Google and Amazon earnings reveal AI capex as strategic necessity

The more consequential signal buried in this earnings cycle is not the raw capex numbers but the rhetorical shift: hyperscalers are now framing AI infrastructure spending as non-discretionary, which changes how analysts should model their cost structures going forward. That framing is doing real work to insulate these companies from shareholder pressure during quarters where AI revenue hasn't yet matched the investment.

This connects directly to the Meta earnings piece from Stratechery on August 3rd, which showed what happens when that framing breaks down. Meta's miss demonstrated that investors will tolerate large AI commitments only as long as the narrative holds, and Google and Amazon are watching that cautionary example closely. Meanwhile, the Alibaba Qwen3.8-Max coverage from August 3rd adds pressure from outside the US that makes the 'we must spend or fall behind' argument easier to sustain internally. The open-weight letter covered by Simon Willison on August 2nd, signed by Amazon among others, further complicates the picture: Amazon is simultaneously lobbying for open model distribution while treating proprietary compute capacity as a moat.

Watch whether Amazon breaks out Anthropic-attributed revenue as a distinct line item in the next two earnings calls. If it does, that signals the investment thesis has matured enough to defend publicly; if it stays bundled inside AWS, the capex justification remains largely faith-based.