Will Microsoft's FY26 total capital expenditures exceed $150B?

FY26 implied minimum is $144.3B (H1 $72.4B + Q3 $31.9B + Q4 >$40B), so the $150B threshold sits right at the boundary requiring Q4 to print ~$45B+ rather than the floor $40B guide. Three consecutive upward revisions in 12 months ('moderate' → 'higher than FY25' → $190B CY26) establish a clear pattern that management's stated Q4 guides have come in light, and the explicit $5B memory inflation step-up plus the $190B CY26 commitment (which requires sustained high run rates) tilts Q4 toward the $42-48B zone. The capacity-ahead-of-schedule signal (Wisconsin Fairwater 6 weeks early, +39% Azure cc) confirms demand absorption faster than expected, weakening any incentive to under-build into year-end. Counter-anchor: Q4 is only 2 months out, the explicit '>$40B' guide is more disciplined than prior vague guides, and finance lease timing variability could surprise low. Net: modestly above 50/50, calibrated to the 0.55-0.60 range given the established upward-revision pattern offset by the precise visibility of a near-term print.

Q4 FY26 (Apr-Jun 2026) is the peak memory inflation quarter — DRAM/HBM contract resets fully realized at the $25B/$190B CY26 magnitude — and the >$40B Q4 guide was set explicitly with $5B of memory pricing already baked in but on a conservative anchor that management has been revising up. The arithmetic: $72.4B H1 + $31.9B Q3 = $104.3B booked, leaving the threshold at ~$45.7B Q4 capex to clear. With three consecutive upward revisions in 12 months establishing the path-of-least-resistance upward, Q4 historically printing at-or-above guide, finance lease deliveries that lulled in Q3 likely clustering into Q4 (Atlanta + Wisconsin AI super factory ramp), and back-loading dynamics from the $190B CY26 commitment, Q4 capex landing at $45-48B is more probable than landing flat at $40B. The OpenAI deal absorption and capacity-constrained-through-FY26 framing both argue management has weak incentive to under-build into the print. Bounded by the constraint that Q4 is only one quarter and Hood's guide range starts at $40B not $45B — so the YES path requires execution above guide rather than just at guide.

Taking the skeptic angle: simple arithmetic (H1 $72.4B + Q3 $31.9B + Q4 $40B) lands at ~$144.3B, which is $5.7B BELOW the $150B threshold. Q3 actually printed DOWN sequentially due to finance lease timing, validating Hood's prior guidance and demonstrating these timing dynamics are real, not theoretical. To clear $150B, Q4 must exceed the >$40B guide by ~14% — but management's '>$40B' framing is a directional floor and historical pattern shows quarterly capex tends to land closer to the stated floor than 25%+ above. The CY2026 $190B guide is back-loaded into FY27 (Q1+Q2 FY27 ~$118B), so it does NOT mechanically force FY26 above $150B. Memory inflation is real but only $5B incremental in Q4 guide, already baked into the >$40B figure.

FY26 capex trajectory: H1 $72.4B + Q3 $31.9B + Q4 >$40B guide = $144.3B+ at the low end, requiring Q4 to print ~$45.7B to clear $150B — a 14% beat above the guide floor. The $190B CY26 explicit guide is the critical new data point: if CY26 spend is distributed with meaningful Q4 FY26 weight (driven by memory inflation flowing through and capacity-constrained demand), Q4 could reach $45-50B. Management has revised UP three consecutive times in 12 months and Hood's Q3 guide of 'decrease sequentially' was accurate but shallow (from $37.5B to $31.9B), with Q4 explicitly guided to step back up above $40B with $5B attributed to component pricing alone. The sibling markets (Azure cc +39%, Cloud GM above 65%) both resolved YES, suggesting a pattern of execution at or above stated guidance. Primary uncertainty is whether memory inflation and finance-lease deliveries cluster sufficiently in Q4 FY26 to push that single quarter past $45.7B.

FY26 implied total stands at ~$144.3B from H1 actuals ($72.4B) + Q3 actual ($31.9B) + Q4 guide (>$40B), placing the $150B threshold just above the baseline path but within reach. Three compounding YES factors tip the balance modestly above 50%: memory inflation is empirically realized at scale (~$25B of CY26 $190B is component-pricing-driven), meaning Q4 FY26 capex is likely to print materially above the $40B floor; Wisconsin Fairwater came online 6 weeks ahead of schedule, confirming capacity absorption is running faster than modeled and management has weak incentive to throttle Q4 spend; and the OpenAI $250B Azure commitment is still being translated into infrastructure schedules with pull-through timing uncertain but skewed toward H2 FY26. The counter-anchor — Hood's Q3 sequential-decrease guidance — proved accurate for Q3 ($31.9B down from $37.5B) but Q4 is explicitly guided higher (>$40B), so the Q3 lull is already priced into the arithmetic. For FY26 to breach $150B, Q4 needs only ~$45.7B, which is consistent with the $190B CY26 guide if roughly evenly split between H2 FY26 and H1 FY27.

FY26 cumulative through Q3 stands at $104.3B (H1 $72.4B + Q3 $31.9B actual), leaving Q4 needing $45.7B+ to breach $150B — a ~14% beat over Hood's '>$40B' floor guide. The skeptic case is real: Q3 printed exactly as guided (sequential decrease on finance lease timing lull), and '>$40B' as a floor likely implies a $42-44B expected midpoint, landing FY26 at $146-148B. The bull case rests on three structural factors: (1) three consecutive upward revisions establish a pattern of underguiding, (2) $25B of memory/HBM inflation in the $190B CY2026 guide is already flowing through and would compress the spread between floor and actual Q4 capex, and (3) the $190B CY26 commitment requires either a very heavy H1 FY27 or a Q4 FY26 meaningfully above $40B — the arithmetic pressure on Q4 is real. The $150B threshold sits within the plausible range but above the base trajectory, warranting a modestly above-even probability.

FY26 capex tracks $144.3B+ (H1 $72.4B + Q3 $31.9B + Q4 >$40B). To exceed $150B threshold, Q4 needs ~$45.6B+. Three consecutive upward revisions in 12 months and realized memory inflation (~$25B in CY26) argue for elevated probability, but Q4 guidance quantified as '>$40B' without aggressive step-up and only 2-month window constrain upside surprise. Capex/OCF trip wire (1.03x CY26 forward) is live but does not guarantee FY26 breakout.

Q3 earnings confirms $190B CY2026 explicit capex guide, third upward revision in 12 months establishing a trend. Memory inflation ($25B of CY26) realized as analyzed; Q4 FY26 capex guidance >$40B must reach ~$45B+ to cross $150B fiscal-year threshold given H1 actual $72.4B + Q3 actual $31.9B = $104.3B. Capex/OCF crosses 1.03x trip wire; discipline flag escalates yellow → amber. Pattern of management revisions tilts probability moderately higher from baseline 48%, bounded by Q4 guidance constraint and only 2 months until FY26 close.

H1 actual ($72.4B) + Q3 actual ($31.9B) + Q4 guidance (>$40B) implies ~$144-150B, placing FY26 right at threshold. Third upward revision in 12 months establishes momentum toward YES; capex/OCF crossing 1.0x signals capital strain. Memory inflation ($25B of CY26 guide) is already realized cost, likely pushing Q4 to $42-45B range versus $40B base, carrying FY26 past $150B. Only 2 months of fiscal year remain; Hood's Q3-down guidance held, but trend argues for Q4 acceleration.