Microsoft (MSFT): A Quality Compounder on Sale — If You'll Underwrite the Capex #

18 June 2026 | Price: $379.40 | Position: none — watch-list buy candidate (own it via VWCE) | 52-week range: $356.28 – $555.45 | Market cap: ~$2.82T

Written as informational analysis, not personalised financial advice. It's a fact-based look at the business and an honest framing of why a wonderful company in a 30% drawdown is — for once — a genuine valuation question rather than a "too expensive to touch" one. Figures were independently verified against Yahoo Finance and Finnhub on 18 June 2026 and will move.

The set-up in one paragraph #

Here is the rare sentence about Microsoft you don't get to write often: it's cheap-ish, and it's down. The stock sits at $379, −19% year-to-date and −30% below its October-2025 all-time high of $539, hovering just above its 52-week low. It has underperformed the S&P 500 by roughly 30 points. Meanwhile the business is doing fine — revenue +15%, Azure accelerating to +33%, an AI run-rate of $13B, a 23-year dividend just raised 9.6%, and a net-cash balance sheet. So why has $700B+ of market value evaporated? One reason, and it's the whole story: capital expenditure has nearly doubled to ~$97B a year to build AI data centres, and it is shredding free cash flow. The market is no longer asking "is Microsoft a great company?" (it is). It's asking "will the AI build-out earn its money back, or is this a $100-billion-a-year bonfire of depreciating GPUs?" That single question is the entire investment case. Let's take both sides honestly.

Part 1: The Bull Case (steelmanned) #

1. For once, the valuation isn't absurd #

At $379 against trailing EPS of $16.78, Microsoft trades on 22.6× trailing / ~19.6× forward (FY27) earnings. Run Graham's intrinsic-value formula — V = EPS × (8.5 + 2g) — on trailing EPS:

The break-even growth rate is ~7%. That is the number the market is implicitly demanding of one of the best businesses on earth. Microsoft is growing revenue 15–17%, with FY27 consensus EPS growth of +15%. You do not have to believe in a miracle to clear a 7% bar — you have to believe Microsoft stays roughly half as good as it currently is.

2. The growth is real and accelerating, not fading #

• Azure grew +33% YoY (constant currency) last quarter — up from 31% the quarter before. Cloud is the opposite of a falling knife.
• Intelligent Cloud segment +21% YoY; the AI business is at a $13B annualized run-rate with 350K+ Microsoft 365 Copilot enterprise seats and 1.4M+ paid GitHub Copilot seats.
• Four straight earnings beats. FY26 revenue tracking ~$330B (+17%).

This matters for the framework: it's mean-reversion on a growing base, not a cyclical or structural decliner. The thing that broke RI.PA — buying a cheap multiple on shrinking revenue — is absent here.

3. A fortress balance sheet and a 23-year dividend #

• Net cash ~$21B ($78B cash & investments vs $57B debt).
• Dividend raised 9.6% to $3.64/yr (0.92% yield), 23 consecutive years of growth, ~21% payout — almost unlimited room to keep raising.
• AAA-equivalent credit, ~46% operating margins, ~69% gross margins. This is as high-quality as large-cap gets.

4. Graham scorecard: 4/7 — and the fails are only about price #

*The current-ratio miss is an accounting artefact: $60B+ of deferred subscription revenue sits in current liabilities. That's prepaid cash from customers, not a bill coming due — the same distortion Graham's filters create for any subscription or financial business. Microsoft passes every quality filter; it fails only the two price filters, which every great compounder does.

Part 2: The Bear Case (the capex problem is the only one that matters) #

1. Free cash flow is being incinerated by AI spend #

This is why the stock is down, and the one thing a buyer must underwrite.

In the December 2025 quarter, Microsoft generated $36B of operating cash flow and converted just $5.9B of it to free cash flow — capex ate $30B. Depreciation is exploding in tandem ($22B → $34B) as those data centres start aging on the income statement.

The result is two different valuations of the same stock:

• On earnings, it's ~22× — reasonable.
• On free cash flow, it's ~39× (FY25) and worse on a trailing basis — expensive.

That gap is the debate. Bulls say the capex is building revenue-generating capacity (Azure is capacity-constrained; AI demand is real). Bears say it's an arms race into hardware that depreciates in 4–6 years with an unproven return, and that free cash flow — not accounting earnings — is the truth. If AI monetisation keeps lagging the spend, the "cheap on earnings" thesis eventually breaks, because earnings will catch up to the depreciation. That's the value-trap risk: not bankruptcy, but capital that never earns its keep.

2. Insiders are selling, not buying (shr-002) #

Every meaningful insider trade in the past year was a sale — CEO Satya Nadella sold ~$2.5B near the all-time high in September 2025, President Brad Smith ~$483M, and several EVPs followed. There was exactly one open-market purchase: director John Stanton, $33M at $397 in February 2026 — notably on the dip, which is the one mildly bullish tell. Most of the selling is pre-scheduled 10b5-1, so it's soft-negative rather than alarming, but there's no insider conviction to lean on at these prices.

3. The overhang is real #

• Regulatory: an EU cloud-market antitrust probe (Microsoft and Amazon) opened in June 2026.
• AI pricing pressure: API price competition is squeezing the whole AI complex (OpenAI, Anthropic, Microsoft).
• Near-term estimates softening: 17 analysts cut the current quarter vs 8 raises — the de-rating may not be over.

Part 3: A portfolio note you can't ignore #

If you hold a global index fund, you already own a lot of Microsoft — it's the single largest position in VWCE (~4–5%). So as a portfolio decision, buying MSFT as a separate line just doubles down on a bet you already carry through the core. The cleanest way to "buy Microsoft cheaper" right now is simply to keep contributing to VWCE: you get MSFT at today's lower price, plus everything else, with zero single-stock risk. A direct position only makes sense if you specifically want to overweight Microsoft above its index weight — which is an active call about the capex question above, not a default.

Bottom line #

Yes — qualified. Unlike most of mega-cap tech, Microsoft at $379 is a genuine quality-at-a-discount setup: a 4/7 Graham name whose only failures are price multiples, de-rated 30% from its high, priced for ~7% growth while delivering ~15%, with a real margin of safety on any sane long-term growth assumption (Graham IV at 10% growth = $478, +26%; analyst mean target $561, +48%). The single thing you must believe to buy it is that the ~$100B/year AI capex will earn its return. If you believe that, this is one of the better risk/rewards in large-cap today. If you don't, the compressed free cash flow is quietly telling you it's a value trap. The company's quality is not in question — the AI ROI is. That's the whole decision.

For most people the non-dramatic answer is: keep DCA-ing the index (you're buying MSFT cheaper anyway), and if you want a deliberate direct overweight, start in tranches nearer the lows — a retest of ~$345–356 takes the break-even growth bar down toward 6% and widens the margin of safety. But that's a personal call about conviction in AI economics, not something to decide off one report.

Sources: Yahoo Finance and Finnhub (live quotes, financials, insider filings, analyst consensus, 18 June 2026); Microsoft Q3 FY2026 results (30 Apr 2026) and prior filings. Prices and figures as of 18 June 2026 and will move. Informational analysis, not investment advice. The author holds no direct MSFT position (indirect exposure via a global index fund).