Every large language model query, every AI image generated, every autonomous agent running in the cloud draws power from a physical data center plugged into a physical grid. The explosion in AI compute since 2023 has turned data center electricity demand from a footnote in utility earnings calls into the single most consequential driver of U.S. power infrastructure investment in a generation. Federal policy — permitting rules, grid interconnection queues, Defense Department cloud contracts, export controls on chips — shapes who builds, who powers, and who profits from this buildout.
The mechanism is straightforward: hyperscalers (Microsoft, Alphabet, Amazon, Meta) have publicly committed to spending hundreds of billions of dollars on data center capacity. That spending flows downstream into power generation, high-voltage transmission equipment, cooling systems, specialty construction, and the real estate investment trusts that own the buildings. Each of those downstream categories has multiple publicly traded proxies. Washington accelerates or impedes this flow through permitting speed, grid interconnection policy, nuclear licensing, and the pace of federal cloud procurement.
This playbook maps the full chain — from the federal trigger to the public market beneficiary — so you can track it in real time and understand which tickers sit closest to the actual money flow.