# /// script # dependencies = [ # "marimo", # "matplotlib==3.11.1", # ] # requires-python = ">=3.12" # /// import marimo __generated_with = "0.24.0" app = marimo.App(width="medium") @app.cell(hide_code=True) def imports(): import marimo as mo import math import json from pathlib import Path import matplotlib.pyplot as plt from matplotlib.ticker import FuncFormatter return FuncFormatter, Path, json, math, mo, plt @app.cell(hide_code=True) def introduction(mo): mo.md(""" # Microsoft: value after the AI build-out **Valuation date:** 8 September 2026. **Financial baseline:** FY26 ended 30 June, reported 29 July. The $493.95 regular-session close is from Microsoft Investor Relations' Refinitiv widget (8 September, 16:00 ET). This is an analyst-assumption model, not management guidance or a forecast with calibrated probabilities. All financial quantities are USD billions; share counts are billions. Ten forward annual periods use the latest fiscal-year revenue as a run-rate baseline. Cash flows are placed at each year end; no precision is claimed for the fiscal/calendar stub or intervening balance-sheet movements. Our base estimate is about **$427/share** at a **9%** nominal dollar discount rate. A 20% margin-of-safety rule gives $341, rounded down to a **$340 entry threshold**. That threshold applies only while the operating thesis remains intact. It is not an order, a guaranteed floor, or a predicted trading path. """) return @app.cell def inputs(): # USD billions, shares billions. Reported FY26 values, not model assumptions. reported = { "period_end": "2026-06-30", "price_date": "2026-09-08 16:00 America/New_York", "price": 493.95, "revenue": {"Productivity": 139.996, "Cloud": 137.791, "Personal": 54.052}, "operating_income": 155.237, "cfo": 182.935, "cash_capex": 115.948, "da_and_other": 38.534, "sbc": 12.405, "finance_lease_additions": 24.608, "finance_lease_principal": 3.101, "cash_and_short_investments": 76.843, "debt_face": 46.136, "finance_lease_liability": 66.594, "shares_diluted": 7.453, "equity_and_other_investments": 36.348, "sources": { "release": "https://www.microsoft.com/en-us/Investor/earnings/FY-2026-Q4/press-release-webcast", "filing": "https://www.sec.gov/Archives/edgar/data/789019/000119312526323660/msft-20260630.htm", "call": "https://www.microsoft.com/en-us/investor/events/fy-2026/earnings-fy-2026-q4" } } # Assumptions are subjective. These are scenarios, not confidence intervals. common = {"tax": .20, "wacc": .09, "terminal_growth": .035, "terminal_roic": .20, "operating_cash_reserve": 20.0, "nwc_per_incremental_revenue": .05} scenarios = { "Downside": { "growth": {"Productivity": [.10,.09,.08,.07,.06,.055,.05,.045,.04,.035], "Cloud": [.24,.20,.16,.13,.10,.08,.07,.06,.05,.04], "Personal": [-.04,0,.01,.02,.02,.02,.02,.02,.02,.02]}, "margin": [.455,.44,.43,.42,.42,.42,.42,.42,.42,.42], "capex": [.52,.48,.43,.38,.33,.29,.27,.25,.24,.23], "da": [.14,.17,.19,.20,.20,.19,.18,.17,.16,.15], "terminal_roic": .15}, "Base": { "growth": {"Productivity": [.13,.12,.11,.10,.09,.08,.07,.06,.05,.04], "Cloud": [.28,.25,.22,.19,.16,.14,.12,.10,.08,.06], "Personal": [-.02,.02,.03,.03,.03,.03,.03,.03,.03,.03]}, "margin": [.46,.455,.45,.445,.445,.445,.445,.445,.445,.445], "capex": [.50,.45,.39,.33,.28,.25,.23,.22,.21,.20], "da": [.14,.17,.19,.20,.20,.19,.18,.17,.16,.14], "terminal_roic": .20}, "Upside": { "growth": {"Productivity": [.15,.14,.13,.12,.11,.10,.09,.08,.07,.06], "Cloud": [.32,.29,.26,.23,.20,.18,.16,.14,.12,.10], "Personal": [0,.03,.04,.04,.04,.04,.04,.04,.04,.04]}, "margin": [.465,.465,.47,.47,.47,.47,.47,.47,.47,.47], "capex": [.50,.45,.39,.33,.28,.25,.23,.21,.19,.18], "da": [.14,.17,.19,.20,.20,.19,.18,.17,.16,.14], "terminal_roic": .25} } return common, reported, scenarios @app.cell def valuation(common, mo, reported, scenarios): def forecast(case, cloud_shift=0.0, reinvestment_shift=0.0): segments = dict(reported["revenue"]) rows = [] for i in range(10): previous = sum(segments.values()) for segment in segments: growth = case["growth"][segment][i] + (cloud_shift if segment == "Cloud" else 0) segments[segment] *= 1 + growth revenue = sum(segments.values()) ebit = revenue * case["margin"][i] nopat = ebit * (1 - common["tax"]) da = revenue * case["da"][i] capex = revenue * (case["capex"][i] + reinvestment_shift) delta_nwc = (revenue - previous) * common["nwc_per_incremental_revenue"] fcff = nopat + da - capex - delta_nwc rows.append({"year": i+1, **segments, "revenue": revenue, "ebit": ebit, "nopat": nopat, "da": da, "capex": capex, "delta_nwc": delta_nwc, "fcff": fcff}) return rows def value(case, wacc=None, terminal_growth=None, cloud_shift=0.0, reinvestment_shift=0.0): wacc = common["wacc"] if wacc is None else wacc growth = common["terminal_growth"] if terminal_growth is None else terminal_growth if wacc <= growth: raise ValueError("Discount rate must exceed perpetual growth") rows = forecast(case, cloud_shift, reinvestment_shift) terminal_fcff = rows[-1]["nopat"] * (1 + growth) * (1 - growth/case["terminal_roic"]) pv_terminal = terminal_fcff / (wacc - growth) / (1 + wacc)**10 pv_explicit = sum(row["fcff"]/(1+wacc)**row["year"] for row in rows) enterprise = pv_explicit + pv_terminal bridge = reported["cash_and_short_investments"] - common["operating_cash_reserve"] - reported["debt_face"] - reported["finance_lease_liability"] equity = enterprise + bridge return {"per_share": equity/reported["shares_diluted"], "enterprise": enterprise, "equity": equity, "bridge": bridge, "terminal_share": pv_terminal/enterprise, "terminal_fcff": terminal_fcff, "pv_explicit": pv_explicit, "pv_terminal": pv_terminal, "rows": rows} def solve(fn, target, lo, hi): flo = fn(lo) - target if flo * (fn(hi)-target) > 0: raise ValueError("Root is not bracketed") for _ in range(80): mid = (lo + hi)/2 if (fn(mid)-target) * flo > 0: lo = mid else: hi = mid return (lo + hi)/2 results = {name: value(case) for name, case in scenarios.items()} reverse_cloud_shift = solve(lambda shift: value(scenarios["Base"], cloud_shift=shift)["per_share"], reported["price"], -.1,.20) implied_discount = solve(lambda rate: value(scenarios["Base"], wacc=rate)["per_share"], reported["price"], .04,.2) sensitivity = [{"wacc": w, "g": g, "value": value(scenarios["Base"], wacc=w, terminal_growth=g)["per_share"]} for w in [.08,.09,.10] for g in [.025,.035,.045]] mo.ui.table([{ "Scenario": name, "Value": round(result["per_share"],2), "Year 5 revenue": round(result["rows"][4]["revenue"],1), "Year 5 FCFF": round(result["rows"][4]["fcff"],1), "Year 10 FCFF": round(result["rows"][-1]["fcff"],1), "Terminal %": round(100*result["terminal_share"],1)} for name,result in results.items()]) return implied_discount, results, reverse_cloud_shift, sensitivity, value @app.cell(hide_code=True) def controls(mo): discount_control = mo.ui.slider(start=7, stop=12, step=.25, value=9, label="Discount rate (%)") growth_control = mo.ui.slider(start=2, stop=5, step=.25, value=3.5, label="Perpetual growth (%)") cloud_control = mo.ui.slider(start=-5, stop=5, step=.5, value=0, label="Annual cloud growth shift (percentage points)") mo.vstack([mo.md("## Test the base case"), discount_control, growth_control, cloud_control]) return cloud_control, discount_control, growth_control @app.cell(hide_code=True) def interactive( cloud_control, discount_control, growth_control, mo, scenarios, value, ): interactive_value = value(scenarios["Base"], wacc=discount_control.value/100, terminal_growth=growth_control.value/100, cloud_shift=cloud_control.value/100) mo.md(f"Under your assumptions: **${interactive_value['per_share']:,.0f}/share**; " f"20% margin-of-safety threshold **${.8*interactive_value['per_share']:,.0f}**. " f"Terminal value contributes **{100*interactive_value['terminal_share']:.1f}%** of enterprise value.") return @app.cell(hide_code=True) def method(mo): mo.md(""" ## Model conventions and limits **Enterprise DCF:** FCFF = EBIT × (1 − 20% tax) + depreciation/amortization − capital expenditure − incremental operating working capital. GAAP-like operating margins include stock compensation; there is no SBC add-back and no assumed buyback-driven share reduction. Current dilution is approximated by FY26's 7.453B weighted diluted shares. **Leases:** Capital expenditure includes cash PP&E and new finance-leased asset investment. Finance lease debt is deducted in the equity bridge; principal payments are not also subtracted from FCFF. Operating rent remains in projected operating margins; operating lease liabilities are therefore not deducted again. Future lease starts must fit either forecast capital investment or rent. This is an aggregate approximation, not a contract-by-contract lease forecast. FY26 had $26.7B of PP&E purchases in payables: our forward cash capex budgets must absorb their settlement; they are not separately deducted as debt. **Equity bridge:** add $76.843B cash/short investments, retain $20B operating liquidity, subtract $46.136B face-value debt and $66.594B finance lease liabilities. Face debt is conservative versus the reported $36.5B estimated market value. No separate premium for OpenAI or other strategic investments. Adding all $36.348B equity/other investments at book would add $4.88/share before tax and overlap; book is not an exit valuation. **Terminal reinvestment:** next-year NOPAT × (1 − perpetual growth / terminal return on invested capital). This charges for growth instead of assuming perpetual growth is free. ROIC is 15%/20%/25% in downside/base/upside. It is a terminal assumption, not a measured Azure return. A 9% discount rate and 3.5% perpetual nominal growth apply to all three cases. The 9% rate is our valuation hurdle, not an estimated CAPM/WACC fact; sensitivity is essential. No scenario probabilities are assigned. **Assumption rationale:** base consolidated revenue growth starts at 16.8%, near FY27 Q1 company guidance of 16–17%, but our annual path is not guidance. Intelligent Cloud growth fades from 28% to 6%; Productivity from 13% to 4%; Personal Computing recovers from −2% to 3%. Growth and margin forecasts already include Copilot/cloud adoption, so no extra AI business value is added. Operating margin fades from 46% to 44.5%, reflecting a larger share of lower-margin cloud sales and operating rent. Tax of 20% follows FY27 guidance. Initial $194B investment is our forward-year budget, compared with company guidance for over $50B in FY27 Q1 and growing FY27 spending; the $175B company figure is calendar 2026, a different period. D&A rises with installation of the asset fleet, from 14% of revenue to 20% in years 4–5 before fading to 14%; modeled investment falls from 50% of sales to 20%. Absolute base investment stays close to $190–205B while revenue almost triples over ten years. This is the principal efficiency bet, not a disclosed maintenance/growth split. A fleet replacement schedule is unavailable; faster hardware obsolescence or higher rent can invalidate the cash recovery. Working-capital funding is 5% of incremental revenue, including operating liquidity needs. No future acquisition program is modeled: acquisition-funded growth or new strategic investments would require additional cash deductions. ## Sources - [Microsoft FY26 release and timestamped quote](https://www.microsoft.com/en-us/Investor/earnings/FY-2026-Q4/press-release-webcast) - [FY26 10-K: financial statements and Notes 6, 10, 13, 15](https://www.sec.gov/Archives/edgar/data/789019/000119312526323660/msft-20260630.htm) - [29 July call: FY27 guidance, lease reclassification, RPO](https://www.microsoft.com/en-us/investor/events/fy-2026/earnings-fy-2026-q4) Public general research; readers decide suitability themselves. Intrinsic value is an uncertain estimate that changes with assumptions, not a price the market must return to. """) return @app.cell(hide_code=True) def exports( Path, common, implied_discount, json, math, mo, reported, results, reverse_cloud_shift, scenarios, sensitivity, ): base = Path(mo.notebook_location()) export = {"reported": reported, "common": common, "scenarios": scenarios} (base / "inputs.json").write_text(json.dumps(export, indent=2)+"\n") (base / "results.json").write_text(json.dumps({"scenarios": results, "sensitivity": sensitivity, "reverse_cloud_shift": reverse_cloud_shift, "implied_discount": implied_discount, "entry_threshold": math.floor(.8*results["Base"]["per_share"]/10)*10}, indent=2)+"\n") print("Saved inputs.json and results.json alongside the notebook") return (base,) @app.cell(hide_code=True) def figures(FuncFormatter, base, mo, plt, reported, results): plt.rcParams.update({"font.family":"DejaVu Sans", "font.size":12, "figure.facecolor":"white", "axes.facecolor":"white", "text.color":"#182739", "axes.spines.top":False,"axes.spines.right":False,"axes.spines.left":False, "axes.spines.bottom":False,"xtick.color":"#526174","ytick.color":"#182739", "svg.fonttype":"path"}) for _compact in [False,True]: _fig,_ax=plt.subplots(figsize=(4.5,3.5) if _compact else (9,3.5),layout="constrained") _names=["Downside","Base","Upside"] _vals=[results[n]["per_share"] for n in _names] _ax.barh(range(3),_vals,color=["#aebac6","#234d7b","#7893aa"],height=.42) for _i,_v in enumerate(_vals): _ax.text(_v,_i+.27,f"${_v:.0f}",va="bottom",ha="center",fontsize=11 if _compact else 14,fontweight="bold") _ax.axvline(reported["price"],color="#a5663c",lw=1.6,ls=(0,(4,3))) _ax.text(reported["price"],2.65,"Market $493.95",ha="center",fontsize=10 if _compact else 12,color="#8d522c") _ax.set_yticks(range(3),_names,fontsize=11 if _compact else 13) _ax.set_xlim(0,690);_ax.set_ylim(-.5,3) _ax.set_xticks([0,200,400,600]);_ax.xaxis.set_major_formatter(FuncFormatter(lambda x,p:f"${x:.0f}")) _ax.tick_params(length=0,pad=8);_ax.grid(axis="x",color="#e4e8ec",lw=.6);_ax.set_axisbelow(True) _suffix="-mobile" if _compact else "" _fig.savefig(base/f"valuation{_suffix}.svg",bbox_inches="tight") _fig.savefig(base/f"valuation{_suffix}.png",dpi=150,bbox_inches="tight");plt.close(_fig) _fig,_ax=plt.subplots(figsize=(4.5,3.8) if _compact else (9,3.5),layout="constrained") _rows=results["Base"]["rows"] _x=[r["year"] for r in _rows] _ax.plot(_x,[r["nopat"] for r in _rows],color="#7893aa",lw=2,label="After-tax operating profit") _ax.plot(_x,[r["capex"]-r["da"]+r["delta_nwc"] for r in _rows],color="#a5663c",lw=2,label="Net reinvestment") _ax.plot(_x,[r["fcff"] for r in _rows],color="#234d7b",lw=2.7,label="Cash flow to the firm") _ax.scatter([5,10],[_rows[4]["fcff"],_rows[9]["fcff"]],color="#234d7b",s=22,zorder=4) for _i in [4,9]: _ax.annotate(f"${_rows[_i]['fcff']:.0f}B",(_i+1,_rows[_i]["fcff"]),xytext=(0,-22 if _i==9 else -25),textcoords="offset points",ha="center",fontsize=10 if _compact else 12,color="#234d7b") _ax.set_xlim(.7,10.5);_ax.set_ylim(-10,390) _ax.set_xticks([1,3,5,7,10],["Year 1","3","5","7","10"]) _ax.set_yticks([0,100,200,300]);_ax.yaxis.set_major_formatter(FuncFormatter(lambda x,p:f"${x:.0f}B")) _ax.tick_params(length=0,pad=8,labelsize=10 if _compact else 11) _ax.grid(axis="y",color="#e4e8ec",lw=.7);_ax.set_axisbelow(True) _ax.legend(loc="upper left",frameon=False,fontsize=9 if _compact else 11) _fig.savefig(base/f"cash-recovery{_suffix}.svg",bbox_inches="tight") _fig.savefig(base/f"cash-recovery{_suffix}.png",dpi=150,bbox_inches="tight");plt.close(_fig) for _path in base.glob("*.svg"): _path.write_text("\n".join(line.rstrip() for line in _path.read_text().splitlines())+"\n") mo.image(base/"valuation.svg") return @app.cell(hide_code=True) def annual_table(mo, results): mo.ui.table(results["Base"]["rows"],label="Base forecast: all annual cash-flow inputs, USD billions") return if __name__ == "__main__": app.run()