TL;DR
TL;DR
At $385.065 (intraday, 2026-07-10) — −29.0% off its $542.07 52-week high (2025-10-28), +9.1% off its $352.83 low (2026-06-25), −23.2% YoY (vs $501.48 on 2025-07-10), after a −34.9% peak-to-trough de-rating; "52-week," not "all-time" (history.json is a 1-year window) — Microsoft trades at 22.8× TTM earnings (the lowest P/E in the hyperscaler cohort: GOOGL 27.4×, AMZN 29.6×, ORCL 24.7×) and ~8.9–9.0× reconciled EV/TTM-revenue, but 39.2× TTM free cash flow. The whole debate reduces to one question — does the AI capex convert to free cash flow? — and both cases name the same event as their falsifier #1: the confirmed Q4/FY2026 print, Wed 2026-07-29.
- Bull. The current company is materially better than the stale FY2025 tape and the market de-rated it anyway: 9M revenue accelerated to +17.8% YoY (above FY2025's +14.9%), operating margin expanded +151 bps to 47.40%, net income +31.3%, Azure +40% (the #2 cloud platform, out-growing AWS), and MSFT is the only aggressive AI-builder still generating strongly positive TTM FCF (+$72.9B vs AMZN −$2.5B, ORCL −$23.7B) from an approximately net-cash balance sheet (+$38,010M). It is the cheapest, highest-FCF name in the cohort, priced as if the capex will not convert.
- Bear. Concede all of that — none of it is contested. But the "cheapest P/E" is an accounting artifact: the E is capex-deferred (9M capex $80,146M is 3.34× the depreciation now in the P&L, ~$75B/yr not yet expensed on a gross PP&E base that nearly doubled to $395B) and OpenAI-mark-flattered (a +$7,583M Q2 non-operating gain). On the un-deferrable metric — cash — the same price is 39.2× FCF growing just +2.8%, a 2.55% yield, and the market has already erased ~$286B in a day on a one-point Azure miss (the −10.0% Q2 print). Re-rating the cash multiple toward 25–30× on flat FCF is −24% to −36%.
- Priced-in break-even. At the adopted broad reconciled EV $2,822,418M the price is a mature 30× free-cash-flow multiple iff capex/revenue falls to ~23.5% (roughly the FY2025 level) from the current 9M 33.1% — i.e. FCF re-expands ~$24B to ~$97B in one mix-shift. If capex stays ~33% and FCF grows at the +2.8% pace, growing into a 30× multiple takes ~10–16 years, so the multiple compresses −24% to −36% instead. The entire conclusion flips on whether AI capex/revenue normalizes from ~33% toward ~23% (FCF converts) or stays suppressed. No buy/sell/hold — research tooling, not advice.
Watch next — a cold reader can act on these:
| Window | Event | If it lands bullish | If it lands bearish |
|---|---|---|---|
| Wed 2026-07-29 (after close) | Q4/FY2026 earnings print — the fulcrum, both cases' falsifier #1; FY2026 full-year + FY2027 capex guide + Azure rate | FY2026 FCF grows, capex/revenue starts down from 33.1%, Azure holds ≥ ~35%, gross margin stabilizes ~67.6% — the de-rating overshot | Azure decelerates below ~35%, the FY2027 capex guide steps up again (above the 9M $80,146M / TTM $97,225M cash-capex run-rate) with FCF still flat, gross margin drifts below 67.63% — the 22.8× P/E is a value trap |
| each 10-Q; next at the Q4 print | Depreciation vs gross-margin recovery | Depreciation growth (+52.9% on the 9M) falls below revenue growth; gross margin stops the 69.05% → 67.63% drift | Depreciation keeps outrunning gross-margin recovery as the $395B PP&E base matures into the cost line |
| each 10-Q; FY2026 10-K ~late-Oct 2026 | OpenAI relationship — equity-method marks + backlog concentration | LongTermInvestments ($33,683M) stabilizes, non-OpenAI backlog carries the growth, the mark stops swinging GAAP NI | The GAAP earnings keep leaning on a swinging OpenAI mark (Q1 −$3.1B / Q2 +$7.6B / Q3 −$14M); OpenAI (external analyst est. ~45% of the $627B RPO — not in the filings; see Sources) shifts more workload off Azure |
| ~mid-September 2026 | Annual dividend raise + buyback refresh (cadence; NOT an 8-K in the index) | Another ~10% raise from $0.91 quarterly; a fresh multiyear authorization, comfortably inside FCF | Raise/buyback pace slows as the capex draw claims FCF |
| ~court-driven; ~Nov 2026 → Feb 2027 | Cloud-antitrust overhang — FTC probe, EU DMA (Azure, ~Nov 2026), UK CMA SMS (~Feb 2027) | Probes close or resolve benign; no conduct remedy on Azure/Copilot licensing | A gatekeeper designation or conduct remedy constrains Azure/Copilot licensing terms |