NVDA / ST01 / INDEPENDENT SHORT-TERM RESEARCH
NVDA 3M：趋势、催化剂与宏观NVDA 3M: trend, catalysts and macroNVDA 3M：趨勢、催化劑與宏觀
ST01是独立的新模型族，不使用H09的EPS、PE、盈利路径或分析师目标价。它研究三个月后的市场价格，不能独立证明当前股票便宜或贵。ST01 is a separate model family. H09 EPS, P/E, earnings paths and analyst targets are not inputs. It studies the market price three months ahead; it cannot independently establish current undervaluation or overvaluation.ST01是獨立的新模型族，不使用H09的EPS、PE、盈利路徑或分析師目標價。它研究三個月後的市場價格，不能獨立證明當前股票便宜或貴。
规则冻结与生成：2026-10-07；价格及技术输入截至2026-10-06收盘，模型信息截面纽约16:30；消息清单截至16:00。历史输入今天重建，不冒充当时保存的PIT记录。目标为三个日历月后，到期比较使用目标日或之后首个交易日收盘。Rules frozen and forecast generated October 7. Price/technical inputs retain the 2026-10-06 close; the model information cutoff is 16:30 New York, and the news ledger cutoff is 16:00. Historical inputs are reconstructed today, not contemporaneously saved PIT records. Target is three calendar months ahead, scored at the close on or after that target.規則凍結與生成：2026-10-07；價格及技術輸入截至2026-10-06收盤，模型信息截面紐約16:30；消息清單截至16:00。歷史輸入今天重建，不冒充當時保存的PIT記錄。目標為三個日曆月後，到期比較使用目標日或之後首個交易日收盤。
| 模型／用途Model / purpose模型／用途 | 目标日Target date目標日 | 价格 USD/股USD/share價格 USD/股 | 状态Status狀態 |
|---|---|---|---|
| ST01 · 3M | 2027-01-06 | $244.31 | 实验短期价格预测Experimental short-term price forecast實驗短期價格預測 |
| 下行情景Downside scenario下行情景 | 2027-01-06 | $215.58 | 历史波动情景，无概率Volatility scenario, no probability歷史波動情景，無概率 |
| 上行情景Upside scenario上行情景 | 2027-01-06 | $276.87 | 历史波动情景，无概率Volatility scenario, no probability歷史波動情景，無概率 |
3M中心怎么计算How the 3M center is calculated3M中心怎麼計算
| 层次Layer層次 | 固定规则Frozen rule固定規則 | 当前贡献Current contribution當前貢獻 |
|---|---|---|
| 趋势Trend趨勢 | 21／63／126日对数动量转换为季度尺度，权重25%／50%／25%；只保留25%，与50／200日趋势冲突再减半，截断±12%。Quarter-scaled 21/63/126-session log momentum at 25/50/25% weights; retain 25%, halve when 50/200-day trend disagrees, cap at ±12%.21／63／126日對數動量轉換為季度尺度，權重25%／50%／25%；只保留25%，與50／200日趨勢衝突再減半，截斷±12%。 | 3.6159% log |
| 宏观利率Macro rates宏觀利率 | 过去63交易日滞后10Y变化×−2，收益率先转小数，截断±2%。系数2是研究假设，未经拟合。Lagged 10Y yield change over 63 sessions × −2, yields in decimals, capped at ±2%. Duration 2 is a research assumption, not a fitted sensitivity.過去63交易日滯後10Y變化×−2，收益率先轉小數，截斷±2%。係數2是研究假設，未經擬合。 | -1.5200% log |
| 消息News消息 | 保留来源、日期、情景触发条件；缺少历史预期和首发资料，价格影响未估计。Retain dated sources and scenario triggers. Missing historical expectation and first-publication panels leave the price effect unestimated.保留來源、日期、情景觸發條件；缺少歷史預期和首發資料，價格影響未估計。 | 未估计；中心不加主观溢价Unestimated; no subjective premium未估計；中心不加主觀溢價 |
mL = ln(Pt / P[t-L]) d = 0.25 × (3 × m21) + 0.50 × m63 + 0.25 × (0.5 × m126) tech = clip(0.25 × d × trend_gate, -0.12, +0.12) macro = clip(-2 × (y_lag,t - y_lag,t-63) / 100, -0.02, +0.02) 3M center = P0 × exp(tech + macro)
本次：$239.24×exp(0.036159 − 0.015200)＝$244.31。合计简单回报2.12%；这是几何中心，非算术期望。This run: $239.24 × exp(0.036159 − 0.015200) = $244.31. Simple forecast return is 2.12%. This is a geometric center, not an arithmetic expected value.本次：$239.24×exp(0.036159 − 0.015200)＝$244.31。合計簡單回報2.12%；這是幾何中心，非算術期望。
价格高于50日均线$219.77和200日均线$201.36。宏观用2026-07-07的4.55%到2026-10-05的5.31%，共上行76bp。当前成交量为20日均量的0.94倍；成交量与QQQ仅作诊断，不加权。Price is above the 50-day average of $219.77 and 200-day average of $201.36. Macro uses 4.55% on 2026-07-07 to 5.31% on 2026-10-05, up 76bp. Volume is 0.94× its 20-day average. Volume and QQQ are context diagnostics only.價格高於50日均線$219.77和200日均線$201.36。宏觀用2026-07-07的4.55%到2026-10-05的5.31%，共上行76bp。當前成交量為20日均量的0.94倍；成交量與QQQ僅作診斷，不加權。
过去利率上行已经可能反映在现价里。这个修正只是假设近期利率变化对后续价格还有预测信息，不是重新计算高利率下的合理PE，也不预设未来降息。其增量预测能力尚未单独证明。Past rate rises may already be reflected in spot. The adjustment only hypothesizes that recent yield changes contain information about subsequent prices. It neither recalculates a fair high-rate P/E nor assumes future rate cuts. Its incremental forecasting value is not yet separately established.過去利率上行已經可能反映在現價裏。這個修正只是假設近期利率變化對後續價格還有預測信息，不是重新計算高利率下的合理PE，也不預設未來降息。其增量預測能力尚未單獨證明。
三年验证：改善有限，保留实验身份Three-year evaluation: limited improvement, experimental status三年驗證：改善有限，保留實驗身份
| 同样本规则Matched rule同樣本規則 | N | MAPE | MALE | P90 APE | 方向命中Direction hit方向命中 |
|---|---|---|---|---|---|
| ST01 | 34 | 18.43% | 0.2016 | 36.02% | 64.71% |
| PriceUnchanged | 34 | 18.86% | 0.2108 | 37.82% | 中性，无方向预测Neutral, no direction中性，無方向預測 |
| Momentum63 | 34 | 27.54% | 0.2754 | 55.97% | 52.94% |
月度origin从2023-10-06到2026-10-06，共37次，34次已成熟、3次待到期；2022年以来的价格只用于指标预热。采用每月6日或之后首个交易日，同一预测日、同一终点比较所有规则；不进行交易收益回测。Monthly origins run from October 6, 2023 to October 6, 2026: 37 forecasts, 34 mature and 3 pending. Prices since 2022 only warm indicators. Each origin is the first session on or after calendar day 6. Rules share identical origins and endpoints; this is not a trading-return backtest.月度origin從2023-10-06到2026-10-06，共37次，34次已成熟、3次待到期；2022年以來的價格只用於指標預熱。採用每月6日或之後首個交易日，同一預測日、同一終點比較所有規則；不進行交易收益回測。
ST01的MAPE较现价不变只降低0.43个百分点；MALE相对改善4.39%，未达到冻结的5%门槛。方向命中22/34＝64.71%，低于始终看涨26/34＝76.47%；8次下跌只识别1次，不能作为已验证的择时信号。ST01 reduces MAPE versus unchanged price by only 0.43 percentage points. Relative MALE improvement is 4.39%, below the frozen 5% gate. Direction hit rate is 22/34 = 64.71%, below always-up at 26/34 = 76.47%; only one of eight declines is recognized. It is not a validated timing signal.ST01的MAPE較現價不變只降低0.43個百分點；MALE相對改善4.39%，未達到凍結的5%門檻。方向命中22/34＝64.71%，低於始終看漲26/34＝76.47%；8次下跌只識別1次，不能作為已驗證的擇時信號。
| 非重叠敏感性队列Nonoverlapping sensitivity cohort非重疊敏感性隊列 | N | ST01 MALE | 价格不变MALEUnchanged MALE價格不變MALE | 因日期重叠剔除Date-overlap omissions因日期重疊剔除 |
|---|---|---|---|---|
| 1 | 10 | 0.2314 | 0.2333 | 1 |
| 2 | 10 | 0.2387 | 0.2481 | 2 |
| 3 | 11 | 0.1728 | 0.1902 | 0 |
月度3M标签互相重叠，34不等于34个独立样本。三个错开队列内部按实际日期剔除重叠，但队列之间仍重叠，不能合并成31个独立样本；效果也不均匀。历史此前已多次用于研究，没有独立样本外结论。Monthly 3M labels overlap; 34 is not 34 independent observations. Each staggered cohort removes internal date overlaps, but the cohorts overlap with one another and cannot be pooled into 31 independent observations. Effects differ by cohort. This history has been reused for research; independent OOS evidence is absent.月度3M標籤互相重疊，34不等於34個獨立樣本。三個錯開隊列內部按實際日期剔除重疊，但隊列之間仍重疊，不能合併成31個獨立樣本；效果也不均勻。歷史此前已多次用於研究，沒有獨立樣本外結論。
| origin年份Origin yearorigin年份 | N | ST01 MAPE | 价格不变MAPEUnchanged MAPE價格不變MAPE |
|---|---|---|---|
| 2023 | 3 | 30.76% | 31.35% |
| 2024 | 12 | 20.99% | 22.05% |
| 2025 | 12 | 17.84% | 17.62% |
| 2026 | 7 | 9.76% | 10.15% |
消息与宏观：研究哪些变化News and macro: changes to study消息與宏觀：研究哪些變化
| 截面前已知资料Known before cutoff截面前已知資料 | 日期Date日期 | 下一次观察什么What to observe next下一次觀察什麼 | 来源Source來源 |
|---|---|---|---|
| Q3收入指引1080亿美元±2%Q3 revenue guidance $108B ±2%Q3收入指引1080億美元±2% | 2026-08-26 | 新指引、交付与利润率是否改变市场预期Whether new guidance, deliveries and margins change expectations新指引、交付與利潤率是否改變市場預期 | NVIDIANVIDIANVIDIA |
| 新增1500亿美元回购授权Additional $150B repurchase authorization新增1500億美元回購授權 | 2026-09-28 | 授权后的实际回购，授权不等于执行Actual purchases after authorization; approval is not execution授權後的實際回購，授權不等於執行 | NVIDIANVIDIANVIDIA |
| CoreWeave Rubin部署公告CoreWeave Rubin deployment reportCoreWeave Rubin部署公告 | 2026-09-30 | 产能、交付及客户需求的独立证据Independent evidence of capacity, delivery and demand產能、交付及客户需求的獨立證據 | NVIDIANVIDIANVIDIA |
| Fed目标利率3.75%–4.00%Fed target rate 3.75%–4.00%Fed目標利率3.75%–4.00% | 2026-09-16 | 10月27–28日及12月8–9日会议；结果未知October 27–28 and December 8–9 meetings; outcomes unknown10月27–28日及12月8–9日會議；結果未知 | FedFedFed |
| CPI／PCE通胀与就业CPI/PCE inflation and employmentCPI／PCE通脹與就業 | 2026-09-11 / 09-30 / 10-02 | 首发值相对事前预期的变化，而非用事后修订值替代First releases relative to prior expectations, without revised-value backfills首發值相對事前預期的變化，而非用事後修訂值替代 | BLSBLSBLS |
上行情景观察：新收入或交付预期改善、突破伴随成交量、利率压力缓解。下行情景观察：指引下修、交付受阻、趋势失守或宏观紧缩加剧。这些是预先登记的观察条件，并非已校准的价格冲击。下一次财报日期尚未在本审计中获官方确认；此前11月18日仅为供应商预计日。Upside observations: improved revenue/delivery expectations, a volume-supported breakout and easing rate pressure. Downside observations: lower guidance, delivery disruption, trend failure or tighter macro conditions. These are registered conditions, not calibrated price shocks. The next earnings date is not officially verified in this audit; the previously discussed November 18 remains a vendor estimate.上行情景觀察：新收入或交付預期改善、突破伴隨成交量、利率壓力緩解。下行情景觀察：指引下修、交付受阻、趨勢失守或宏觀緊縮加劇。這些是預先登記的觀察條件，並非已校準的價格衝擊。下一次財報日期尚未在本審計中獲官方確認；此前11月18日僅為供應商預計日。
10月7日会议纪要的内容不进入10月6日截面。现有消息清单不是完整三年首发语料，CPI／PCE／就业也缺少完整历史预期和vintage面板，因此没有声称三层体系全部完成回测。October 7 minutes content is excluded from the October 6 cutoff. The news register is not a complete three-year first-publication corpus; CPI/PCE/jobs lack complete historical expectation and vintage panels. Full three-layer backtesting is not claimed.10月7日會議紀要的內容不進入10月6日截面。現有消息清單不是完整三年首發語料，CPI／PCE／就業也缺少完整歷史預期和vintage面板，因此沒有聲稱三層體系全部完成回測。
波动情景，不能当成概率Volatility scenarios are not probabilities波動情景，不能當成概率
| 模型／用途Model / purpose模型／用途 | 目标日Target date目標日 | 价格 USD/股USD/share價格 USD/股 | 状态Status狀態 |
|---|---|---|---|
| ST01 · 3M | 2027-01-06 | $244.31 | 实验短期价格预测Experimental short-term price forecast實驗短期價格預測 |
| 下行情景Downside scenario下行情景 | 2027-01-06 | $215.58 | 历史波动情景，无概率Volatility scenario, no probability歷史波動情景，無概率 |
| 上行情景Upside scenario上行情景 | 2027-01-06 | $276.87 | 历史波动情景，无概率Volatility scenario, no probability歷史波動情景，無概率 |
情景＝中心×exp(±最近21个日对数收益率的样本标准差×√63)。它反映近期波动尺度，不是公允价值上下限，也不是68%或70%置信区间。历史34次中覆盖24次，只是事后描述，最弱错开队列覆盖率50%。Scenarios = center × exp(±sample standard deviation of the last 21 daily log returns × √63). They express recent volatility scale, not fair-value limits or a 68%/70% confidence interval. Coverage of 24 of 34 observations is descriptive; the weakest staggered cohort covers 50%.情景＝中心×exp(±最近21個日對數收益率的樣本標準差×√63)。它反映近期波動尺度，不是公允價值上下限，也不是68%或70%置信區間。歷史34次中覆蓋24次，只是事後描述，最弱錯開隊列覆蓋率50%。
保持趋势输入不变，额外+50bp／−50bp利率变化的假设敏感性分别为$241.88／$246.76。这不是未来利率预测。Holding trend inputs fixed, hypothetical additional +50bp/−50bp yield-change sensitivities are $241.88 / $246.76. These do not forecast future rates.保持趨勢輸入不變，額外+50bp／−50bp利率變化的假設敏感性分別為$241.88／$246.76。這不是未來利率預測。
与0M／6M／12M估值怎样一起看How to read it beside 0M/6M/12M valuations與0M／6M／12M估值怎樣一起看
| 模型与期限Model and horizon模型與期限 | 目标日 → 盈利窗口末日Target → earnings-window end目標日 → 盈利窗口末日 | Core EPS | 倍数Multiple倍數 | USD/股USD/shareUSD/股 |
|---|---|---|---|---|
| H09 · 0M | 2026-10-06 → 2027-10-06 | 13.8662 | 17.2535× | $239.24 |
| H09 · 6M | 2027-04-06 → 2028-04-06 | 16.6289 | 17.2535× | $286.91 |
| H09 · 12M | 2027-10-06 → 2028-10-06 | 18.9070 | 17.2535× | $326.21 |
按本轮授权分别展示：ST01只定义3M短期价格；H09保留0M／6M／12M统一Core EPS×固定隐含PE的旧截面参考。两套模型不拼成一条时间曲线、不相互校准。H09的0M是现价锚；6M／12M是盈利兑现的条件价格，不是ST01延长至6M／12M。As authorized for this study, ST01 defines only a 3M short-term price. H09 retains its frozen 0M/6M/12M Core EPS × held implied P/E reference. Do not splice the systems into one price curve or calibrate them to each other. H09 0M is spot-anchored; its 6M/12M values are conditional earnings-realization prices, not extensions of ST01.按本輪授權分別展示：ST01只定義3M短期價格；H09保留0M／6M／12M統一Core EPS×固定隱含PE的舊截面參考。兩套模型不拼成一條時間曲線、不相互校準。H09的0M是現價錨；6M／12M是盈利兑現的條件價格，不是ST01延長至6M／12M。
3M用于描述未来一季的价格路径与事件风险，6M／12M用于研究盈利兑现。两者差距提醒你区分“基本面空间”和“短期何时被市场兑现”，差距本身不是额外买入信号。3M describes a one-quarter price path and event risk; 6M/12M study earnings realization. Their gap separates fundamental potential from near-term market realization. The gap itself is not an additional buy signal.3M用於描述未來一季的價格路徑與事件風險，6M／12M用於研究盈利兑現。兩者差距提醒你區分“基本面空間”和“短期何時被市場兑現”，差距本身不是額外買入信號。
返回NVDA个股研究Back to NVDA equity research返回NVDA個股研究下一步怎样升级What is required for an upgrade下一步怎樣升級
保留这次冻结预测，等目标到期后核验，不能事后改目标。补建财报、产品、政策消息的首发时间与事前预期面板；用新增、未用于选型的样本验证趋势及宏观各自的增量作用，再校准消息与波动区间。之后再决定ST02是否有足够证据。Preserve this frozen forecast and evaluate it at maturity without changing the target afterward. Build first-publication and prior-expectation panels for earnings, products and policy. Use new observations not used for selection to evaluate incremental trend and macro effects, then calibrate news and risk ranges before considering ST02.保留這次凍結預測，等目標到期後核驗，不能事後改目標。補建財報、產品、政策消息的首發時間與事前預期面板；用新增、未用於選型的樣本驗證趨勢及宏觀各自的增量作用，再校準消息與波動區間。之後再決定ST02是否有足夠證據。
本轮公开原创方法、派生结果和聚合误差；原始行情、利率CSV、逐日记录、StockAnalysis Pro数据与账户资料留在内部。日线端点与官方结算收盘存在少量历史来源差异；当前239.24已与官方前日收盘和Nasdaq档案核对。历史利率未完成ALFRED vintage核验，滞后不等于完整PIT证明。Publish original methods, derived outputs and aggregate errors. Raw prices, rate CSVs, daily records, StockAnalysis Pro data and account information remain internal. Historical bar endpoints have small differences from the archived settled-price source; the current $239.24 matches the official prior-session close and Nasdaq archive. Historical rates lack ALFRED vintage verification; lagging is not proof of full PIT.本輪公開原創方法、派生結果和聚合誤差；原始行情、利率CSV、逐日記錄、StockAnalysis Pro數據與賬户資料留在內部。日線端點與官方結算收盤存在少量歷史來源差異；當前239.24已與官方前日收盤和Nasdaq檔案核對。歷史利率未完成ALFRED vintage核驗，滯後不等於完整PIT證明。
Fed H.15发布口径Fed H.15 release conventionFed H.15發佈口徑 · FRED DGS10FRED DGS10FRED DGS10
Protocol SHA256 1a054ae7e4ffb08a2a1dc17037a1d58e48cbf9ed7e750b8174d0f1929be3871f
Public record SHA256 0124a552ff671cd4df67b410552ff77e89d0f65827901d3b92f79a7b2144e55f