Senate Finance Democrat proposes excise tax on new AI-focused data centers

A draft white paper from Sen. Ron Wyden, D-Ore., ranking member of the Senate Finance Committee, proposes removing existing tax incentives for data center construction and introducing a new excise tax on data centers whose workloads primarily handle AI compute. The framework aims to shift incentives amid a national boom in data center development, with the stated goal of offsetting consumer burdens such as higher energy prices, workforce disruption, and broader resource strain. In a press release, Wyden said American communities are questioning whether the rapid buildout of data centers will benefit them as local disruptions rise and long-term career prospects for millions of workers are affected.

The proposal targets two categories of data centers: those owned, operated, or rented by massive data center operators, and those used by major data center payors, including hyperscalers Amazon and Meta. The excise tax would function as a gross receipts tax levied on total revenue or sales of a business. The draft paper specifies that the tax would apply to the largest actors driving the data center boom and most able to pay, with selective exemptions for assets constructed before 2024.

Wyden’s framework includes a provision for space-based data centers. If a space-based data center attempts to avoid the tax by being labeled as a non-U.S. asset, a withholding tax would apply to payments made by U.S. taxpayers to use the facility. This addresses proposals from space-focused firms, including Blue Origin and SpaceX, to build millions of orbital data centers.

The draft white paper is open for public feedback, with the Senate Finance Committee’s minority staff accepting comments until Aug. 31. Discussion draft language is expected to arrive in the coming fall. The proposal follows heightened political attention to the impacts of data center construction, including a pledge by several major tech companies to absorb utility costs for data centers handling their AI workloads.