AI Capex Faces a New Test: Can Amazon and Meta Turn Spending Into Returns?

Amazon and Meta are in the middle of an AI spending cycle that has already reshaped their capital expenditure plans, but the bigger question for investors is whether that spending will translate into sustainable returns. In a market-focused piece published by Yahoo Finance, analyst Mark Mahaney argues that there are growing indications that major internet companies could be moving from AI experimentation toward a phase where the investment supports business outcomes.

The core of the argument is that AI has become a capital-intensive foundation for everything from recommendation and search relevance to advertising targeting and cloud-based services. Mahaney’s point, as framed in the report, is that the current wave of compute and infrastructure buildouts may start to show up as incremental revenue, improved margins, or both, rather than remaining a cost center. The thesis is essentially about timing, conversion, and efficiency, not about whether AI remains strategically important.

For Amazon, that conversion question is closely tied to AWS, the company’s cloud platform business. While the Yahoo Finance piece centers on the broader pattern across large platforms, Amazon’s AI economics typically revolve around AWS demand for accelerated computing and AI-related services, and on whether those workloads translate into higher retention and greater share-of-wallet from enterprise customers. Amazon also operates retail and advertising businesses where AI can improve customer engagement and ad performance, but the immediate financial linkage tends to be harder to isolate than cloud consumption.

For Meta, the ROI pathway is often described through how AI improves feed ranking, content discovery, and ad delivery. These improvements can affect advertiser demand and user engagement, which then flow through to ad pricing and ad load. In the Yahoo Finance framing, the idea is that when AI models and systems become embedded in core product workflows, they can be monetized more directly than early-stage pilots.

The report suggests that internet behemoths may be approaching a period when AI capital spending is no longer purely speculative, but becomes associated with specific operational improvements that support earnings power. That does not mean every dollar spent will deliver the same payback rate, or that the market will agree on when the returns will be visible. Instead, the claim is that the balance of evidence is shifting toward the possibility of payoffs.

Even so, neither Amazon nor Meta typically discloses a neat, quarter-by-quarter “ROI from AI capex” number. Companies may discuss AI investment levels, infrastructure capacity, and product adoption, but they do not always provide granular attribution tying capital spending directly to future cash flows from specific models. That leaves room for uncertainty around amortization, adoption curves, and whether AI performance gains translate into measurable commercial metrics at scale.

Investors and analysts are likely to watch for indicates such as rising demand for AI compute capacity, evidence that AI-driven product improvements are lifting revenue per user or advertiser effectiveness, and indications that operating margins can stabilize or expand despite ongoing infrastructure buildouts. In particular, AWS usage trends and the rate at which AI workloads become “must-have” rather than “nice-to-have” are often treated as key indicators of whether capex is becoming a recurring revenue engine.