US lawmakers are considering two bills that would strip AI‑focused data centers of the ability to claim full expensing for equipment and structures, a tax break that currently lets firms deduct the entire cost of qualifying investments in the year incurred.
Warner’s Data Center Tax Accountability and Disclosure Act (July 2026)
- Scope – Disallows full expensing under § 168(k) for any property used in an “AI data center,” defined as a facility dedicated to IT and telecommunications equipment that is at least 20 % used for developing or operating AI.
- LEED exemption – Facilities that earn LEED Platinum or Gold certification may still claim full expensing, linking the tax benefit to high energy‑efficiency standards.
- Revenue impact (Tax Foundation TAG model)
- Central estimate: $29.9 billion in conventional revenue over 2027‑2036.
- Dynamic estimate (accounting for reduced GDP): $18.2 billion over the same period.
- Long‑run GDP effect: < 0.05 % reduction, reflecting a modest increase in the cost of capital.
- Scenario range – If the share of affected investment varies, revenue could be $17.5 billion (low) to $46.5 billion (high) over ten years.
Wyden’s Data Center Tax White Paper (August 2026)
- Scope – Proposes a gross receipts tax on data‑center operators, with an undefined rate described as “low single digits.”
- Exemptions – “Internet infrastructure,” corporate IT departments, “small” data centers, and pre‑2024 facilities (except the largest operators) would be exempt.
- Deemed minimum regime – Requires data‑center users above certain asset and expense thresholds to pay a minimum tax, aimed at preventing avoidance.
- Potential effect – Even a 4 % gross receipts tax can translate into an effective tax on net income of 40 %–200 % depending on a firm’s expense structure, making the nominal rate potentially burdensome.
Common Concerns
- Administrative ambiguity – Both bills leave key definitions unclear (e.g., how the 20 % AI usage threshold is measured; what qualifies as “small” or “internet infrastructure”).
- Investment displacement – Neither proposal addresses non‑US data‑center investment, raising the risk that developers shift new facilities abroad, which could erode the United States’ projected $581 billion AI capital‑expenditure advantage in 2026.
- Tax policy rationale – Existing corporate income, capital‑gains, and property taxes already capture supernormal returns from AI enterprises. Targeted taxes on AI data centers may duplicate revenue while discouraging investment.
Economic Context
- AI investment outlook – Goldman Sachs forecasts $1 trillion of global AI‑related spending in 2026, with $581 billion occurring in the United States.
- Full expensing benefits – Allows immediate deduction of equipment and building costs, encouraging rapid deployment of AI infrastructure and supporting higher wages and job creation. Removing this incentive shifts deductions to depreciation over time, which can delay but not increase long‑run tax revenue.
Policy Implications
- A neutral tax framework that permits cost recovery while taxing profits may better capture the broader economic gains from AI without jeopardizing US competitiveness.
- Imposing new excise or gross‑receipts taxes on AI data centers could add complexity, distort investment decisions, and potentially push activity overseas, undermining the projected economic benefits of AI adoption.
Source article: taxfoundation.org
