MSFT · Microsoft Corporation · Stock research

.

MSFT — Microsoft Corporation

Current Thesis

The July note framed the 2026-07-29 print as a binary landmine into a hostile tape. It resolved to the upside, and the narrative leg reversed polarity: Microsoft is now the megacap that gets paid for AI capex rather than punished for it. Q4 FY26 (reported 2026-07-29 AMC) delivered revenue $90.01B (+18% YoY) against a ~$87.6B consensus, non-GAAP EPS $4.74 versus $4.33 expected, and Azure +43% YoY against a 39–40% guide — with Azure crossing $100B of annual revenue for the first time (+41% for the fiscal year). The line that re-rated the stock was not growth but spend: calendar-2026 capex guidance came down to roughly $175B from the ~$190B set on 2026-04-29, and management framed FY27 as steady rather than a step-up. Shares moved +8.13% after hours from the $390.54 close, while Meta fell 9.6% and Alphabet 15% in the same reporting window. The follow-through has been mechanical: the 2026-08-14 close of $495.40 sits 8.0% under the $538.66 record close of 2025-10-28, three-month return +17.7%, RSI(14) 84.8. The leg an investor is buying now is "capex discipline plus $678B of contracted backlog equals the only positive-free-cash-flow hyperscaler in 2027" — a narrative that is fresh, working, and arriving into a ten-week catalyst vacuum at the most overbought reading of the cycle.

Bullish and bearish views on Microsoft Corporation

The model's bull view on Microsoft Corporation (MSFT), in brief: Azure beat its own guide by 300–400bps. +43% YoY in Q4 FY26 (2026-07-29) versus the 39–40% constant-currency guide, after +40% in Q3 FY26. Two consecutive quarters of re-acceleration against a bear case built on deceleration. Commercial RPO $678B, +84% YoY (2026-07-29).… The bear view: RSI(14) at 84.8 on 2026-08-14 with no company-specific event scheduled until the Q1 FY27 print in late October. Both cases follow in full.

Bull Case

• Azure beat its own guide by 300–400bps. +43% YoY in Q4 FY26 (2026-07-29) versus the 39–40% constant-currency guide, after +40% in Q3 FY26. Two consecutive quarters of re-acceleration against a bear case built on deceleration.
• Commercial RPO $678B, +84% YoY (2026-07-29). Contracted-but-unrecognized revenue of that size converts the capex argument from speculative buildout to pre-sold capacity; management stated demand exceeds current supply.
• The capex number fell, not rose. CY2026 guidance moved to ~$175B from the ~$190B set on 2026-04-29 — the single differentiator versus Alphabet (-15%) and Meta (-9.6%) in the same week. Q4 capex was $41B; FY26 property-and-equipment purchases totalled $115.95B.
• Copilot seat curve steepened: 30M paid seats at Q4 versus 20M at Q3 FY26; GitHub Copilot users 50M (2026-07-29). Grok 4.6 shipping inside GitHub Copilot (2026-08-14) extends the model-agnostic surface rather than narrowing it.
• Sell-side capitulated upward within two weeks. Wells Fargo to $700 from $650 (2026-08-12, the Street high), JPMorgan to $625 from $550 (2026-08-13), Bernstein to $660 from $647. Consensus sits near $569.56 across 56 analysts — versus the July note's $558.77 consensus against a $393.82 reference close, the gap to consensus has compressed sharply as price did the work.
• The free-cash-flow framing is the new differentiator: a 2026-08-14 sell-side call argued Microsoft emerges as big tech's FCF leader in 2027 while Alphabet and Meta stay negative — the same logic that drove the 07-29 divergence, now being extrapolated a year forward.

Bear Case

• RSI(14) at 84.8 on 2026-08-14 with no company-specific event scheduled until the Q1 FY27 print in late October. Extension of this order is normally worked off by time or price, and there is no earnings date inside the window to reset expectations.
• The capex cut is partly cosmetic. Coverage of the 07-29 call attributes a meaningful part of the CY26 reduction to an accounting-classification change effective FY2027. Pre-print previews framed FY27 capex expectations at $255–260B. If the disclosed FY27 figure lands at that level, the "discipline" that separated Microsoft from Alphabet and Meta on 07-29 becomes a timing artifact.
• Coatue went the other way, +18.4% to 2.9M shares, and Pershing Square added while exiting Alphabet. The snapshot predates the print and says nothing about August flow, but it is the only positioning data on the tape.
• Component-cost inflation has not gone away. The April CY26 plan carried roughly $25B of memory-price inflation. Memory strength is a revenue tailwind for suppliers and a direct cost line for Microsoft.
• AI financing conditions are tightening at the edges. WSJ reported 2026-08-15 that Nvidia is downsizing plans for a $250B guarantee of an OpenAI data centre; AMD is issuing $5B of bonds (2026-08-14) to fund AI server expansion. The capital stack behind the buildout is now a visible variable.
• The model layer is commoditizing around Microsoft. OpenAI annualized revenue topped $40B ahead of a reported $1T IPO push (2026-08-14); Anthropic posted >$11.5B of Q2 revenue (Bloomberg, 2026-08-14) with bankers reportedly modelling $200B by 2028. Cheaper, more numerous frontier models cut both ways for a company that both hosts and competes with them.

Setup & Price Structure

The 2026-08-14 close of $495.40 is 8.0% below the $538.66 record close of 2025-10-28, against a 52-week low of $349.20 printed during the June derating. The structural sequence is clean: the $400 shelf lost in June was reclaimed on the 07-29 gap; the $455 area — the level the July note named as the threshold a momentum leg would require — was taken out in early August and is now the shelf beneath the advance rather than resistance above it. The post-print gap zone near $422, where shares traded in the 07-29 after-hours session from a $390.54 close, marks the floor of the entire re-rating leg.

What the structure does not offer is a defined-risk entry. Price is roughly 18 trading sessions into a vertical move, RSI(14) at 84.8, with the nearest reference support a long way below. Crowding observables, stated plainly and without a verdict: three separate price-target raises inside 48 hours on 2026-08-12/13 after the move rather than before it; a Street-high target of $700 published with the stock quoted near $503.81 on 2026-08-12; consensus $569.56 now sitting only modestly above spot where it sat far above it a month ago; and the absence of any scheduled company catalyst for roughly ten weeks. The narrative is accelerating — the 07-29 print produced new attention, the participation is visibly expanding through the 08-12/13 target revisions and the 08-14 13F disclosures, and the "FCF king 2027" framing is a fresh headline as of 2026-08-14. The mechanical stretch is a separate question from the narrative label, and it is the one that governs entry timing here.

Catalyst Calendar (next 30 days)

• ~2026-08-26 (est.) — Nvidia fiscal Q2 report. Not Microsoft's event, but the cleanest read-through on whether hyperscaler order books are still expanding; the 07-29 divergence was driven entirely by relative capex messaging.
• 2026-09-10 — $0.91 per-share dividend payable.
• ~2026-09-15 (est.) — Next quarterly dividend declaration inferred from the June-declaration cadence; date unconfirmed. Microsoft has historically paired a September board action with its annual raise, so the size of any increase is the datapoint, not the declaration itself.
• ~2026-10-28 (est.) — Q1 FY27 print. The next genuine binary: first disclosed FY27 capex against the $255–260B the pre-print previews framed, and the first Azure comparison against the +43% bar.
• 2026-12-08 — Annual shareholders meeting, virtual, 8:30am PT (outside the window, listed for completeness).

No Microsoft-specific catalyst falls inside the next 30 days. The stock is extended into an empty calendar.

Elapsed catalysts

• 2026-08-20 — Record date for the $0.91 quarterly dividend declared 2026-06-10. Mechanical, not a thesis event. (passed 6d ago)

What Would Change Our Mind

The structure that would have to break is the post-print re-rating shelf. A weekly close below $455 gives back the level the July structure required for any momentum leg and returns price into the pre-print regime; a subsequent loss of the ~$422 post-earnings gap zone would close the 07-29 gap entirely and mark the capex-discipline story as a one-quarter event.

On fundamentals, the specific datapoint that would flip the read is the FY27 capex disclosure at the Q1 FY27 print (~2026-10-28 est.): a figure in the $255–260B range that the pre-print previews framed, with the CY26 reduction confirmed as an accounting reclassification rather than a spending decision, removes the exact differentiator that separated Microsoft from Alphabet and Meta on 07-29. Azure printing below 40% against the +43% comparison would do the same from the revenue side.

On the theme, a flip to saturated would be dated by the target-raise cycle exhausting itself — a Street-high above $700 with no incremental estimate revisions behind it — combined with price failing to make a new high above $538.66 on a retest. The bull case survives an orderly pullback that holds the $455 shelf; it does not survive a fill of the 07-29 gap.

Correlation Notes

• The megacap-AI cohort has decoupled on capex messaging, not on growth. On 2026-07-29 Microsoft rose 8.13% after hours while Meta fell 9.64% and Alphabet 15% — same week, same demand environment, opposite spend framing. Correlation to GOOGL/META is now conditional on that variable rather than stable.
• Memory pricing runs inverted. DRAM/HBM strength is revenue for suppliers and a cost line for Microsoft; the April CY26 plan embedded roughly $25B of component inflation. Semiconductor rallies driven by memory pricing are not a clean positive read-through here.
• AI credit is the shared factor to watch. AMD's $5B bond sale (2026-08-14) and the WSJ report that Nvidia is downsizing a $250B OpenAI data-centre guarantee (2026-08-15) put financing conditions, not chip supply, at the centre of the complex. A widening of AI-related credit spreads would hit the whole cohort regardless of individual capex discipline.
• Neocloud prints are a second-order signal. CoreWeave's post-earnings move and the risk commentary around it (2026-08-16) speak to whether third-party GPU capacity is absorbing or competing with hyperscaler supply — relevant to the "demand exceeds supply" claim made on the 07-29 call.
• OpenAI and Anthropic economics feed back directly. OpenAI at a >$40B annualized run-rate ahead of a reported $1T IPO push and Anthropic at >$11.5B of Q2 revenue (both 2026-08-14) are simultaneously Azure consumption, competitive pressure on the Copilot layer, and — for OpenAI — an equity-method line in Microsoft's own accounts.