Meta Platforms and the Cost of Defending the Moat
What the business is
Meta Platforms owns a fine toll road and is spending heavily to ensure that tomorrow’s traffic still uses it. The toll road remains attractive. The construction budget requires more imagination.
Facebook, Instagram, WhatsApp and Messenger gather consumer attention, which Meta sells to advertisers through targeted placement and campaign tools. Advertising represents about 98% of revenue. In the first quarter of 2026, it contributed $55.0 billion of the company’s $56.3 billion in total revenue.
The moat begins with network effects. Users attract other users, and the resulting audience attracts advertisers. Advertisers, in turn, finance continued product development and distribution. Meta also has accumulated social graphs, behavioural data and campaign infrastructure. Switching is possible, but an advertiser must rebuild performance histories, pixels and operating routines elsewhere. Friction need not be dramatic to be economically useful.
Recent operating evidence supports the moat. First-quarter advertising impressions rose 19%, while the average price per advertisement increased 12%. Selling more inventory at a higher price is generally preferable to explaining why neither happened. AI recommendation systems can strengthen this engine by improving content selection, engagement and targeting. Adoption of Meta’s AI advertising tools has doubled, potentially making sophisticated campaign creation easier for smaller advertisers.
The strategic difficulty is that AI may strengthen the franchise while making it far more capital-intensive. Assistants and answer engines could divert attention from feeds, and generated content could weaken feed quality. Reality Labs adds a separate record of large losses without demonstrated recovery of the capital committed. I therefore regard the advertising franchise as understandable, but the return on the infrastructure now being built as unproven.
What the numbers say
Fiscal 2025 revenue was $200.97 billion, operating income was $83.28 billion and the operating margin was 41%. The five-year average return on equity is 27.5%, while debt-to-equity is 0.27. Those figures describe a highly profitable core business with a manageable historical balance-sheet profile.
Capital requirements are the less agreeable part of the account. Fiscal 2025 capital expenditure was $69.7 billion, and 2026 guidance calls for $125–145 billion. The owner-earnings calculation used here mechanically treats all capital expenditure as maintenance spending. That may understate earning power if the infrastructure produces durable growth, but the evidence does not establish how much spending is genuinely optional. I would rather leave uncertainty visible than improve the answer by adjusting the question.
Capital allocation has also shifted. Meta previously repurchased shares during weakness and introduced a dividend in 2024. More recently, repurchases stopped while debt issuance and possible equity financing were considered to support AI expenditure. The dual-class structure means outside shareholders have little practical ability to restrain these commitments. Management may be right, but permission is not the scarce commodity.
- Market cap (price-adjusted, 2026-09-15 close): $1663.49 billion
- Owner-earnings yield (price-adjusted): 0.56%
- Share price (2026-09-15 close): $670.24
- 10-year Treasury: 5.00%
A 0.56% owner-earnings yield against a 5.00% Treasury leaves a spread of -4.44 percentage points. The supplied parsed intrinsic-value range is $60.0–70.0 billion. The price-adjusted market capitalization carries a 2672% premium to the low end and a 2276% premium to the high end. A superior business can survive many things, but valuation arithmetic has an inconvenient habit of surviving too.
What Wall Street believes vs what I see
The available headlines lean optimistic, though not without nerves. One argues for a substantially higher share price, another says AI spending is beginning to pay off, and a third says AI is working while emphasizing the size of the bill. The shared belief is that infrastructure spending will eventually produce better advertising economics or meaningful new revenue streams.
I see a plausible thesis burdened by a demanding starting valuation. The advertising operation continues to show higher impressions and higher prices, while AI tools may improve recommendation quality and campaign performance. Those are meaningful signs of franchise strength. They are not yet proof that the present investment program will earn an adequate return on all the capital absorbed.
The mechanical owner-earnings calculation may be severe because it treats every dollar of capital expenditure as necessary to maintain the business. Reality could be kinder if a substantial portion creates incremental revenue and durable pricing power. Yet a 0.56% owner-earnings yield against a 5.00% Treasury requires more than a mildly favourable reinterpretation. It requires the distinction between growth capital and maintenance capital to matter a great deal.
That is the central contest. If AI infrastructure creates stronger engagement, new revenue and lasting advertising advantages, current cash economics may obscure future earning power. If comparable spending becomes necessary merely to preserve Meta’s position, it is maintenance capital in economic clothing. The label on the invoice does not pay it.
Verdict
Verdict: pass.
The 0.56% owner-earnings yield is below the 5.00% Treasury, so the shares fail the rubric’s watch threshold and fall well short of the additional 2 percentage points required for circle. The price-adjusted market capitalization is also nowhere near 70% or less of the supplied intrinsic range’s $60.0 billion low end. I would change the verdict if verified owner earnings raised the yield above the Treasury and the valuation moved into the rubric’s required margin-of-safety range.
This analysis is AI-generated, educational, and not investment advice. Figures may contain errors or be delayed. Disclaimer