {
    "bull": 22,
    "base": 50,
    "bear": 28
  },
  "analyst_consensus_target": 146.8,
  "analyst_target_high": 156,
  "analyst_target_low": 125,
  "analyst_target_median": 150,
  "analyst_target_upside_pct": 20.4,
  "analyst_grades_consensus": "Buy",
  "analyst_bullish_pct": 70.5,
  "analyst_coverage_count": 44,
  "analyst_recent_target_count_qtr": 2,
  "analyst_recent_target_avg": 154.5,
  "fmp_rating": "B",
  "fmp_overall_score": 3,
  "recent_upgrades_30d": 0,
  "recent_downgrades_30d": 0,
  "next_update_date": "2026-08-31",
  "next_check_date": "2026-08-31",
  "next_update_basis": "default +14d from 2026-08-16, rolled to a trading day (2026-08-30 was a Sunday). Two dated events DO fall inside the window - the 19 Aug FOMC minutes (the live September-hike tail trigger) and the 20 Aug macro-report refresh - but neither is a company-specific catalyst of the kind the scheduling rule keys on, and recurring macro releases are excluded by rule except for high-macro-sensitivity sectors inside the 3-day WAIT window, which SE (Medium sensitivity) is not. Next company event: earnings 2026-11-10.",
  "next_earnings_date": "2026-11-10",
  "analysis_status": "on-going",
  "finder_ticker": "SE",
  "finder_exchange": "\ud83c\uddfa\ud83c\uddf8 NYSE",
  "delta_vs_prior": {
    "prior_report": "2026-08-07",
    "prior_price": "was US$111.00 at the report of 2026-08-07",
    "price_change_pct": 9.9,
    "signals": "FLIPPED on all three horizons: HOLD -> DO NOT BUY (prior calibration-SE-20260807-0755.json carried HOLD/HOLD/HOLD)",
    "moved_keys": {
      "quality_score": "76 -> 74 (fell)",
      "valuation_score": "38 -> 32 (fell)",
      "timing_score": "52 -> 60 (rose)",
      "driver_score": "66 -> 59 (fell; label Tailwind -> Neutral)",
      "economic_alignment_conviction": "48 -> 42 (fell)",
      "warranted_ratio": "rose from 1.42 to 2.25",
      "actual_multiple": "rose from 34 to 50.7 (now on the step-7b CLEAN basis the anchor requires, not the reported basis)",
      "signal_short": "HOLD -> DO NOT BUY",
      "signal_medium": "HOLD -> DO NOT BUY",
      "signal_long": "HOLD -> DO NOT BUY",
      "do_not_buy_triggers": "empty -> Trigger 2 arm (a) Valuation Extreme",
      "hard_gate_state": "caution -> donotbuy",
      "total_debt_usd_bn": "restated from 1.22 to 2.22 (non-current borrowings of 908.2m had been omitted)",
      "scenario_probabilities": "moved from the 7 Aug published 22/50/28 to 18/52/30: bull cut 22->18, base held 50->52, bear raised 28->30",
      "warranted_multiple": "24 -> 22.5 (fell; SOTP guardrail replaces the single-sector line)",
      "entry_groups_met": "0 -> 2 (rose)",
      "entry_conviction": "Wait -> Full-Size",
      "short_entry_confirmed": "false -> true",
      "overall_confidence": "40 -> 60 (rose)",
      "nonop_pct_of_net_income": "0 -> 19.1 (rose; the prior run's step-7b decomposition returned zero in error)",
      "clean_pe": "43 (= trailing_pe, i.e. no decomposition) -> 54.5",
      "analyst_consensus_target": "133.6 -> 146.8 (rose)",
      "macro_report_date": "2026-07-30 -> 2026-08-12",
      "fmp_rating": "changed from B+ to B",
      "stop_loss": "raised from 83 to 107",
      "fair_value_est": "fell from 121 to 96",
      "scenario_base_target": "rose from 125 to 128",
      "scenario_bull_target": "rose from 162 to 168",
      "scenario_bear_target": "unchanged at 84"
    },
    "unchanged_verified": [
      "competitive_share_trajectory stable",
      "competitive_threat_level elevated",
      "analysis_status on-going",
      "exit_action Hold",
      "val_band expensive",
      "currency USD"
    ],
    "gate_changes": "Gate 2 Earnings-Event Risk CLEARED (Q2 reported 11 Aug; next print 10 Nov). Gate 3 Valuation-Ceiling remains TRIGGERED and is now double-confirmed on both the ratio and the guardrail arm. NEW: Do-Not-Buy Trigger 2 arm (a) FIRED, which supersedes the Gate 3 HOLD cap and sets all three horizons to DO NOT BUY. hard_gate_state moves caution -> donotbuy.",
    "headline": "Q2 2026 was a big beat on revenue (+48.1%) - but diluted EPS grew just 1.4%, adjusted EBITDA 10.6%, and all three segment margins compressed. The stock rose 9.9% while the clean multiple went to 2.25x its warranted level, and Do-Not-Buy Trigger 2 arm (a) fired: the growth is exceptional and proven but does not reach the earnings line the multiple prices. All three horizons move from HOLD to DO NOT BUY."
  },
  "framework_amendment_proposed": "RECORDED FOR THE SKILL OWNER, NOT SELF-APPLIED. DNB Trigger 2 arm (a) conditions its carve-out on there being no exceptional, proven, durable GROWTH but supplies no noun, so the adjudication has to be reconstructed from three indirect signals (the warranted-multiple identity, the g_near definition, and the L445 >20-percent-grower wording). Proposed text: ...with no exceptional, proven, durable growth IN THE EARNINGS STREAM THE MULTIPLE CAPITALISES. This would stop a Do-Not-Buy ever turning on a lexical argument about an adjective, which is exactly what happened on SE 2026-08-16.",
  "calibration_notes": "POST-AUDIT ROUNDS 1-3: this report failed its first independent audit and the signal moved HOLD -> DO NOT BUY. Corrections made: (a) the armed-tail register was incomplete (FOUR tails, two live) so DNB arm (b) was never actually tested - all four are now enumerated and ruled on by name, and arm (b) still does not fire; (b) the arm (a) growth carve-out was re-stated on corrected inputs, the un-haircut 24.7% consensus CAGR was dropped as inadmissible double-counting, conversion was put on the record, and the carve-out FAILED - hence the DNB; (c) total debt understated by 908.2m, re-propagated across net cash, EV, FCF yield, SOTP per share, ROIC and Gate 1; (d) actual_multiple/warranted_ratio moved onto the step-7b clean basis (50.7x / 2.25x); (e) section 9 daily row was one session stale; (f) the year-ago Rule of 40 was 54.0 not 59.3, so the composite improved rather than held. NOT adopted: an audit request for FY2025 revenue growth of +48.3%, which is arithmetically impossible against the filed quarters - see section 15. Separately, three corrections to the 2026-08-07 report are disclosed in this run: (1) the step-7b earnings-quality decomposition returned nonop_pct_of_net_income 0 and clean_pe == trailing_pe, when non-operating income is in fact 19.1% of TTM net income; (2) the actual multiple was taken from an unrestated consensus EPS that is not on a GAAP diluted basis; (3) the guardrail was the single-sector Consumer-Discretionary line rather than a segment-weighted one for a three-business conglomerate. All three corrections move the ratio AGAINST the stock."
}