Meta Earnings: Strong Ad Revenue Momentum, but an Avalanche of AI Costs Is Coming

Key Morningstar Metrics for Meta Platforms

• Fair Value Estimate: $850
• Morningstar Rating: ★★★
• Morningstar Economic Moat Rating: Wide
• Morningstar Uncertainty Rating: High

What We Thought of Meta Platforms’ Earnings

Meta reported solid third-quarter earnings, with sales up 26% to $51 billion and operating margins contracting 300 basis points to 40%, as artificial intelligence-related costs rose year over year. The firm indicates that its capital expenditures, mostly on AI, will be over $100 billion next year.

Why it matters: Meta’s ad business continues to perform exceptionally well, with management attributing improvements in engagement and monetization to the firm’s increased use of AI, as well as continued ad inventory (Threads, WhatsApp) growth, and strong engagement (Facebook, Instagram).

• Meta has two primary levers of driving ad sales: volume and price. On volume, the firm can increase its ad inventory, that is, the number of ads shown, by adding more surfaces and by increasing its user count. To this end, we saw an impressive 8% growth in daily users to 3.54 billion.
• On price, we see Meta’s AI investments as key to driving higher value to advertisers via improved ad targeting and higher returns on ad spending, which then drives higher ad prices. This dynamic was evident in the 10% increase in ad prices, mostly driven by improved ad performance.

The bottom line: We maintain our $850 fair value estimate for wide-moat Meta, with the firm’s strong performance offset by the higher-than-expected capital and operating expenditures commentary for 2026. With shares trading down after hours, we continue to view them as undervalued.

• We understand investor fears around investments in AI. After a scarring Reality Labs misallocation of capital, investors are again doubting Meta’s capital allocation strategy for AI. We are not as pessimistic.
• From a strategic perspective, we think it is important that Meta invests in its own foundation models. Simply put, the costs of running someone else’s model at Meta’s scale are not economically feasible and would cause material margin compression.