NVDA / HORIZON-09 / 2026-10-07
H09：实际价格与分析师共识双重检验H09: dual tests against realized prices and analyst consensusH09：實際價格與分析師共識雙重檢驗
输入截面2026-10-06收盘，本轮2026-10-07纽约08:06再次核验收盘$239.24。原财报、利率与完整历史序列沿用冻结资料，不称全部刷新到10月7日。Input cutoff is the2026-10-06 close, rechecked on2026-10-07 at08:06 New York:$239.24. Preserve frozen financial, rate and historical series inputs; this is not a full refresh to October7.輸入截面2026-10-06收盤，本輪2026-10-07紐約08:06再次核驗收盤$239.24。原財報、利率與完整歷史序列沿用凍結資料，不稱全部刷新到10月7日。
H09同比日收入增长候选未采用：三年共识差异改善，但十年12M实际价格误差与EPS误差增大。保留H07／H03市场P/E不变参考。两项检验分开保留，不把接近分析师共识称为实际价格准确率。H09 daily-revenue YoY candidate is not adopted: three-year consensus gaps improve, but decade12M realized-price and EPS errors worsen. Retain the H07/H03 held-market-P/E reference. Preserve both evaluations separately; closeness to analysts is not realized-price accuracy.H09同比日收入增長候選未採用：三年共識差異改善，但十年12M實際價格誤差與EPS誤差增大。保留H07／H03市場P/E不變參考。兩項檢驗分開保留，不把接近分析師共識稱為實際價格準確率。
保留的当前参考输出：H07 v0.1零权重Retained current reference: H07 v0.1 zero weight保留的當前參考輸出：H07 v0.1零權重
| 期限Horizon期限 | 目标日 → 盈利窗口末日Target → earnings-window end目標日 → 盈利窗口末日 | Core EPS | P/E | 计算输出 USD/股Calculated USD/share計算輸出 USD/股 |
|---|---|---|---|---|
| 0M | 2026-10-06 → 2027-10-06 | 13.88266 | 17.23301× | $239.24 |
| 6M | 2027-04-06 → 2028-04-06 | 16.66056 | 17.23301× | $287.11 |
| 12M | 2027-10-06 → 2028-10-06 | 18.95277 | 17.23301× | $326.61 |
主表版本RA-NVDA-HORIZON-07-v0.1，零权重分支；等同H03市场P/E不变参考。Headline version RA-NVDA-HORIZON-07-v0.1, zero-weight branch; equivalent to the H03 held-market-P/E reference.主表版本RA-NVDA-HORIZON-07-v0.1，零權重分支；等同H03市場P/E不變參考。
H09未采用候选诊断H09 unadopted candidate diagnosticsH09未採用候選診斷
| 期限Horizon期限 | 目标日 → 盈利窗口末日Target → earnings-window end目標日 → 盈利窗口末日 | Core EPS | P/E | 计算输出 USD/股Calculated USD/share計算輸出 USD/股 |
|---|---|---|---|---|
| 0M | 2026-10-06 → 2027-10-06 | 13.86619 | 17.25348× | $239.24 |
| 6M | 2027-04-06 → 2028-04-06 | 16.62886 | 17.25348× | $286.91 |
| 12M | 2027-10-06 → 2028-10-06 | 18.90702 | 17.25348× | $326.21 |
诊断版本RA-NVDA-HORIZON-09-v0.1，候选未采用。Diagnostic version RA-NVDA-HORIZON-09-v0.1, candidate not adopted.診斷版本RA-NVDA-HORIZON-09-v0.1，候選未採用。
候选自己的CoreEPS0为13.866187，PE为17.253482×；原参考为13.882658与17.233011×。三个期限分别在同一分支体系内计算，不混用。0M是现价恒等式，6M／12M是条件价格预测。The candidate uses its own CoreEPS0of13.866187 and PE17.253482×; the reference uses13.882658 and17.233011×. Each branch computes all horizons consistently without mixing.0M is a current-price identity;6M/12M are conditional price forecasts.候選自己的CoreEPS0為13.866187，PE為17.253482×；原參考為13.882658與17.233011×。三個期限分別在同一分支體系內計算，不混用。0M是現價恆等式，6M／12M是條件價格預測。
增长锚改动与估值公式Growth-anchor change and valuation formula增長錨改動與估值公式
旧模型以相邻季度日均收入变化扣除季节差，可能放大单季增长。H09保持第一季度指引完全一致，将初始动量替换为下一指导季度与上一财年同季度实际日均收入比的log/4。严格同FY/fq匹配、按实际季度天数归一化；同季同比不再重复扣除季节差。The old model removes seasonal differences from adjacent-quarter daily-revenue changes, potentially amplifying one-quarter growth. H09 preserves the first guided quarter exactly and replaces initial momentum with log of next guided daily revenue divided by actual daily revenue from the same fiscal quarter one year earlier, divided by four. Match exact fiscal year/quarter and actual days; do not subtract seasonality again from same-quarter YoY growth.舊模型以相鄰季度日均收入變化扣除季節差，可能放大單季增長。H09保持第一季度指引完全一致，將初始動量替換為下一指導季度與上一財年同季度實際日均收入比的log/4。嚴格同FY/fq匹配、按實際季度天數歸一化；同季同比不再重複扣除季節差。
| 步骤Step步驟 | 定义Definition定義 |
|---|---|
| Growth anchor | g₀ = log[(bias-adjusted guided Revenue₁/days₁)/(known Revenue_same-quarter-last-year/days_lag)] / 4 |
| Quarter j ≥ 2 | DailyRevenueⱼ = DailyRevenueⱼ₋₁ × exp[seasonⱼ − seasonⱼ₋₁ + g₀ × 2^(-(j−1)/4)] |
| Accounting | EPSⱼ = (Revenueⱼ × GAAP GM − Opexⱼ) × 85% / diluted shares |
| Own price anchor | PE₉ = P₀ / EPS₉₀; P₉ₕ = EPS₉ₕ × PE₉ |
保留12个已知季度的季节拟合、4季度半衰期、已成熟指引偏差、H03第一季度费用金额及其后TTM费用率、GAAP GM、股数、15%正常化Core税率和日数聚合。首季度整个记录保持原样，新动量只用于第2季度起。缺失或无效同比季度明确不可用，没有未来季度补值。Keep seasonality from twelve known quarters, four-quarter half-life, matured guidance bias, H03 first-quarter dollar expenses and later TTM expense ratio, GAAP GM, shares,15%normalized Core tax and day aggregation. Preserve the entire first-quarter record; new momentum applies from quarter2. Missing or invalid prior-year actuals are unavailable, with no future-quarter substitution.保留12個已知季度的季節擬合、4季度半衰期、已成熟指引偏差、H03第一季度費用金額及其後TTM費用率、GAAP GM、股數、15%正常化Core税率和日數聚合。首季度整個記錄保持原樣，新動量只用於第2季度起。缺失或無效同比季度明確不可用，沒有未來季度補值。
分析师目标价只作外部比较，没有用于拟合、特征或加权定价。P/E仍以当期价格重新归一，H09没有独立估计未来退出P/E；不能与旧数值倍数混用。Analyst targets are external comparison evidence, not training data, features or blended pricing inputs. P/E remains renormalized to the current price; H09 does not independently estimate future exit P/E and cannot use the old numerical multiple with its new EPS.分析師目標價只作外部比較，沒有用於擬合、特徵或加權定價。P/E仍以當期價格重新歸一，H09沒有獨立估計未來退出P/E；不能與舊數值倍數混用。
十年实际股价检验Ten-year realized-price evaluation十年實際股價檢驗
范围2016-10-06至2026-10-06：原40个季度起点，加固定起始边界，共41个；不是逐日十年预测。6M／12M已到期样本39／37，未来目标日取当日或其后首个交易日收盘。0M现价恒等式不评分。Range2016-10-06 through2026-10-06: forty original quarterly origins plus one fixed start boundary,41 in total; not daily ten-year forecasts. Mature6M/12M samples are39/37. Labels are the first trading close on or after the target date. The0M observed-price identity is not scored.範圍2016-10-06至2026-10-06：原40個季度起點，加固定起始邊界，共41個；不是逐日十年預測。6M／12M已到期樣本39／37，未來目標日取當日或其後首個交易日收盤。0M現價恆等式不評分。
| 期限Horizon期限 | 分支Branch分支 | N | MALE | 中位绝对误差Median APE中位絕對誤差 | P90误差P90 APEP90誤差 | 方向命中Direction方向命中 |
|---|---|---|---|---|---|---|
| 6M | H09 | 39 | 0.309121 | 24.05% | 57.46% | 69.23% |
| 6M | MarketPEHeld | 39 | 0.319356 | 26.17% | 50.90% | 74.36% |
| 6M | PriceUnchanged | 39 | 0.355869 | 33.01% | 54.14% | — |
| 12M | H09 | 37 | 0.537194 | 44.36% | 72.52% | 64.86% |
| 12M | MarketPEHeld | 37 | 0.493833 | 37.30% | 73.51% | 67.57% |
| 12M | PriceUnchanged | 37 | 0.622806 | 48.23% | 68.55% | — |
较匹配的市场P/E不变基准，6M MALE改善3.20%，但P90误差恶化12.89%，只改善2/4财季面板；12M MALE恶化8.78%、中位误差恶化18.91%，只改善1/4面板。股价不变为零收益中性预测，方向栏留空。MALE越低越好，不是准确率。Versus the matched held-market-P/E baseline,6M MALE improves3.20%, but P90 error worsens12.89% and only two of four fiscal-quarter panels improve.12M MALE worsens8.78%, median error worsens18.91%, and only one of four panels improves. Unchanged price is a neutral zero-return forecast; leave its direction blank. Lower MALE is better; it is not an accuracy rate.較匹配的市場P/E不變基準，6M MALE改善3.20%，但P90誤差惡化12.89%，只改善2/4財季面板；12M MALE惡化8.78%、中位誤差惡化18.91%，只改善1/4面板。股價不變為零收益中性預測，方向欄留空。MALE越低越好，不是準確率。
| 期限Horizon期限 | 对照Control對照 | MALE改善MALE improvementMALE改善 | 配对95%误差差区间Paired95% loss-difference interval配對95%誤差差區間 | 门槛Gate門檻 |
|---|---|---|---|---|
| 6M | MarketPEHeld | 3.20% | [-0.05109, 0.03268] | false |
| 6M | PriceUnchanged | 13.14% | [-0.09892, 0.00646] | true |
| 12M | MarketPEHeld | -8.78% | [-0.03577, 0.14834] | false |
| 12M | PriceUnchanged | 13.75% | [-0.17865, 0.01793] | true |
三年同日分析师平均目标价检验Three-year same-date analyst average-target evaluation三年同日分析師平均目標價檢驗
范围2023-10-06至2026-10-06，12个原季度历史起点。将model12M(t)与同日StockAnalysis All Analysts平均一年目标价(t)比较。日期与金额来自2026-10-07悬停提示，按显示精度记录；当前截面另列。0M／6M无同期限目标价，本项不评分。Range2023-10-06 through2026-10-06, twelve original quarterly historical origins. Compare model12M(t) with same-date StockAnalysis All Analysts average one-year target(t). Dates and amounts were read from hover tooltips on2026-10-07 at displayed precision. Separate the current snapshot.0M/6M have no matched-horizon targets and are not scored here.範圍2023-10-06至2026-10-06，12個原季度歷史起點。將model12M(t)與同日StockAnalysis All Analysts平均一年目標價(t)比較。日期與金額來自2026-10-07懸停提示，按顯示精度記錄；當前截面另列。0M／6M無同期限目標價，本項不評分。
| 分支Branch分支 | N | MALE | 平均绝对差异Mean absolute gap平均絕對差異 | 中位差异Median gap中位差異 | P90差异P90 gapP90差異 | 平均有符号差异Mean signed gap平均有符號差異 | ±10% | ±20% |
|---|---|---|---|---|---|---|---|---|
| H09 | 12 | 0.146112 | 16.86% | 8.81% | 35.93% | 13.24% | 7/12 | 9/12 |
| MarketPEHeld | 12 | 0.222992 | 22.64% | 20.79% | 43.27% | 9.76% | 3/12 | 5/12 |
共识MALE改善34.48%，平均绝对差异22.64%→16.86%，中位20.79%→8.81%；±10%内3→7个，±20%内5→9个。不过模型相对共识的平均带符号差异从+9.76%升至+13.24%，改善绝对差异后整体仍高于共识。Consensus MALE improves34.48%; mean absolute gap falls22.64%→16.86%, median20.79%→8.81%. Counts within±10%rise3→7 and within±20%rise5→9. However, mean signed gap to consensus rises from+9.76%to+13.24%; the model remains higher than consensus despite smaller absolute gaps.共識MALE改善34.48%，平均絕對差異22.64%→16.86%，中位20.79%→8.81%；±10%內3→7個，±20%內5→9個。不過模型相對共識的平均帶符號差異從+9.76%升至+13.24%，改善絕對差異後整體仍高於共識。
共识点估计门槛通过，但配对95%区间[-0.21244,+0.04499]跨零。12个重叠、集中于AI周期的样本不足以证明稳健优势。共识不是正确价值标签，也不是未来成交结果。Consensus point-estimate gates pass, but the paired95%interval[-0.21244,+0.04499]includes zero. Twelve overlapping observations concentrated in the AI cycle do not establish a robust advantage. Consensus is neither a true-value label nor a future realized outcome.共識點估計門檻通過，但配對95%區間[-0.21244,+0.04499]跨零。12個重疊、集中於AI週期的樣本不足以證明穩健優勢。共識不是正確價值標籤，也不是未來成交結果。
StockAnalysis平均目标价图表StockAnalysis average-target chartStockAnalysis平均目標價圖表
EPS防护检验EPS guard evaluationEPS防護檢驗
| 期限Horizon期限 | 分支Branch分支 | N | MALE | 中位误差Median APE中位誤差 | P90误差P90 APEP90誤差 | 防护Guard防護 |
|---|---|---|---|---|---|---|
| 0M | H09EPS | 36 | 0.441139 | 28.47% | 107.05% | false |
| 0M | H03EPS | 36 | 0.394212 | 18.00% | 94.45% | — |
| 6M | H09EPS | 34 | 0.630331 | 37.96% | 137.12% | false |
| 6M | H03EPS | 34 | 0.537500 | 31.49% | 90.06% | — |
| 12M | H09EPS | 32 | 0.750211 | 51.46% | 93.86% | false |
| 12M | H03EPS | 32 | 0.654627 | 36.30% | 91.75% | — |
EPS0／6／12 MALE分别恶化11.90%／17.27%／14.60%，三个防护均未通过。EPS样本要求完整未来盈利窗口财报已公布，不能拿价格到期样本替代。EPS0/6/12 MALE worsens11.90%/17.27%/14.60%; all three guards fail. EPS maturity requires financial publication for the full future earnings window and cannot be replaced by price maturity.EPS0／6／12 MALE分別惡化11.90%／17.27%／14.60%，三個防護均未通過。EPS樣本要求完整未來盈利窗口財報已公佈，不能拿價格到期樣本替代。
样本范围与诊断Sample scope and diagnostics樣本範圍與診斷
| 期限Horizon期限 | 起点Origins起點 | 可计算Available可計算 | 成熟EPSMature EPS成熟EPS | 到期价格Mature price到期價格 |
|---|---|---|---|---|
| 0M | 41 | 41 | 36 | — |
| 6M | 41 | 41 | 34 | 39 |
| 12M | 41 | 41 | 32 | 37 |
原40起点与近期已到期子集Original forty origins and recent mature subsets原40起點與近期已到期子集
| Scope | Horizon | Model | N | MALE | Median APE |
|---|---|---|---|---|---|
| original40 | 6M | H09 | 38 | 0.308175 | 23.36% |
| original40 | 6M | MarketPEHeld | 38 | 0.322496 | 26.33% |
| original40 | 6M | PriceUnchanged | 38 | 0.354629 | 32.22% |
| original40 | 12M | H09 | 36 | 0.527800 | 42.70% |
| original40 | 12M | MarketPEHeld | 36 | 0.490068 | 36.87% |
| original40 | 12M | PriceUnchanged | 36 | 0.612595 | 46.14% |
| consensus-date mature-price subset | 6M | H09 | 10 | 0.147441 | 12.21% |
| consensus-date mature-price subset | 6M | MarketPEHeld | 10 | 0.226953 | 20.03% |
| consensus-date mature-price subset | 6M | PriceUnchanged | 10 | 0.253708 | 19.73% |
| consensus-date mature-price subset | 12M | H09 | 8 | 0.162365 | 10.46% |
| consensus-date mature-price subset | 12M | MarketPEHeld | 8 | 0.240083 | 7.67% |
| consensus-date mature-price subset | 12M | PriceUnchanged | 8 | 0.420237 | 32.34% |
原40起点保留相同拒绝结论。近期共识日期对应的价格到期样本只有6M10个、12M8个；12M全部8个实际上涨，方向100%也等于一直预测上涨。近期子集不能替代完整十年检验，12M中位误差在该子集仍恶化36.40%。The original forty origins retain the same rejection. Mature prices at recent consensus dates number only ten at6M and eight at12M. All eight12M outcomes rise, so100%direction also equals always predicting up. Recent subsets cannot replace the full-decade evaluation; their12M median error still worsens36.40%.原40起點保留相同拒絕結論。近期共識日期對應的價格到期樣本只有6M10個、12M8個；12M全部8個實際上漲，方向100%也等於一直預測上漲。近期子集不能替代完整十年檢驗，12M中位誤差在該子集仍惡化36.40%。
固定起始边界仅新增一个早期样本，季节拟合不足时沿用原回退；包括该边界共2个季节回退。所有41个起点均可计算，没有因表现删样本。The fixed boundary adds one early observation and uses the inherited fallback when seasonal history is insufficient; two seasonal fallbacks occur including that boundary. All41 origins are available, with no performance-based removal.固定起始邊界僅新增一個早期樣本，季節擬合不足時沿用原回退；包括該邊界共2個季節回退。所有41個起點均可計算，沒有因表現刪樣本。
预先规则、独立检查与决定Prespecified rules, independent checks and decision預先規則、獨立檢查與決定
方案在首次输出前固定，仅运行一个候选：实际价格6M／12M均需较市场PE不变和股价不变改善MALE≥10%、中位误差≥5%，P90恶化≤10%，至少3/4财季面板改善；共识12M需较市场PE不变达到相应绝对差异门槛；EPS各期MALE和P90恶化不超过10%。Freeze before first output and run one candidate. Both6M/12M actual-price tests require≥10%MALE and≥5%median improvement against held-market-P/E and unchanged price,≤10%P90 worsening, and improvement in at least three of four fiscal-quarter panels.12M consensus requires corresponding gap improvements versus held-market-P/E. EPS MALE and P90 may not worsen by more than10%at any horizon.方案在首次輸出前固定，僅運行一個候選：實際價格6M／12M均需較市場PE不變和股價不變改善MALE≥10%、中位誤差≥5%，P90惡化≤10%，至少3/4財季面板改善；共識12M需較市場PE不變達到相應絕對差異門檻；EPS各期MALE和P90惡化不超過10%。
| 检验Evaluation檢驗 | 通过Passed通過 |
|---|---|
| 实际价格门槛Actual-price gates實際價格門檻 | false |
| 分析师共识点估计门槛Analyst-consensus point-estimate gates分析師共識點估計門檻 | true |
| EPS防护EPS guardsEPS防護 | false |
| 双重开发门槛Dual development gates雙重開發門檻 | false |
| 独立检查Independent checks獨立檢查 | 项数Count項數 | 通过Passed通過 |
|---|---|---|
| 经营、日数、价格锚与原控制Operating paths, day counts, price anchors and old controls經營、日數、價格錨與原控制 | 5933 | true |
| 未来信息修改、删除与追加Future-information mutation, deletion and append未來信息修改、刪除與追加 | 226 | true |
5,933项数值检查和226项信息隔离检查通过，审计未读取新评分。旧研究与分析师基准的8,916个文件哈希保留。检查证明实现与信息边界，不证明预测有效。5,933numerical checks and226information-isolation checks pass; auditors did not read new scores. Preserve hashes of8,916old research and analyst-benchmark files. Checks establish implementation and information boundaries, not predictive validity.5,933項數值檢查和226項信息隔離檢查通過，審計未讀取新評分。舊研究與分析師基準的8,916個文件哈希保留。檢查證明實現與信息邊界，不證明預測有效。
H09未采用，当前参考保持H07。历史多轮重用、窗口重叠、共识仅12点，均限制结论。原始财报时点是盘后可得性重建；分析师历史原始发布时间、修订和拆股换算沿革尚未独立验证。没有独立样本外验证或正式Champion。网页发布只记录研究，未部署预测模型。H09 is not adopted; retain H07. Reused history, overlapping windows and only twelve consensus points limit conclusions. Financial timing reconstructs after-close availability; original analyst publication times, revisions and split-adjustment lineage remain unverified. No independent OOS validation or formal Champion. Publishing records research and does not deploy a prediction model.H09未採用，當前參考保持H07。歷史多輪重用、窗口重疊、共識僅12點，均限制結論。原始財報時點是盤後可得性重建；分析師歷史原始發佈時間、修訂和拆股換算沿革尚未獨立驗證。沒有獨立樣本外驗證或正式Champion。網頁發佈只記錄研究，未部署預測模型。
逐期价格、Pro目标价、财报和指引路径留内部；公开仅方法、派生模型值及聚合结论。下一研究应区分收入周期与同比滞后、盈利路径与倍数状态；本轮结果后没有试第二候选。Keep origin-level prices, Pro targets, financials and guidance paths internal; publish only methods, derived model outputs and aggregates. Future research should distinguish revenue cycles and YoY lag, earnings paths and multiple states. No second candidate was tried after these results.逐期價格、Pro目標價、財報和指引路徑留內部；公開僅方法、派生模型值及聚合結論。下一研究應區分收入週期與同比滯後、盈利路徑與倍數狀態；本輪結果後沒有試第二候選。
滚动时点验证方法Rolling-origin evaluation method滾動時點驗證方法
Protocol SHA256 28e03f79822a272641c8d50c1f136ed00175905acffec79a07e449a208d6a3ae
Record SHA256 a7212801ab4b9ffd77851ed02b93dfafee237c3b6e5c0c009672e746e6181fd9
H08及原三年共识比较H08 and original three-year consensus comparisonH08及原三年共識比較 · H07保留参考Retained H07 referenceH07保留參考 · H06趋势档案H06 trend archiveH06趨勢檔案 · H05档案H05 archiveH05檔案 · H04档案H04 archiveH04檔案 · H03档案H03 archiveH03檔案 · H02档案H02 archiveH02檔案