Big Tech’s AI spend may be understated by accounting methods, with Meta and Alphabet at the center of a new estimate

Wall Street Journal reporting, carried by Yahoo Finance, argues that how companies classify technology spending can make artificial intelligence outlays look smaller than they are. The article’s central claim is that nine major technology firms, including Alphabet and Meta Platforms, show roughly $3 trillion more in AI-related spending than commonly implied by surface-level estimates.

The gap, the report suggests, is tied to what counts as “AI spending.” Instead of limiting the accounting to clearly labeled AI development costs, the broader view incorporates infrastructure and operational expenditures that support AI systems over time, including compute capacity and the systems used to deliver models and AI features to users.

Alphabet and Meta are singled out because they sit near the core of the modern AI stack, combining large-scale data, specialized chip and server ecosystems, and consumer and enterprise platforms where AI features are increasingly embedded. In that environment, the line between “AI” and “standard technology” can blur, particularly when the same data centers and software platforms are used for both AI and non-AI workloads.

For Meta, AI has increasingly become part of how its platforms operate, from content discovery to ad delivery and integrity efforts. While the report frames the issue primarily as an accounting and classification question, it also reflects a practical reality for companies running global services: AI workloads compete for the same capacity that powers everything else, so AI-related costs may be distributed across multiple budget categories rather than tagged as a single initiative.

The Yahoo Finance post also frames the $3 trillion difference as potentially large enough to change how investors and analysts interpret the pace and scale of AI investment across the sector. If spending totals are understated, companies could appear to be investing less aggressively than they are, which could affect expectations around future capacity buildouts, model development timelines, and cost pressure.

Meta’s own public communications on AI and infrastructure often emphasize the role of applied AI in its products and the engineering work required to operate those systems at scale. Its official newsroom provides updates on research, product rollouts, and engineering infrastructure themes, but it does not, in that public feed, offer a single consolidated dollar figure that reconciles with the broader, multi-company estimate described in the Yahoo Finance report.

The article also does not provide enough detail in the excerpt available here to verify how each of the nine companies’ AI-related numbers were constructed, including which line items were included or excluded and how the methodology treated shared infrastructure and ongoing operations versus one-time build costs. Without that company-by-company breakdown in the material provided, the $3 trillion figure should be treated as an estimate of classification effects rather than a confirmed audited total for any single firm.

Going forward, the key question for the market is whether companies provide more consistent disclosure about AI-related spending categories, or whether analysts continue to adjust accounting interpretations to approximate the “true” figure. For Meta and peers, watch for changes in how they describe AI infrastructure capacity, cost drivers, and operating expense trends, since those indicates can help narrow the uncertainty around what is truly AI-specific versus broadly technological spending.