Amazon's $220B data center bet wins investor faith as cloud revenue soars

The market's response to Amazon versus Meta illustrates a critical divide in how investors evaluate AI infrastructure spending. For Amazon, the calculus is straightforward: AWS is generating substantial and accelerating revenue (up 37% year over year), which provides concrete justification for the $173 billion fiscal-year capex and the raised $220 billion 2026 forecast. Jassy's statement that AWS and Amazon Bedrock can succeed without owning a frontier AI model—because "there's not going to be a single model to rule them all"—signals that Amazon's revenue engine is not dependent on proprietary model dominance but on providing compute capacity that multiple models can run on. This positions the company as cloud infrastructure provider rather than pure AI play, a distinction investors appear to value.

Meta's contrasting experience reveals that capex alone is not sufficient to win investor confidence. Despite comparable infrastructure ambitions, Meta's stock fell 8% because investors see spending without an offsetting revenue stream. The $7.6 billion decline in Amazon's cash reserves over the past 12 months and its first period of negative free cash flow in 2024 could have triggered similar skepticism, but AWS growth and margin trajectory expectations have insulated the company. The data-center-to-revenue lag—which Jassy frames as a design feature of AWS's model—actually reinforces investor confidence, because present spending on capacity can be monetized years later as demand materializes. This temporal mismatch works in Amazon's favor precisely because its cloud business has a track record of converting infrastructure into revenue, whereas Meta's capex spending remains speculative from an investor perspective.