Archived thesis
Meta Platforms
Buying operating earnings at a discount created by uncertainty over how the compute build gets monetized.
At a glance
My views, condensed from the full thesis below.
- Investment case
- Compute is becoming important enough that the infrastructure proves valuable over time, and Muse can turn the spending into a second revenue line.
- What the market may be missing
- Investors are treating the capex as a cost rather than an investment, and the stock has been punished partly because management has not said how the compute buildout will be monetized.
- Business economics
- Advertising turns an enormous user base into high-margin operating earnings, and AI-driven ranking and targeting has been lifting the price per impression.
- Catalysts and timing
- Operating income growing again in 2027.
Principal risks
Monetization stays vague while depreciation accumulates, so margins compress before any benefit appears. Meta, Alphabet and Amazon share the same AI capex cycle, so they are not three independent bets.
What would prove the thesis wrong
Operating income still falling through 2027, or no path to paid revenue for Muse by then.
Investment case
Scale, operating earnings and the advertising engine
Meta can continue increasing its scale and competitive position. The advertising business converts an enormous user base into operating earnings at high margin, and that engine is what makes the rest of the story affordable.
Advertising revenue decomposes into impressions delivered and price per impression. AI-driven ranking and targeting has been lifting the second of those, which is a measurable mechanism rather than a narrative.
The compute buildout and the monetization question
The market has punished the stock partly because management has been unclear about exactly how its large compute buildout will be monetized. That criticism is fair. The capital being committed is large and the return path has not been specified.
The view taken here is that compute is becoming such an important resource that the infrastructure will prove valuable over time, even where the specific product route is currently undefined. The risk in holding that view is obvious. It argues from scarcity rather than from a demonstrated revenue line, and if monetization stays vague while depreciation accumulates, margins compress before any benefit appears.
Valuation and sizing discipline
The valuation has been attractive relative to the quality of the business, which is the entry argument.
Sizing stays controlled regardless, because the portfolio already has substantial mega-cap technology and compute exposure. Meta, Alphabet and Amazon share sensitivity to the same AI capex cycle, so treating them as three independent positions would understate the concentration actually being run.
Why it runs above the 11.5% ceiling
Meta is my largest holding at about 12%, above the 11.5% soft ceiling, and I have re-underwritten it there deliberately. It has the fastest revenue growth of the three hyperscalers and the lowest bar in price terms, which tells me investors are treating its capex as a cost rather than an investment. I think Muse turns that spending into a second revenue line, and that Meta's operating income grows again in 2027.
If operating income is still falling through 2027, or Muse has shown no path to paid revenue by then, I am wrong and the 12% weight is too large.
Business economics
Reporting period through Q2 2026
Figures as filed with the SEC, and margins and growth rates calculated from them. Each chart has a data table.
Operating income
Reported
Quarterly, US$ billions, Q3 2024 to Q2 2026
Filed as operating income (loss). Fourth quarters are derived as the fiscal-year total less the other three quarters, since filings report no standalone Q4 figure.
Financial data through Q2 2026. Source: SEC filings (opens in a new tab)
Data table: Operating income
| Period | Operating income |
|---|---|
| Q3 2024 | $17.4B |
| Q4 2024 | $23.4B |
| Q1 2025 | $17.6B |
| Q2 2025 | $20.4B |
| Q3 2025 | $20.5B |
| Q4 2025 | $24.7B |
| Q1 2026 | $22.9B |
| Q2 2026 | $18.8B |
Operating margin
Calculated from reported figures
Quarterly, percent, Q3 2024 to Q2 2026
Calculated as operating income divided by revenue, both as filed.
Financial data through Q2 2026. Source: SEC filings (opens in a new tab)
Data table: Operating margin
| Period | Operating margin |
|---|---|
| Q3 2024 | 42.7% |
| Q4 2024 | 48.3% |
| Q1 2025 | 41.5% |
| Q2 2025 | 43.0% |
| Q3 2025 | 40.1% |
| Q4 2025 | 41.3% |
| Q1 2026 | 40.6% |
| Q2 2026 | 30.9% |
Revenue
Reported
Quarterly, US$ billions, Q3 2024 to Q2 2026
Filed as revenue from contract with customer excluding assessed tax. Fourth quarters are derived as the fiscal-year total less the other three quarters, since filings report no standalone Q4 figure.
Financial data through Q2 2026. Source: SEC filings (opens in a new tab)
Data table: Revenue
| Period | Revenue |
|---|---|
| Q3 2024 | $40.6B |
| Q4 2024 | $48.4B |
| Q1 2025 | $42.3B |
| Q2 2025 | $47.5B |
| Q3 2025 | $51.2B |
| Q4 2025 | $59.9B |
| Q1 2026 | $56.3B |
| Q2 2026 | $60.8B |
Diluted EPS
Reported
Quarterly, US$ per share, Q3 2024 to Q2 2026
Filed as earnings per share diluted. Fourth-quarter EPS is not shown: it is reported only within annual results, and annual EPS less three quarters is not a valid per-share figure because the share count changes.
Financial data through Q2 2026. Source: SEC filings (opens in a new tab)
Data table: Diluted EPS
| Period | Diluted EPS |
|---|---|
| Q3 2024 | $6.03 |
| Q4 2024 | Not shown |
| Q1 2025 | $6.43 |
| Q2 2025 | $7.14 |
| Q3 2025 | $1.05 |
| Q4 2025 | Not shown |
| Q1 2026 | $10.44 |
| Q2 2026 | $6.18 |
Valuation
Attractive relative to the quality of the business, with the lowest bar in price terms of the three hyperscalers.
No valuation inputs, such as forward multiples or peer comparisons, are recorded for this thesis yet, so the valuation view is qualitative.
Risks and thesis invalidation
Principal risks. Monetization stays vague while depreciation accumulates, so margins compress before any benefit appears. Meta, Alphabet and Amazon share the same AI capex cycle, so they are not three independent bets.
What would prove it wrong. Operating income still falling through 2027, or no path to paid revenue for Muse by then.
The thesis argues the case against itself in The compute buildout and the monetization question, Valuation and sizing discipline and Why it runs above the 11.5% ceiling.
Position sizing
About 12%, above the 11.5% soft ceiling, re-underwritten deliberately because it pairs the fastest revenue growth of the three hyperscalers with the lowest bar in price terms.
The sizing discussion in the sizing summary was written when the position was about 12%. In the holdings snapshot of October 16, 2026 it is 15.96%. The text has not been revised for the new weight.
Related decisions
AddWeight change unavailable
Added after Muse and the Muse Wonder models launched. Meta now ships frontier models to 3.6 billion daily users, and I think the market still prices it as an ad business. I'm wrong if Muse shows no paid subscriber or transaction revenue within two quarters.
In the decision log
Sources and revisions
- Company filings: SEC EDGAR (opens in a new tab), through Q2 2026
- Holdings snapshot: October 16, 2026
- The analysis and views are Isaac Toffel's own.
Text revisions: Oct 16, 2026 · Sep 27, 2026 · Aug 20, 2026.