Taxing AI | Bond Buyer

While states and counties wrestle with tax policies affecting data center development, the federal government is also weighing strategies for tapping the billions flowing into an artificial intelligence industry that could decimate employment. 

Processing Content

“Americans feel the looming fear of AI job losses, and Congress has a responsibility to help the country prepare,” said Sen. Ron Wyden D-Ore.

“The support meant to protect workers from financial devastation when they lose their jobs is brittle and out of date and needs to be updated before the next economic crisis.” 

Wyden, who is the ranking member on the Senate Finance Committee has drafted legislation that would establish a transition wages program to provide financial assistance to workers displaced by artificial intelligence. 

The bill announced last week dovetails with a tax package Wyden is working on that would levy an excise tax on data center operators and “removes unnecessary tax breaks for data centers, such as bonus depreciation, opportunity zone benefits and research and development tax credits.” 

Data centers generate large amounts of property tax revenue at the local level and many states exempt them from paying sales tax on computer hardware that needs to be updated on a regular basis. 

Local opposition to the centers, usually centered on environmental and utility rate concerns has slowed development and become a political issue. 

The potential for job loss is also a rising concern as tax revenues are seen as a potential remedy for helping workers transition to other employment. 

“Artificial intelligence could raise productivity across a wide range of occupations and tasks. But it could also replace workers who currently perform some of those tasks,” said Eleana Patel, and Aparna Mathur, writing for Brookings.

“How many workers AI might replace, and how quickly, remains an open question.” 

Patel is a co-director at the Urban-Brookings Tax Policy Center, Mathur is a resident scholar of economic policy at the American Enterprise Institute.

As the demand for AI continues to grow, strategies for taxing the industry are forming into possible levies on computing power, electricity usage, units of AI output, windfall profits, and automation.

According to research from Brookings, taxing the computing power is difficult as it relies on self-reporting. 

A tax on units of AI output “taxes AI as a business input and cannot distinguish between productive and wasteful uses.” 

A tax on windfall profits wouldn’t kick in until the AI firm actually shows a profit. 

Taxing automation would be tied to job displacement and wouldn’t provide any revenue until after the job was eliminated.   

Virginia, the leader in data center placements has already enacted a tax on kilowatt hours used by the industry. 

Some states are leaning into existing taxes on software to extract money from AI subscribers.

According to Numeral, an AI-native financial software company that tracks taxes on tech, “States often treat AI as software as a service, meaning customers access it online through a subscription instead of downloading it.” 

“Other states treat it as prewritten software, meaning off-the-shelf software built for many buyers instead of one.” 

Kentucky has determined that AI components comply with their definition of “pre-written computer software,” and is subject to sales tax. 

Texas includes AI as data processing, making it generally taxable. 

Washington state considers it to be digital services and is generally taxable. 

California and Colorado both have AI teed up for sales tax starting next year. 

Chicago has decided AI is included in its Personal Property Lease Transaction tax with a rate of 15%. 

“Because AI is a new technology, no state has yet established a sales tax regulation that treats artificial intelligence as a separate product or service category with its own tax rules,” said Numeral.