State tax policy affecting AI

Jared Walczak, vice president of state projects at the Tax Foundation.
Jared Walczak, vice president of state projects at the Tax Foundation.

The rise of artificial intelligence and the data centers necessary to it are subject to tax policies enacted by states to attract the revenues they produce. 

Processing Content

"If lawmakers want data center investment in their states, they do need to ensure that the tax code does not uniquely penalize these operations, as is often the case," wrote Jared Walczak, vice president of state projects at the Tax Foundation. 

"States can't legislate their climate, and some regions lack the available open land for data center scalability," he wrote. "Tax structures, by contrast, are the policy lever most within reach of state lawmakers."   

The Tax Foundation has crunched hypothetical numbers for locations in twelve different states by setting up a model that places a $1 billion data center in each one. The fictional company has $220 million in revenue and 15% net income before taxes.  

"A data center is unlikely to be profitable in its first year given its high upfront costs, but comparing first-year taxes to long-run average profits yields an astonishing 263% rate in Santa Clara, where servers and other equipment are taxed under a high-rate sales tax," per the report.  

By contrast, the same data center in Charlotte, N.C. would have a 7% rate. Loudoun County, Va., which is known as a preferred location for data centers, also breaks the model by clocking in at 56%, the second highest in the study.  

"Silicon Valley, with its tech agglomeration and centrality to the new artificial intelligence boom, and Loudoun County, Virginia, the historic 'capital of the internet,' have built-in advantages that continue to attract data center operations even at high tax costs," per the report. 

Loudoun County's built-in advantages include affordable land, favorable electricity rates, an established fiber optic network, proximity to Washington D.C. and real-world tax breaks at the county level that combine to trump the impacts from the model. 

Data center investors are increasingly looking towards debt as a way to finance and refinance construction costs. 

Public power companies, which rely on bond sales to build infrastructure, are stymied by Internal Revenue Service regulations restricting their participation in the data center game. 

Under current IRS private use rules for tax-exempt bonds, public power utilities are restricted to three-year contracts with non-government customers, and less than 10% of a bond issue, or $15 million, could go to a private use.

The Trump administration has waded into the fray as the Federal Energy Regulatory Commission last week ordered the largest Regional Transmission Organization in the U.S. to develop rules for "co-locating" data centers next to power plants. 

FERC is an independent agency operating within the Department of Energy and is governed by five commissioners nominated by the President and confirmed by the Senate. 

According to McKinsey & Co., global "data center investments are projected to reach nearly $7 trillion by 2030. More than $4 trillion will be allocated to computing hardware." 

"With more than 40% of this spending expected to occur in the United States, each state can carefully assess costs and benefits to ensure optimal outcomes for their communities."  

For reprint and licensing requests for this article, click here.
Artificial intelligence Tax Politics and policy Energy industry Private activity bonds
MORE FROM BOND BUYER