MSFT — Microsoft
The easy part is liking Microsoft. The useful part is deciding what the AI bill is worth.
Microsoft sells a pile of enterprise products companies really do not want to rip out: Azure, Microsoft 365, security, GitHub, Windows and everything glued around them. AI gives Microsoft more things to sell through customer relationships it already owns. That is the easy part. The part I keep coming back to is whether the enormous infrastructure bill turns into enough per-share earnings to justify the price.
Microsoft is the kind of company where the bull case almost writes itself. Azure is growing like crazy, Microsoft 365 is buried in the workday, security is everywhere, and AI gives the company more stuff to sell through customer relationships it already owns.
Why I care about this one
The thesis is not “Microsoft has AI.” Plenty of companies have AI.
Microsoft has distribution. That is the whole cheat code.
Azure already hosts the workloads. Microsoft 365 already has the seats. GitHub already has the developers. Security is already buried in enterprise budgets. Microsoft does not need to invent a new customer relationship every time it launches an AI product; it gets to sell into the ones it already owns.
That is the advantage I care about. It is also why the stock is rarely sitting around waiting for somebody to notice.
FY2026 revenue reached $331.8B, up 18%, while operating income rose 21% to $155.2B. In Q4, Azure and other cloud services grew 43%, Microsoft Cloud revenue reached $59.3B, and commercial RPO finished at $678B. (FY26 Q4 results)
The FY2027 starting guide is strong too: 16%–17% total revenue growth and roughly 45% constant-currency Azure growth in Q1, with operating margins expected to be relatively flat year over year.
So no, demand is not the part I am struggling with.
The number I keep coming back to is $115.9B. That is how much Microsoft added to property and equipment in FY2026. Q4 capex alone was $41B, and management expects more than $50B in Q1 FY2027 with full-year capex still growing. (FY26 Q4 earnings call)
Azure growing 43% is fantastic.
I just want the receipt.
FY2026 Q4 + FY2027 Q1 outlook
Primary sources: FY2026 Form 10-K — SEC ↗ · FY26 Q4 earnings release ↗ · FY26 Q4 investor metrics ↗ · FY26 Q4 earnings call + FY27 outlook ↗ · Sep. 15 dividend increase ↗
The Sep. 28 review moved the earnings anchor to $17.28 of OpenAI-adjusted EPS, lowered terminal multiples to 22x / 28x / 32x and kept the position at hold / no add. The core question did not change: the AI infrastructure bill still has to show up in per-share economics.
My position
What I actually bought and sold.
| Date | Type | Account | Dollars | Shares | Price | Return now* |
|---|---|---|---|---|---|---|
| Feb 24, 2026 | Buy | Roth IRA | $5,000 | 13.035 | $384 | +34.6% |
| Mar 13, 2026 | Buy | Taxable | $2,000 | 4.995 | $400 | +28.9% |
| Mar 23, 2026 | Buy | Roth IRA | $2,465 | 6.391 | $386 | +33.8% |
| Mar 30, 2026 | Buy | Taxable | $2,500 | 6.921 | $361 | +42.9% |
| Apr 17, 2026 | Buy | Roth IRA | $1.28 | 0.003 | $427 | +21.0% |
| Jun 11, 2026 | Reinvest | Taxable | $10.84 | 0.027 | $401 | +28.6% |
| Jun 11, 2026 | Reinvest | Roth IRA | $17.68 | 0.044 | $402 | +28.5% |
| Jun 12, 2026 | Buy | Taxable | $1,800 | 4.638 | $388 | +33.0% |
| Jun 24, 2026 | Buy | Taxable | $4,924 | 13.236 | $372 | +38.8% |
| Sep 10, 2026 | Reinvest | Taxable | $27 | 0.055 | $493 | +4.6% |
| Sep 10, 2026 | Reinvest | Roth IRA | $17.72 | 0.036 | $492 | +4.9% |
| Sep 10, 2026 | Sell | Roth IRA | $5,000 | 10.199 | $490 | — |
*Return now uses the latest available quote for buys and reinvestments. Sales and fees show “—”.
Valuation & scenario math
For the operating anchor, I use Microsoft’s $17.28 of FY2026 OpenAI-adjusted EPS rather than the $17.95 GAAP figure, which included $0.67 per share of net gains tied to the OpenAI investment.
From there, the model assumes 9% / 15% / 18% EPS growth across bear / base / bull and 22x / 28x / 32x terminal P/Es. I am deliberately not asking Microsoft to keep nearly all of today’s premium five years from now.
The base case lands around $973 per share in September 2031.
Whether that is exciting or mediocre depends on the price in front of me, so I care much more about the live return to that endpoint than the return from one old reference close.
I still want the business to create the return. A generous earnings anchor and a generous terminal multiple should not quietly do the work for it.
26x / 24.5x / 23x FY2026 OpenAI-adjusted EPS
15.0% above $449 starter
| Band | Buy price | Base return to Sep 2031 |
|---|---|---|
| Starter | $449 | +16.9% |
| Normal | $423 | +18.3% |
| Aggressive | $397 | +19.8% |
The entry prices come from the $17.28 OpenAI-adjusted FY2026 EPS anchor at 26x / 24.5x / 23x. They stay fixed until the underwriting changes.
The model is not assuming Microsoft becomes mediocre. It is asking whether today's AI spending turns into per-share earnings fast enough to justify paying a premium now. I use $17.28 of OpenAI-adjusted FY2026 EPS and make growth and the terminal multiple do separate jobs instead of letting one optimistic assumption hide inside the other.
| Normalized EPS today | $17.28 |
| Annual EPS growth | 15% |
| EPS by Sep 25, 2031 | ~$35 |
| Terminal P/E | 28x |
| Implied base price | ~$973/share |
The base case is basically ~$35 of Sep 25, 2031 EPS at 28x, or about $973/share. Same math, just without making the reader decode equation soup.
| Case | What has to be true | Target range | Annualized return from today |
|---|---|---|---|
| Bear | 9% annual EPS growth · 22x terminal P/E. AI demand stays real, but growth normalizes, the infrastructure bill keeps margins under pressure and the market stops paying quite so much for the story. | $544–$626 center ~$585 | +2.6% |
| Base | 15% annual EPS growth · 28x terminal P/E. Azure, Microsoft 365, security and AI keep compounding earnings around the mid-teens while the capex wave earns a good—not magical—return. | $905–$1,041 center ~$973 | +13.7% |
| Bull | 18% annual EPS growth · 32x terminal P/E. AI becomes a durable monetization layer across Azure and the application stack, capacity gets absorbed quickly and the spending eventually produces real operating leverage. | $1,176–$1,354 center ~$1,265 | +19.8% |
Returns use the current quote and the time remaining to Sep 25, 2031. The scenario targets stay fixed until the underwriting changes.
Capital allocation
Fantastic cash engine. I still want receipts for the AI bill.
Azure is growing, new capacity is getting absorbed quickly and the balance sheet is absurdly strong. That earns a high score. The missing points are the things that still need proof: returns on the current infrastructure wave, a mixed record around gaming and strategic bets, and $12.4B of SBC making the buyback headline look much prettier than the actual share-count change.
Stage-relative, not some objective rating-agency grade. The number is there to force the judgment into the open. See the rubric →
What management is actually doing with the cash
Microsoft can absolutely afford the AI build. That is not the same thing as saying every dollar earns a great return. Azure growth, fast capacity absorption and the balance sheet deserve a lot of credit. I still take points off because the infrastructure bet is enormous, gaming and other strategic bets have been mixed, and buybacks have not translated into much diluted share reduction once SBC is included.
What I’m doing with the next dollar
Microsoft already has a meaningful place in the portfolio, and the target is 4%–7%. The live position size above is the number I care about rather than a weight frozen into this paragraph.
A better price makes the expected return better. It does not automatically make Microsoft the best use of the next dollar.
At this size I want both things at once: a valuation that is genuinely more attractive and enough sizing room that another purchase still makes sense. If something else offers the better return, I can keep owning Microsoft without forcing another buy.
What can break this
Microsoft Cloud gross margin was 65% in Q4 and 66% for FY2026, down from 69% the year before. That does not automatically mean the AI build is a bad idea. Building capacity before the revenue arrives is kind of the point.
But “demand is real” and “the incremental return on capital will be great” are two different claims.
The bear case can be painfully boring. Azure keeps growing. Copilot is useful. Microsoft earns more money every year. And yet the company spends so much to support the next leg of growth that the per-share economics end up merely good while the stock was priced for something better.
That is a much more believable bear case to me than everyone waking up one morning and deciding Word no longer exists.
The risk I care about most is not “AI was fake.” It is that AI is very real and Microsoft still earns a mediocre return on the marginal dollar spent chasing it.
Several years of capex outrunning owner-earnings growth would get my attention. So would a structural slide in cloud gross margin, meaningful Azure share loss, premium AI products that users like but do not pay enough for, or more large strategic bets that absorb capital without much per-share payoff.
On the other side, if Azure keeps outrunning supply, AI products create real ARPU, cloud margins stabilize and free cash flow starts catching the infrastructure bill, I am probably being too conservative.
A 15% stock drop by itself tells me almost none of that.
For now, I own enough that it matters and not so much that I can stop asking what the next dollar can earn.
That feels about right.
The evidence I’m watching
Azure keeps outrunning supply, premium AI products create real incremental revenue, Microsoft Cloud gross margin stabilizes and free cash flow starts catching up with the infrastructure bill. If that happens without me needing a richer terminal multiple, I am probably being too conservative.
Capex keeps outrunning owner-earnings growth for years, cloud gross margin stays structurally lower, Azure meaningfully loses share, AI products fail to create enough incremental revenue, or more giant strategic bets absorb cash without much per-share payoff.