Artificial intelligence has moved from a market narrative to a measurable economic force, and the good news for entrepreneurs and investors is that you do not need proprietary data to track it. A 2026 note from staff at the Federal Reserve Board identifies publicly available series—corporate capital expenditures, national equipment investment data, and aggregate earnings measures—that together reveal how much of the "AI buildout" is real spending rather than sentiment. Fidelity analysts frame the same question as a bubble test: five directional indicators, from earnings growth to the interest-rate cycle, that show whether AI-driven valuations rest on genuine productivity gains.
Morgan Stanley's research institute goes further, calling AI "an industrial buildout" with nearly $3 trillion in infrastructure spending still ahead as of its March 2026 analysis. This article assembles those threads into a monitoring framework you can update from public filings and government statistics.
Tracking the AI Buildout: Infrastructure and Investment
The first category to monitor is the physical buildout itself. The Federal Reserve note measures capital expenditure as total purchases of "property and equipment" as reported on companies' 10-Q filings, tracking the major technology firms—including Oracle and Coreweave, the latter of which went public in the first quarter of 2025—through U.S. Securities and Exchange Commission data. This gives a direct, quarterly read on what hyperscalers are actually spending on AI-related hardware and data centers.
The scale is substantial. Morgan Stanley Research estimates $2.9 trillion in global capital expenditure on data centers between 2025 and 2028, of which roughly $1.4 trillion would be covered by hyperscaler cash flows, with the remainder drawn from roughly $200 billion in corporate debt issuance, $150 billion in securitized credit issuance, about $800 billion in potential private credit via asset-based finance and joint-venture debt funding, and roughly $350 billion in other capital, primarily private equity, venture capital, and sovereign funds. For an investor, the composition of that financing stack matters as much as its size: who is funding the buildout, and with what kind of money.
Two caveats from the Fed note deserve attention. First, because hyperscalers increasingly lease data center capacity rather than own it, headline capex figures could progressively understate total AI-related infrastructure investment. Second, broader national investment data—such as Bureau of Economic Analysis figures on equipment investment—includes substantial non-AI activity, so any AI-specific component must be approximated, for instance by extrapolating a pre-2023 baseline, and that approximation may become less reliable over time. The Fed's own suggestion is to triangulate across these investment indicators rather than rely on any single series.
One structural observation frames the whole category: as of 2026, the Fed staff note that financial markets have been highly responsive to changes in the AI narrative, while aggregate output and labor market data show limited signs of broad-based transformation. The economy is reorganizing around the technology, but the real effects are concentrated in certain areas. Watching whether that gap narrows is itself an indicator.




