Meta Platforms reports Q2 2026 earnings today (July 29) after raising its full-year capital expenditure outlook to $125–$145 billion for AI infrastructure, data centres, and computing capacity — while analysts expect revenue of ~$60 billion for the quarter, up ~27% year-on-year.
Meta is spending more on AI infrastructure this year than the entire GDP of a mid-sized country. That number — $125–$145 billion in capex — is extraordinary. But here is what a CFO reads before cheering or panicking about it: does this investment clear the IRR hurdle?
IRR — internal rate of return — is the answer to a simple question: what annual return does this investment generate over its life? Every company has a cost of capital (WACC — the blended rate it pays to debt and equity holders). If IRR > WACC, the investment creates value. If IRR < WACC, you're burning shareholder money with a smile.
Meta's gross margin sits at roughly 82%. That is a phenomenal base — it means every dollar of incremental AI-driven revenue drops most of the way to profit. But capex of this scale means the denominator — the investment — is enormous. To justify $145 billion, Meta needs AI to unlock revenue streams that are both large and durable. The cash flows have to compound fast enough to beat the clock.
A CFO would model three scenarios: bull (AI monetisation accelerates ad pricing and opens new product lines), base (steady ad growth, capex earns market returns), and bear (capex bloats without a revenue catalyst, FCF craters). The Q2 call today is really a Q: what is management's own IRR assumption — and do you believe them?
Every founder making a big infrastructure bet faces the same question at smaller scale. Module 16 is where you learn to answer it honestly.
📚 Learn the concept: NPV & IRR
▶ Play the 90-second CFO game All daily posts