NVDA / HORIZON-08 / 2026-10-06
H08：费用路径连续性，候选未通过H08: expense-path continuity; candidate failsH08：費用路徑連續性，候選未通過
研究截面2026-10-06收盘$239.24，收盘价于纽约23:55再次核验。原财报与指引记录沿用内部留档，本轮只变更费用路径。16:30是收盘后可得性重建约定，不是同期保存的PIT或16:00成交输入。Research cutoff: the2026-10-06 close of$239.24, rechecked at23:55 New York. Preserve the internally archived filings and guidance; change only the expense path.16:30 is an after-close availability reconstruction, not contemporaneous PIT or a16:00 execution input.研究截面2026-10-06收盤$239.24，收盤價於紐約23:55再次核驗。原財報與指引記錄沿用內部留檔，本輪只變更費用路徑。16:30是收盤後可得性重建約定，不是同期保存的PIT或16:00成交輸入。
H08-GUIDE-OPEX-CONTINUITY／协议v0.1未通过，保留H07市场P/E不变参考。候选价格误差仅减少约0.20%／0.007%，EPS误差反而增大，不能替换参考模型。0M是现价锚，6M／12M是条件价格预测，均不代表独立公允价值。H08-GUIDE-OPEX-CONTINUITY / protocol v0.1 fails; retain the H07 held-market-P/E reference. Candidate price errors fall only about0.20%/0.007%, while EPS errors increase, so it does not replace the reference.0M is an observed-price anchor;6M/12M are conditional price forecasts, not independent fair values.H08-GUIDE-OPEX-CONTINUITY／協議v0.1未通過，保留H07市場P/E不變參考。候選價格誤差僅減少約0.20%／0.007%，EPS誤差反而增大，不能替換參考模型。0M是現價錨，6M／12M是條件價格預測，均不代表獨立公允價值。
本轮保留的当前参考输出Current reference retained this round本輪保留的當前參考輸出
| 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 · 现价参考Price reference現價參考 | 2026-10-06 → 2027-10-06 | 13.8827 | 17.2330× | $239.24 |
| 6M | 2027-04-06 → 2028-04-06 | 16.6606 | 17.2330× | $287.11 |
| 12M | 2027-10-06 → 2028-10-06 | 18.9528 | 17.2330× | $326.61 |
追加研究：过去三年同日平均目标价比较Added study: three-year same-date average-target comparison追加研究：過去三年同日平均目標價比較
2026-10-07追加，范围2023-10-06至2026-10-06。逐一悬停读取StockAnalysis的All Analysts平均一年目标价，匹配原模型12个季度历史起点的12M预测；日期及金额均按提示实际显示精度读取，未从曲线像素估算。Added on2026-10-07; range2023-10-06 through2026-10-06. Read StockAnalysis All Analysts average one-year targets from hover tooltips and match twelve original quarterly model origins to their12M predictions. Dates and amounts use displayed tooltip precision, not values estimated from curve pixels.2026-10-07追加，範圍2023-10-06至2026-10-06。逐一懸停讀取StockAnalysis的All Analysts平均一年目標價，匹配原模型12個季度歷史起點的12M預測；日期及金額均按提示實際顯示精度讀取，未從曲線像素估算。
主比较为Model12M(t)对Average12MTarget(t)，不对未来实际成交价或未来发布的新目标价评分。0M／6M没有同期限分析师目标价，本次不评分；当前截面单列，未混入12个历史样本。模型的当期价格锚保留，参数未调整。Compare Model12M(t) with Average12MTarget(t), without scoring future traded prices or newly published future targets.0M/6M lack matched-horizon analyst targets and are not scored. The current snapshot is separate from the twelve historical samples. Preserve the model’s current-price anchor and parameters.主比較為Model12M(t)對Average12MTarget(t)，不對未來實際成交價或未來發布的新目標價評分。0M／6M沒有同期限分析師目標價，本次不評分；當前截面單列，未混入12個歷史樣本。模型的當期價格錨保留，參數未調整。
| 分支Branch分支 | N | MALE | 平均绝对差异Mean absolute gap平均絕對差異 | 中位绝对差异Median absolute gap中位絕對差異 | 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% |
|---|---|---|---|---|---|---|---|---|
| 保留H07参考Retained H07 reference保留H07參考 | 12 | 0.222992 | 22.64% | 20.79% | 43.27% | 9.76% | 3/12 | 5/12 |
| H08未采用诊断Rejected H08 diagnosticH08未採用診斷 | 12 | 0.226602 | 23.14% | 21.10% | 43.42% | 10.37% | 3/12 | 5/12 |
保留参考平均绝对差异22.64%，中位20.79%；9/12个起点高于共识，3/12低于共识。H08平均绝对差异23.14%，没有改善。该比较不改变H08未采用的决定，也不构成正式Champion或独立样本外验证。The retained reference has a22.64%mean absolute gap and20.79%median gap; nine of twelve origins exceed consensus and three fall below it. H08’s23.14%mean absolute gap is not an improvement. This comparison does not change H08’s rejection or establish a formal Champion or independent OOS validation.保留參考平均絕對差異22.64%，中位20.79%；9/12個起點高於共識，3/12低於共識。H08平均絕對差異23.14%，沒有改善。該比較不改變H08未採用的決定，也不構成正式Champion或獨立樣本外驗證。
相对差异=模型12M/同日平均目标价−1；绝对差异取绝对值，MALE为平均绝对对数差。这些是与共识的差异指标，不是预测实际股价的准确率。小样本与AI周期集中限制结论。Relative gap=model12M/same-date average target−1; take its absolute value for absolute gaps. MALE is mean absolute log difference. These quantify differences from consensus, not realized-price accuracy. Small samples and concentration in the AI cycle limit conclusions.相對差異=模型12M/同日平均目標價−1；絕對差異取絕對值，MALE為平均絕對對數差。這些是與共識的差異指標，不是預測實際股價的準確率。小樣本與AI週期集中限制結論。
当前2026-10-06冻结参考仍为0M $239.24、6M $287.11、12M $326.61。12M相对同日页面的一年平均目标价差异为−0.64%；只是一个当前截面，不代表三年样本已经达到这个差异水平。The frozen2026-10-06 reference remains0M$239.24,6M$287.11 and12M$326.61. Its12M gap versus the same-date page’s average one-year target is−0.64%; this single current snapshot does not describe the three-year sample’s gap.當前2026-10-06凍結參考仍為0M $239.24、6M $287.11、12M $326.61。12M相對同日頁面的一年平均目標價差異為−0.64%；只是一個當前截面，不代表三年樣本已經達到這個差異水平。
历史数值在2026-10-07读取。历史日期标签不证明原始PIT：原始发布时间、后续修订及拆股换算沿革尚未独立核验。模型按收盘后财报可得性重建，目标价日内更新时间未知。历史已用于模型研究，因此本次为回溯共识比较。公开只发布模型输出、方法及聚合结论；逐日期原始目标价和提示文本保留内部。Historical values were read on2026-10-07. Date labels do not establish original PIT: publication times, subsequent revisions and split-adjustment lineage are not independently verified. Models reconstruct after-close financial availability; intraday consensus-update times are unknown. This history has been used in model research, making this a retrospective consensus comparison. Publish model outputs, methods and aggregates only; date-level vendor targets and tooltip text stay internal.歷史數值在2026-10-07讀取。歷史日期標籤不證明原始PIT：原始發佈時間、後續修訂及拆股換算沿革尚未獨立核驗。模型按收盤後財報可得性重建，目標價日內更新時間未知。歷史已用於模型研究，因此本次為回溯共識比較。公開只發布模型輸出、方法及聚合結論；逐日期原始目標價和提示文本保留內部。
StockAnalysis历史目标价图表StockAnalysis historical target chartStockAnalysis歷史目標價圖表
以下保留原H08验证档案：实际价格检验不属于上方新增的三年共识比较。Original H08 validation archive follows: realized-price tests are separate from the added three-year consensus comparison above.以下保留原H08驗證檔案：實際價格檢驗不屬於上方新增的三年共識比較。
H08候选诊断输出，未采纳H08 candidate diagnostics, not adoptedH08候選診斷輸出，未採納
| 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 · 现价参考Price reference現價參考 | 2026-10-06 → 2027-10-06 | 14.1581 | 16.8978× | $239.24 |
| 6M | 2027-04-06 → 2028-04-06 | 17.0053 | 16.8978× | $287.35 |
| 12M | 2027-10-06 → 2028-10-06 | 19.3450 | 16.8978× | $326.89 |
H08使用自己的EPS0重新归一P/E：239.24/14.158083=16.897768×。参考模型使用13.882658与17.233011×。EPS变高时不能继续套用旧数值P/E，混成另一套价格主值。H08 renormalizes P/E using its own EPS0:239.24/14.158083=16.897768×. The reference uses13.882658 and17.233011×. Higher EPS cannot be combined with the old numerical P/E to invent a mixed headline price.H08使用自己的EPS0重新歸一P/E：239.24/14.158083=16.897768×。參考模型使用13.882658與17.233011×。EPS變高時不能繼續套用舊數值P/E，混成另一套價格主值。
单独修正费用口径跳变Change only the expense-basis transition單獨修正費用口徑跳變
旧H03第1季度采用已公布指引的费用金额，第2季度起改用历史TTM费用／收入比例。在AI加速期曾出现收入预测接近实际、营业利润率却机械下跳的情况。H08保持第1季度完全一致，其后固定使用第1季度费用比例；第10季度之后的延伸也保持同一规则。Old H03 uses guided dollar expenses in quarter1, then switches to historical TTM expenses/revenue. During AI acceleration this could mechanically lower operating margin even when revenue was close to actual. H08 preserves quarter1 exactly, then applies its expense ratio to later quarters, including extensions beyond quarter10.舊H03第1季度採用已公佈指引的費用金額，第2季度起改用歷史TTM費用／收入比例。在AI加速期曾出現收入預測接近實際、營業利潤率卻機械下跳的情況。H08保持第1季度完全一致，其後固定使用第1季度費用比例；第10季度之後的延伸也保持同一規則。
| 规则Rule規則 | 定义Definition定義 |
|---|---|
| Expense anchor | r = Opex₁ / bias-adjusted Revenue₁ |
| Future expenses | Opexⱼ = Revenueⱼ × r |
| Operating profit | OPⱼ = Revenueⱼ × guided GAAP GM − Opexⱼ |
| Normalized Core EPS | EPSⱼ = OPⱼ × 85% / diluted shares |
| Price forecast | Pₕ = EPSₕ × (P₀/EPS₀) |
收入递推、12季度季节拟合、4季度半衰期、指引偏差修正、毛利率、股数及15%正常化税率保持不变。所有期限由同一季度路径按实际日数聚合；没有拟合价格、费用权重或新EPS残差参数。Revenue recurrence,12-quarter seasonality,4-quarter half-life, guidance-bias correction, gross margin, shares and15%normalized tax remain unchanged. All horizons aggregate the same quarter path by actual days. No price, expense-weight or new EPS-residual parameters are fitted.收入遞推、12季度季節擬合、4季度半衰期、指引偏差修正、毛利率、股數及15%正常化税率保持不變。所有期限由同一季度路徑按實際日數聚合；沒有擬合價格、費用權重或新EPS殘差參數。
| 当前模型字段Current model field當前模型字段 | H08H08H08 | 旧后续路径Old later path舊後續路徑 |
|---|---|---|
| 费用／收入Expenses/revenue費用／收入 | 8.1247% | 9.4604% |
| 模型营业利润率Model operating margin模型營業利潤率 | 65.8753% | 64.5396% |
上述65.8753%是模型指引锚季度的利润率，收入含已成熟指引误差修正，不是公司直接给出的营业利润率。将费用固定为收入比例，仍是研究假设，忽略固定成本、研发投资节奏及规模效应。The65.8753%margin belongs to the model’s guidance-anchored quarter and includes a matured guidance-error revenue adjustment; it is not a directly stated company operating-margin forecast. Expenses proportional to revenue remain a research assumption, omitting fixed costs, R&D timing and scale effects.上述65.8753%是模型指引錨季度的利潤率，收入含已成熟指引誤差修正，不是公司直接給出的營業利潤率。將費用固定為收入比例，仍是研究假設，忽略固定成本、研發投資節奏及規模效應。
EPS水平与价格预测的关系How EPS levels relate to price forecastsEPS水平與價格預測的關係
H08固定未来毛利率、费用比例、税率和股数后，EPS窗口正比于收入窗口，所以Pₕ=P₀×RevenueWindowₕ/RevenueWindow₀。利润率的绝对水平在现价锚中抵消；本轮价格变化来自旧第1季度与后续季度费用切换所扭曲的盈利增长比例得到改变。With future gross margin, expense ratio, tax and shares fixed, EPS windows are proportional to revenue windows, so Pₕ=P₀×RevenueWindowₕ/RevenueWindow₀. The absolute margin level cancels in the current-price anchor. Price changes arise from changing the earnings-growth ratio distorted by the old quarter1-to-later expense transition.H08固定未來毛利率、費用比例、税率和股數後，EPS窗口正比於收入窗口，所以Pₕ=P₀×RevenueWindowₕ/RevenueWindow₀。利潤率的絕對水平在現價錨中抵消；本輪價格變化來自舊第1季度與後續季度費用切換所扭曲的盈利增長比例得到改變。
因此，这个分支修正了费用路径一致性，但没有独立估计未来P/E，也不能把EPS水平提高直接称为更高公允价值。This branch improves expense-path consistency but does not independently estimate future P/E; higher EPS levels do not directly imply higher fair value.因此，這個分支修正了費用路徑一致性，但沒有獨立估計未來P/E，也不能把EPS水平提高直接稱為更高公允價值。
盈利检验：误差增大，未通过Earnings test: errors increase; gates fail盈利檢驗：誤差增大，未通過
| 期限Horizon期限 | N | H08 MALEH08 MALEH08 MALE | H03 MALEH03 MALEH03 MALE | 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 |
|---|---|---|---|---|---|---|
| 0M | 35 | 0.4017 | 0.3973 | -1.10% | [-0.03531, 0.04335] | false |
| 6M | 33 | 0.6008 | 0.5418 | -10.89% | [0.02417, 0.12171] | false |
| 12M | 31 | 0.7508 | 0.6619 | -13.44% | [0.05123, 0.15910] | false |
EPS0／6／12的MALE分别恶化1.10%／10.89%／13.44%。6M与12M四个财季面板全部未改善。费用跳变被消除，不等于盈利预测更准确；收入外推误差与费用误差此前有相互抵消。MALE越低越好，不是准确率。EPS0/6/12 MALE worsens by1.10%/10.89%/13.44%. None of the four fiscal-quarter panels improve at6M or12M. Removing the expense discontinuity does not improve earnings accuracy; revenue-extrapolation and expense errors previously offset one another. Lower MALE is better; it is not an accuracy rate.EPS0／6／12的MALE分別惡化1.10%／10.89%／13.44%。6M與12M四個財季面板全部未改善。費用跳變被消除，不等於盈利預測更準確；收入外推誤差與費用誤差此前有相互抵消。MALE越低越好，不是準確率。
展开同样本盈利对照Expand same-sample earnings controls展開同樣本盈利對照
| Horizon | Model | N | MALE | Median APE | P90 APE |
|---|---|---|---|---|---|
| 0M | H08EPS | 35 | 0.4017 | 20.3% | 99.0% |
| 0M | H03EPS | 35 | 0.3973 | 17.4% | 96.8% |
| 0M | naiveEPS | 35 | 0.7242 | 47.1% | 90.3% |
| 6M | H08EPS | 33 | 0.6008 | 37.9% | 92.6% |
| 6M | H03EPS | 33 | 0.5418 | 30.4% | 90.1% |
| 6M | naiveEPS | 33 | 0.8989 | 59.3% | 85.5% |
| 12M | H08EPS | 31 | 0.7508 | 49.5% | 96.7% |
| 12M | H03EPS | 31 | 0.6619 | 37.6% | 91.8% |
| 12M | naiveEPS | 31 | 1.0248 | 62.8% | 89.5% |
naiveEPS是上一年同季盈利重复；它只是对照，未作为新候选运行。naiveEPS repeats prior-year same-quarter earnings; it is a control, not another new candidate.naiveEPS是上一年同季盈利重複；它只是對照，未作為新候選運行。
价格检验：微小差异不足以替换参考Price test: tiny differences do not justify replacement價格檢驗：微小差異不足以替換參考
| Horizon | Model | N | MALE | Median APE | P90 APE | 方向命中Direction方向命中 |
|---|---|---|---|---|---|---|
| 6M | H08 | 38 | 0.321846 | 25.4% | 53.8% | 73.7% |
| 6M | MarketPEHeld | 38 | 0.322496 | 26.3% | 51.4% | 73.7% |
| 6M | PriceUnchanged | 38 | 0.354629 | 32.2% | 54.4% | 0.0% |
| 12M | H08 | 36 | 0.490035 | 37.8% | 75.9% | 66.7% |
| 12M | MarketPEHeld | 36 | 0.490068 | 36.9% | 74.7% | 66.7% |
| 12M | PriceUnchanged | 36 | 0.612595 | 46.1% | 68.7% | 0.0% |
股价不变是零收益中性预测，方向一栏留空，不作为普通涨跌方向基准。Unchanged price is a neutral zero-return forecast; its direction cell is left blank rather than treated as a regular up/down forecast.股價不變是零收益中性預測，方向一欄留空，不作為普通漲跌方向基準。
| 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 | 0.2016% | [-0.00744, 0.00830] | false |
| 6M | PriceUnchanged | 9.2442% | [-0.10079, 0.02849] | false |
| 12M | MarketPEHeld | 0.0067% | [-0.00735, 0.01049] | false |
| 12M | PriceUnchanged | 20.0067% | [-0.29504, 0.02268] | false |
对市场P/E不变仅改善0.2016%／0.0067%，两个配对区间均跨零，没有达到10%门槛。12M中位误差反而恶化，方向命中与基准相同。更高或更连续的EPS路径不能替代价格验证。Improvement versus held-market-P/E is only0.2016%/0.0067%; both paired intervals include zero and fail the10%threshold.12M median error worsens and direction matches the baseline. A higher or more continuous EPS path cannot substitute for price validation.對市場P/E不變僅改善0.2016%／0.0067%，兩個配對區間均跨零，沒有達到10%門檻。12M中位誤差反而惡化，方向命中與基準相同。更高或更連續的EPS路徑不能替代價格驗證。
盈利到期与价格到期分别计算Separate earnings maturity from price maturity盈利到期與價格到期分別計算
| 期限Horizon期限 | 原起点Origins原起點 | 可计算Available可計算 | EPS已成熟Mature EPSEPS已成熟 | 价格已到期Mature price價格已到期 |
|---|---|---|---|---|
| 0M | 40 | 40 | 35 | — |
| 6M | 40 | 40 | 33 | 38 |
| 12M | 40 | 40 | 31 | 36 |
40个季度起点全部可计算，没有删去困难起点。EPS样本要求完整盈利窗口的财报已经公布，价格样本只要求目标收盘已到期，因此35／33／31与38／36不能混为同一批样本。最早起点2016-11，接近10年而非每日十年预测；1个历史季节拟合不足起点保留原回退。0M价格恒等式不评分。All40 quarterly origins are available; no difficult origin is removed. EPS scoring requires fully published earnings windows, while price scoring requires only mature target closes, so35/33/31 and38/36 are different samples. Origins start in2016-11, nearly ten years rather than daily ten-year forecasts. One inherited seasonal-fit fallback is retained. The0M price identity is not scored.40個季度起點全部可計算，沒有刪去困難起點。EPS樣本要求完整盈利窗口的財報已經公佈，價格樣本只要求目標收盤已到期，因此35／33／31與38／36不能混為同一批樣本。最早起點2016-11，接近10年而非每日十年預測；1個歷史季節擬合不足起點保留原回退。0M價格恆等式不評分。
验证与保留决定Validation and retention decision驗證與保留決定
| 检查组Check group檢查組 | 检查数Count檢查數 | 通过Passed通過 |
|---|---|---|
| 独立经营路径与代数核验Independent operating-path and algebra validation獨立經營路徑與代數核驗 | 4958 | true |
| 未来信息扰动与正向控制Future-information perturbations and positive controls未來信息擾動與正向控制 | 435 | true |
原H02至H07的8,884个文件hash保留。方案在首次输出前固定；结果出现后未调整费用比率、收入增长、回退规则或筛选另一候选。检查通过说明实现遵循协议，不能证明预测有效。All8,884 protected H02–H07 files retain their hashes. The specification was frozen before first output. Expense ratios, revenue growth and fallback rules were not changed after results, and no alternative candidate was selected. Passing checks establishes protocol compliance, not predictive performance.原H02至H07的8,884個文件hash保留。方案在首次輸出前固定；結果出現後未調整費用比率、收入增長、回退規則或篩選另一候選。檢查通過説明實現遵循協議，不能證明預測有效。
H08未采纳，当前参考保持H07／H03市场P/E不变。收入周期恢复、季节拟合放大、固定未来利润率和股数仍需进一步研究。旧历史已反复使用，期限重叠，没有独立样本外验证；网页发布不等于正式模型部署。H08 is not adopted; retain the H07/H03 held-market-P/E reference. Revenue-cycle recovery, seasonal amplification, fixed future margins and shares still need research. History has been repeatedly reused, horizons overlap and independent OOS validation is incomplete. Publishing a page does not deploy a validated model.H08未採納，當前參考保持H07／H03市場P/E不變。收入週期恢復、季節擬合放大、固定未來利潤率和股數仍需進一步研究。舊歷史已反覆使用，期限重疊，沒有獨立樣本外驗證；網頁發佈不等於正式模型部署。
公开仅派生输出和聚合结论，逐期价格、财报和指引源记录留在内部。Publish derived outputs and aggregates only; individual price, filing and guidance source records remain internal.公開僅派生輸出和聚合結論，逐期價格、財報和指引源記錄留在內部。
公司指引来源Company guidance source公司指引來源 · 滚动时点验证方法Rolling-origin evaluation method滾動時點驗證方法
Protocol SHA256 01720d989c775c38b332f7037bac543f962dfdb3df23134accadca9a8393a688
Record SHA256 df590da26fc87c031c9d0136692c9cef5daab65cb4f5eb9db26e52dca1cbf6f6
查看H08验证Read H08 evaluation查看H08驗證 · 保留H07参考与档案Retained H07 reference and archive保留H07參考與檔案 · H06趋势回测H06 trend auditH06趨勢回測 · 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