The de-rating happened on a real concern — capex is no longer transitory — but it overshot a set of operating fundamentals that have not flinched. Azure is supply-bound at +39% with a $627B RPO queue behind it. Margins hardened through the worst of the build. Copilot crossed 20M seats with seat-add growth accelerating, not decelerating. The cleanest argument for owning Microsoft here is that the forward multiple (21.8x) now trades a half-turn below peer-cluster average despite the franchise's structurally higher margin profile and lower revenue concentration. We rate BUY with a $491 12-month target — modest premium to peers, but a clean expression of the asymmetry.

1. Q4 FY26 print (late July 2026). Azure cc growth re-acceleration above 40% would confirm capacity unlock; deceleration toward 36% would mark the demand-digestion bear thesis. Watch the AI-vs-non-AI contribution split.
2. CY26 capex realised vs the $190B target. Lower than $190B is a positive FCF surprise; higher means the component-pricing inflation is structural, not cyclical. The $5B Q4 capex component-price callout from Q3 FY26 is the bellwether.
3. Copilot Pricing Engine — June 2026 GitHub Copilot switch. The pivot to user-plus-usage is the leading test of whether enterprise AI unit economics support consumption pricing. Early ARPU data will inform the FY27 Copilot revenue ramp.

Four signals emerge from the eight-quarter cadence. 1) Agent / agentic frequency has 10x'd from 6 mentions (Q4 FY24) to 65 (Q3 FY26) — the clearest rhetorical pivot in the cycle, signaling that the platform thesis has migrated from foundation-model wrapping to autonomous workflow capture. 2) Capex frequency has nearly doubled (10 → 18), tracking the absolute dollar trajectory; this is the rare case where management is openly leaning into the question rather than minimising it. 3) Inference references spiked Q2 FY25 (12 mentions) alongside the cost-curve commentary, then resurged in FY26 as throughput gains became a balance-sheet defence. 4) OpenAI mentions tracked the contractual cycle — peak Q1 FY26 (18) on the new $250B deal, moderating thereafter as the conversation shifted to consumption execution.

“We will increase our total AI capacity by over 80% this year and roughly double our total data center footprint over the next 2 years, reflecting the demand signals we see. When I look at the entirety of these high-value agent systems and when we look at the efficiency and fungibility of our fleet, that's what gives us the confidence to invest, both the capital and the R&D talent to go after this opportunity.”

“I have said we are now likely to be short capacity to serve the most important things we need to do, which is Azure, our first-party applications. We need to invest in product R&D and we're doing end-of-life replacements in the fleet. I thought we were going to catch up; we are not. Demand is increasing. Azure probably does bear most of the revenue impact, and it is safe to say that the number could be higher.”

“At the end of the day, it is going to come from some eval and outcome that a business has, where these agents that are working on behalf of users or with users is creating value. When they start seeing that the task trajectory is compressing the workflow, improving revenue, decreasing costs, that is what's driving usage. We have a structural position in knowledge work, coding, security — then you couple that with the right business model, which is user plus usage.”

The eight-quarter narrative arc reads as a capacity-bound demand thesis hardening into a capital-allocation question. In FY25 the framing was 'short power and space'; in FY26 it has migrated to 'short GPUs and CPUs' as long-lived shells came online and the bottleneck moved upstream. Nadella's vision quote is the bull case in plain language: the bet is that high-value agentic workflows convert eval-driven usage into revenue. Hood's downside admission is the bear case in plain language: the number 'could be higher' is a candid acknowledgement that the capex curve has not stopped bending. The conditional-bull quote names the test: when enterprise customers measure agent outcomes against workflow compression, revenue growth, and cost reduction — that's when the platform monetises.

Microsoft trades at 21.8x forward earnings versus a 22.6x peer-cluster average — a ~3% discount, despite a forward EBITDA margin (52%) that runs ~17 points ahead of the cluster (35%). On forward EV/EBITDA the discount is wider: 15.8x vs 17.3x, a ~9% gap. CRM is the only structurally cheaper name in the cluster, but it grows roughly two-thirds the pace; ORCL trades higher on a faster growth print (32% FY27) but at higher leverage; GOOGL and AMZN both command premiums on platform breadth. The honest read: Microsoft's franchise quality is not currently reflected in the multiple, and the gap is wider than the modest growth differential alone can justify.

We rate BUY with a 12-month target of $491, ~16% above spot. The setup is a clean expression of post-correction asymmetry: forward multiples now trade at a discount to peers despite structurally higher margins, Azure printing four quarters of 39% cc capacity-bound growth, RPO compounding at +99% YoY, and a Copilot consumption-pricing pivot scheduled for June 2026. Risks are real — CY26 capex of ~$190B leaves no room for component-price inflation upside, and OpenAI concentration in RPO is now ~45%. We see ~1.9:1 reward-to-risk on Bull ($624, +47%) vs Bear ($317, -25%) outcomes. FMP's standalone DCF cross-check at $331 reflects the bear case's WACC and margin assumptions; our blended $491 reflects a more capacity-unlock-positive read.