META vs MSFT — Post-Print Teardown

As of 30 Jul 2026, 09:15 ET (both reported last night, 29 Jul, post-close)

1. The scoreboard

Metric	META Q2'26	MSFT Q4 FY26
EPS actual vs est	$6.18 vs $7.18 → −13.9% miss	$4.74 vs $4.23 → +12.1% beat
Revenue	$60.8B vs $60.2B (+28.0% YoY)	$90.0B vs $87.6B (+17.8% YoY)
Operating margin	30.9% (vs 43.0% LY)	45.1% (vs 44.9% LY)
Next-Q guide vs consensus	$61.0–64.0B vs $62.68B → midpoint below	$89.85–90.95B vs $89.63B → above
Stock reaction (premkt)	−8.9%	+9.5%

---

2. Your capex thesis is only half right — and that half is the wrong half

You asked whether the divergence was capital spending. No. If capex intensity were the driver, MSFT should have been punished harder in absolute dollars — it spent $35.8B of capex in the quarter vs Meta's $30.1B. That's the bias in your framing: the market in July 2026 is not penalizing capex level, it is penalizing capex without a contracted revenue offset.

Look at the intensity ratios:

Quarter	META capex/rev	MSFT capex/rev	META capex/OCF	MSFT capex/OCF
Q2'25 (LY)	34.8%	22.3%	64.7%	40.0%
Q2'26 (now)	49.5%	39.8%	94.5%	64.6%

Meta is the more intense spender on both measures, and it is converting almost nothing to cash: Meta FCF collapsed to $1.7B (2.9% FCF margin) from $13.2B last quarter and $9.0B a year ago. Microsoft still printed $19.6B FCF (21.8% margin) in its heaviest capex quarter ever.

Capex intensity: Meta's step-function to ~50% of revenue in one quarter is what broke the model, not the dollar amount.

The actual differentiator is proof of monetization. Microsoft handed the market receipts:

• Azure +43% growth; Azure crossed $100B annual revenue for the first time
• Cloud revenue $214B full-year, with ~90% from customers outside frontier-model labs
• Contract backlog +25% ex-OpenAI; commercial RPO was $627B as of last quarter
• Microsoft 365 Copilot at 30M+ paid seats, Copilot revenue +60% QoQ
• 88 new data centers added this year, "customer demand continues to exceed available capacity"
• CFO Amy Hood explicitly framed the spend as reversible: "you just slow down" on GPUs if demand cools

Meta handed the market a promise:

• CEO: "it would be foolish to basically just sell all of the compute and take a short-term profit"
• CFO: "we are demand constrained... near-term infrastructure capacity is more valuable than long-term capacity"
• Monetization plan described as a future "business in a box" with subscriptions + volume pricing, and enterprise agents — i.e., pre-revenue
• Raised the low end of the 2026 capex range while missing EPS by 14%

That is the whole trade. Microsoft's incremental dollar is landing on a $600B+ backlog. Meta's incremental dollar is landing on a bet on superintelligence plus a rumored ~$10B Anthropic compute lease that isn't in the numbers yet.

---

3. The balance sheet divergence nobody is talking about loudly enough

This is the part I'd want on your risk sheet:

	META	MSFT
Total debt Q2'25 → Q2'26	$49.6B → $112.3B	$60.6B → $56.8B
Net debt / (net cash)	+$22.1B net debt	−$20.0B net cash
Debt added in last 2 quarters	+$25.5B (Q1→Q2 alone)	−$0.8B

Meta went from a $20.7B net cash position in Q1'25 to $22.1B net debt today — a ~$43B swing in five quarters. On top of that there is off-balance-sheet structure: BlackRock led a $12.3B bond sale for Meta's El Paso data center, with Meta investing $10B+ into a ~1GW venture. Microsoft has similar structures (Goldman pitching $5.4B for a Microsoft-tied DC), but it is doing it from a net-cash position with 45% operating margins.

Meta's FCF went to near-zero this quarter; Microsoft absorbed record capex and still generated ~$20B.

Earnings-quality red flag on the MSFT side — do not skip this. On the call the CFO disclosed that effective FY27, Microsoft is extending the estimated useful life of data centers and office buildings from 15 to 25 years. Mechanically that lowers depreciation and flatters future EPS. Management said the impact on FY27 operating income is "minimal," but this is exactly the kind of change that becomes material in FY28–29 and that you should model separately rather than take at face value. If you're buying MSFT here, buy it on Azure growth and backlog, not on reported EPS.

---

4. Valuation — Meta is now genuinely cheap, and that's the bear trap

At current premarket prices:

Metric	META	MSFT	Tech sector median
P/E (TTM, at premkt price)	18.0x	24.7x	—
P/E (TTM, quarter-end basis)	21.2x	20.8x	0.5x*
EV/EBITDA	14.0x	14.1x	5.4x
P/Sales (TTM)	6.29x	8.35x	2.78x
P/FCF (TTM)	35.0x	41.4x	9.6x
ROIC (quarterly)	5.8%	6.7%	~0.0%

*Sector median P/E is distorted by loss-making constituents — treat it as noise, not a benchmark.

Meta trades at a 27% discount to Microsoft on trailing earnings while growing revenue 10 points faster (28% vs 18%). That is the entire bull case, and it's the one Ackman is running ("META and AMZN are cheap stocks"), plus Ark and Rothschild ($1,000 PT).

Here's the problem with that discount: Meta's denominator is falling. EPS went from +13.7% YoY growth in Q1'26 to −13.5% in Q2'26. The 18x is on a TTM number that includes two strong quarters that will roll off. If you annualize the Q2 run-rate ($6.18 × 4 = $24.72), Meta is at 21.6x on current earnings power — and that's before Q3, where guidance midpoint ($62.5B) is below the $62.68B consensus with capex still climbing. Anchoring on trailing P/E for a company whose margin just fell 1,220bps is the single most common way to lose money in this setup.

---

5. Sell-side reaction — the vote was unanimous in direction

META — every single post-print action was a price-target cut:

Firm	Rating	Old → New PT
Rosenblatt	Buy	$1,015 → $883
BofA	Buy	$835 → $810
Barclays	Overweight	$830 → $780
KeyBanc	Overweight	$790 → $780
Cantor	Overweight	$770 → $680

MSFT — mixed, skewed up: Piper $540 → $550; Cantor $502 → $522; Barclays $545 → $512. Standing bulls: Bernstein $646, Guggenheim $586, Truist $575.

Note the asymmetry in ratings vs targets: nobody downgraded Meta, they just cut targets by 4–13%. Sell-side is still Buy-rated but is marking down its own capex/margin assumptions. Wedbush already went to Neutral, $671 on 16 Jul, pre-print, explicitly on the "monetization gap." That was the right call.

---

6. Technical / positioning setup

Trailing 12 months indexed. META −28.9%, MSFT −18.4%. Both have de-rated hard; today's gap opens the first real divergence.

	META	MSFT
Last close / premkt	585.61 / 533.45	390.54 / 427.65
50-day SMA	604.55 (below)	397.91 (premkt above)
200-day SMA	636.11 (−16% below)	434.13 (premkt −1.5%, testing)
12-mo change	−28.9%	−18.4%

Structurally: both broke their 200-day, and press coverage flagged the Mag-7 basket breaking its 200-DMA for the 4th time since 2022. But today MSFT gaps back to kiss its 200-day from below — a level that either becomes support on a reclaim or a rejection. META gaps down to ~$533, which is below its March 2026 low-zone close ($525.72 week of 23 Mar) — it is making new 12-month lows on a fundamental miss, which is a materially worse tape.

One more positioning note: MSFT short interest hit ~92M shares, the highest since May 2015 going into the print. A 9.5% gap on peak short interest is partly a squeeze — do not extrapolate the whole move as fundamental re-rating.

---

7. Risk registers (from the 10-Ks, filed 29 Jul 2026 for MSFT / Jan 2026 for META)

META — the sharper risk profile:

• Single-revenue-stream concentration: "substantially all of our revenue is currently generated from marketers advertising on Facebook and Instagram." Everything else is a cost center.
• Reality Labs: reduced 2025 operating profit by ~$19.19B, and Meta guides 2026 losses "to remain similar." That's a ~$19B/yr structural drag before AI capex.
• AI explicitly flagged as unproven: "our AI strategy and investments may not be successful in the foreseeable future, or at all."
• Litigation calendar: "several bellwether trials in our youth-related litigation matters are scheduled for 2026 and beyond." A Tennessee Instagram-addiction trial is live right now; FTC is suing Hims & Hers over data sent to Meta; a novel AI-layoff discrimination case is in arbitration.
• EU regulatory: European Commission ruled the "subscription for no ads" model non-compliant; Meta warns further forced modifications could cause "a significant impact to our European business and revenue."
• Signal loss: ATT/GDPR/DMA-driven erosion of ad targeting signals is a permanent, recurring headwind.
• Dual-class structure — you have effectively zero governance recourse on the capex decision. Zuckerberg controls the vote. That matters enormously when management says "my personal bet is that the people who invest in this are going to be rewarded."

MSFT — the more diversified but not risk-free profile:

• Capex risk stated directly: investments are "in advance of fully developed revenue streams" and "may not be realized in the expected timeframes or at expected levels"; overestimation "may lead to impairment of assets on our balance sheet."
• Power/land/permitting as a hard physical constraint, plus "community opposition, state and local moratoriums."
• OpenAI relationship dependency — though the renegotiated framework (royalty-free IP through GPT-32-equivalent, rev-share through 2030, elimination of their rev-share to OpenAI) is a net positive vs. the prior deal.
• $28.9B IRS transfer-pricing NOPA still outstanding for tax years 2004–2013, plus penalties and interest. That is a real, quantified tail liability.
• New: UK CMA consumer-protection investigation opened into Microsoft Ireland Operations (29 Jul).
• Consumer segment is shrinking — Windows OEM guided down high-teens, Xbox content down low-teens.

---

8. The call: I'd own MSFT for the next 12 months

If you force me to one name for a 1-year holding period, I take Microsoft. Reasoning, in order of weight:

1. Direction of estimate revisions. MSFT just guided Q1 above consensus and guided FY27 to "another year of double-digit revenue and operating income growth." META guided Q3 midpoint below consensus with capex going up. Over a 12-month horizon, the single highest-correlation variable with equity return is the direction of forward EPS revisions, and those two arrows point opposite ways right now. Every one of five post-print Meta notes cut its target.

2. Margin defense is proven vs. unproven. MSFT held 45.1% operating margin through a quarter where capex hit 39.8% of revenue. META's fell to 30.9% at 49.5% intensity. Microsoft has demonstrated it can absorb the AI build inside its margin structure; Meta has demonstrated it cannot — yet.

3. Contracted revenue backing the spend. $627B RPO, backlog +25% ex-OpenAI, Azure +43% on a $100B+ base, 30M Copilot seats. Meta's offsetting revenue is a rumored Anthropic lease and a "business in a box" that doesn't exist yet.

4. Balance sheet optionality. $20B net cash vs $22B net debt with $25.5B of debt added in one quarter. In a tape where "fund managers now fear AI capex more than Iran, tariffs, or a recession" and CDS spreads on hyperscalers have doubled, the levered spender is the one that gets repriced first if credit tightens.

5. Reversibility. Hood said out loud that the GPU portion of spend can be throttled — two-thirds of MSFT capex is short-lived assets that track revenue. Zuckerberg said the opposite: selling compute for short-term profit would be "foolish." One is an option, the other is a commitment.

Where I could be wrong, and it's not a small risk. Meta's asymmetry is real: 18.0x trailing / ~21.6x on Q2 run-rate against 28% revenue growth, at new 12-month lows, with sell-side still Buy-rated and $786 average targets (+47% implied). If Q2'26 proves to be the margin trough — capex intensity plateaus, the Anthropic-style compute-leasing deals get signed and disclosed, and ad revenue keeps compounding at 25%+ — Meta re-rates violently and beats MSFT by 20+ points. Zuckerberg's own framing on the call ("we see the technology working, we're happy with the trajectory of the lab") is the tell that he will not blink, which is bad for 2026 EPS and potentially great for 2028.

How I'd actually express it if I ran your book: overweight MSFT as the core, hold a smaller, later-sized position in META and let it prove the margin trough on the Q3 print (late Oct) before adding. You are not paid to catch the exact bottom on a name whose CEO has unilateral voting control and just told you he's spending through your P&L. And if you want the pair on: long MSFT / short META is the clean expression of "capex with a backlog beats capex with a thesis" — but recognize you're short a 18x-earnings, 28%-growth asset, which is a painful borrow if sentiment flips.

Not trading advice. I can't make this decision for you, and I have no ability to backtest whether this framework has historically produced returns — that would need to go through the Strategy Tester.

---

Want me to go deeper on any leg of this?

Pull the full OPRA options chain for both names to show you exactly what implied vol is pricing for the next 30/90 days and where the max-pain strikes sit post-gap, compare META and MSFT against the rest of the hyperscaler cohort (GOOGL, AMZN, ORCL) on capex/OCF and FCF margin so you can rank the whole AI-capex complex, or set up alerts on MSFT reclaiming its 200-day and on META losing its March 2026 low, so you get notified rather than watching the tape.