Morgan Stanley lifted its capital expenditure estimates for both Amazon and Meta, reflecting Wall Street's recognition that AI infrastructure investment is running hotter than previously modeled. For Amazon, which generated $716.9B in revenue (+12.4% YoY) at a 10.8% net margin, higher CapEx directly pressures free cash flow given AWS's role as the primary AI buildout vehicle. Meta's situation is similar but more acute: at $201.0B in revenue (+22.2% YoY) and a robust 30.1% net margin, the company has more cushion to absorb spend, yet its AI outlays are growing rapidly enough to warrant a formal estimate revision.

The revision matters because sell-side CapEx estimate increases tend to trickle into lower near-term free cash flow targets and can weigh on multiples that are already pricing in significant AI monetization. Both AMZN and META trade at elevated valuations relative to historical norms, so any signal that the payback period on AI infrastructure is lengthening is a genuine risk to sentiment.

The bull case rests on the idea that massive CapEx is itself a moat signal — only a handful of companies can fund this level of infrastructure, and those that do tend to capture disproportionate AI revenue as enterprise adoption scales. META's 30%+ net margin gives it unusual capacity to self-fund the buildout without dilution. AMZN's AWS still commands premium cloud pricing and has a direct monetization path for every dollar of AI infrastructure it deploys.

The bear case is straightforward: CapEx that rises faster than revenue monetization compresses free cash flow yields, and both stocks already embed optimistic assumptions about AI revenue ramp timelines. If the buildout cycle extends another 12-18 months before meaningful AI revenue inflects, current multiples leave little margin of safety. Watch for any guidance commentary on CapEx intensity versus revenue growth at the next earnings prints as the key catalyst to resolve the tension.