MSFT / M01 / MODEL RESEARCH
Microsoft M01：经营成本和股数为何没有提高目标价准确度Microsoft M01: why costs and shares did not improve target-price accuracyMicrosoft M01：經營成本和股數為何沒有提高目標價準確度
返回完整研究演进Back to model evolution返回完整研究演進 · 返回MSFT个股模型值Back to MSFT equity model values返回MSFT個股模型值
研究问题与公式Research question and equations研究問題與公式
保持H09收入路径，独立预测GAAP经营成本C=收入−营业利润，再由收入减成本得到利润率。稀释股数用前后两组四季的日数加权股数变化，按四季度半衰期递推；并非假设公司承诺未来回购。Keep H09 revenue, forecast GAAP operating cost C=revenue−operating income, and derive margins from revenue minus cost. Diluted shares follow the change between two four-quarter day-weighted share blocks with a four-quarter half-life; this does not assume promised future repurchases.保持H09收入路徑，獨立預測GAAP經營成本C=收入−營業利潤，再由收入減成本得到利潤率。稀釋股數用前後兩組四季的日數加權股數變化，按四季度半衰期遞推；並非假設公司承諾未來回購。
Cost_q = GAAPRevenue_q − GAAPOperatingIncome_q gC = ln(DailyCost_latest / DailyCost_sameQuarterLastYear) /4 gS = ln(DayWeightedShares_recent4 / DayWeightedShares_prior4) /4 Shares_j = LatestShares × exp[gS × Σ(k=1..j)2^(-(k−1)/4)] OperatingIncome_j = ForecastRevenue_j − ForecastCost_j CoreEPS_j = OperatingIncome_j ×0.85 / Shares_j PE0 = P0 / M01CoreEPS0; Price_h = M01CoreEPS_h × PE0
收入与成本各用最近12季季节性和四季增长半衰期；股数路径同样四季衰减。成本含折旧、股权薪酬及营业费用，现金资本开支未重复全额扣减。Q4推导股数保留近似标记。Revenue and costs use separate 12-quarter seasonal estimates and four-quarter growth half-lives; share drift also decays over four quarters. Costs retain depreciation, stock compensation and operating expenses; cash capital expenditure is not deducted again in full. Derived Q4 shares retain approximation labels.收入與成本各用最近12季季節性和四季增長半衰期；股數路徑同樣四季衰減。成本含折舊、股權薪酬及營業費用，現金資本開支未重複全額扣減。Q4推導股數保留近似標記。
输入截面与失败决定Input dates and rejection輸入截面與失敗決定
当前研究输入为2026-10-07官方收盘，三年历史标签统一截至2026-10-06。0M仍为市价锚；本版本价格不是10月8日当前主表。The live research input uses the October 7 official close; historical labels remain frozen through October 6.0M remains a market-price anchor. This version’s prices are not the October 8 retained main table.當前研究輸入為2026-10-07官方收盤，三年曆史標籤統一截至2026-10-06。0M仍為市價錨；本版本價格不是10月8日當前主表。
经营成本与净稀释股数递推未提高目标价准确度，保留失败版本。Forecasting operating costs and net diluted shares did not improve target-price accuracy. The rejected version is retained.經營成本與淨稀釋股數遞推未提高目標價準確度，保留失敗版本。
两个期限的价格误差高于M00；0M和6M利润率误差也恶化。成本单独、股份单独消融仅诊断，未替换失败的联合主候选。Both price horizons had larger errors than M00;0M and 6M margin errors worsened. Cost-only and shares-only ablations were diagnostic and did not replace the rejected joint candidate.兩個期限的價格誤差高於M00；0M和6M利潤率誤差也惡化。成本單獨、股份單獨消融僅診斷，未替換失敗的聯合主候選。
保留原三年月度研究：不与季度样本混排Preserve original monthly studies: separate from quarterly samples保留原三年月度研究：不與季度樣本混排
M01/M02使用原三年月末与分析师观察日面板，标签截至10月6日；M03另用财报后季度面板和10月8日截点。两个面板不是同一批起点，不可直接按误差大小排序版本。M01/M02 retain the original three-year month-end and analyst-observation panels through October 6. M03 instead uses post-earnings quarterly origins and an October 8 cutoff. These are different origin panels; their errors should not be ranked as if they came from identical samples.M01/M02使用原三年月末與分析師觀察日面板，標籤截至10月6日；M03另用財報後季度面板和10月8日截點。兩個面板不是同一批起點，不可直接按誤差大小排序版本。
| 版本Version版本 | 6M：成熟30 / 待到期66M:30 mature /6 pending6M：成熟30 / 待到期6 | 12M：成熟24 / 待到期1212M:24 mature /12 pending12M：成熟24 / 待到期12 |
|---|---|---|
| M00 | 15.15% | 10.41% |
| M01 | 15.30% | 11.15% |
| M02 | 14.67% | 11.11% |
| 价格不变Price unchanged價格不變 | 14.66% | 10.04% |
| 版本 / 对照Version / reference版本 / 對照 | 到期收盘MAPETerminal-close MAPE到期收盤MAPE |
|---|---|
| M00 | 10.10% |
| M01 | 10.74% |
| M02 | 10.86% |
| 分析师平均目标Mean analyst target分析師平均目標 | 13.20% |
| 价格不变Price unchanged價格不變 | 9.82% |
36个分析师真实观察日按同日重算模型，其中24个12M共同终点成熟。目标意见差与真实误差分开，分析师未参与训练。M01与M02均未通过完整冻结筛选。Models were recomputed on 36 actual analyst-observation dates, with 24 mature common 12M endpoints. Target disagreement is separate from actual error, and analysts were not training inputs. M01 and M02 failed their complete frozen screens.36個分析師真實觀察日按同日重算模型，其中24個12M共同終點成熟。目標意見差與真實誤差分開，分析師未參與訓練。M01與M02均未通過完整凍結篩選。
计算通过，模型有效性仍待验证Arithmetic passed; model validity remains unproved計算通過，模型有效性仍待驗證
独立复核确认计算、日期和来源约束一致；这不等于采用筛选通过。历史资料按已公开日期重建，Q4股数与日分配标签含近似，PIT原始版本、独立样本外和前瞻有效性尚未证明。Independent review confirmed arithmetic, dates and source-availability constraints. This is distinct from passing adoption screens. Historical inputs are reconstructed by public dates, with Q4-share and daily-allocation approximations. Original immutable PIT vintages, independent out-of-sample results and prospective validity remain unproved.獨立複核確認計算、日期和來源約束一致；這不等於採用篩選通過。歷史資料按已公開日期重建，Q4股數與日分配標籤含近似，PIT原始版本、獨立樣本外和前瞻有效性尚未證明。
原预测和失败版本均保留。当前M00是条件研究参照，并非正式已验证模型。M03完整CoreEPS窗口0M/6M/12M分别35/33/31个，三个期限对数误差都比M00大；峰值评分并没有证明盈利预测改善。Original forecasts and rejected versions are preserved. M00 is a conditional research reference, not a validated production model. M03 has 35/33/31 complete Core EPS windows at 0M/6M/12M; all three log-error panels worsened versus M00. Peak scores did not validate better earnings forecasts.原預測和失敗版本均保留。當前M00是條件研究參照，並非正式已驗證模型。M03完整CoreEPS窗口0M/6M/12M分別35/33/31個，三個期限對數誤差都比M00大；峯值評分並沒有證明盈利預測改善。
本页只展示自身模型预测、原创方法与聚合对照结论。商业原始逐日分析师数据、逐季度精确目标记录和供应商原图线保留内部。This page publishes own forecasts, original methods and aggregate comparative conclusions. Commercial daily analyst records, exact per-quarter vendor targets and original vendor charts remain internal.本頁只展示自身模型預測、原創方法與聚合對照結論。商業原始逐日分析師數據、逐季度精確目標記錄和供應商原圖線保留內部。
微软财报来源Microsoft issuer earnings微軟財報來源 · SEC原始申报SEC original filingsSEC原始申報 · 分析师数据来源Analyst data source分析師數據來源