How Machine Learning May Impact Data Center Tax Incentives

A data center, technically speaking, is defined as a large group of networked computer systems that store, process, or distribute massive amounts of data. As such, they require and depend on locations that have access to sufficient and affordable power and space. Since the cost of starting and running a data center can easily (and often does) run into the tens of millions, the taxes associated are sky-high.

Data centers are subject to both sales and property taxes, including those on real estate and personal property (or equipment). Sales taxes would be incurred on a one-time basis, such as when building materials, IT equipment, or mechanical and electrical equipment is purchased. As an example, $10 million of technology delivered to a center in Ohio would incur a staggering $700,000 in sales taxes.

Real estate taxes are paid annually for data center structure itself and depend on building value and the local tax rate. Another example: a center valued at $30 million in Kansas would see real estate taxes reach around $4.6 million over five years, or about $930,000 annually.

While most businesses would struggle to stay above board with such extreme taxes, the U.S. government weighs the benefits offered to states and communities from the presence of data centers against these costs. The IRS is hardly a forgiving agency, well-known for reaching a decade into the past to collect back taxes, but data centers offer more than the average American Joe: they can stimulate local and state economies, create jobs (mostly through their lengthy, high-end construction period), and encourage new technologically-based business to set up shop.

As long-term investments, data centers receive some breathing room from the Internal Revenue Service: they see sales tax breaks from select states (about 17 have implemented customized incentives for data centers since 2005), property tax abatements that cover the facility and equipment, and even cash grants to make public infrastructure improvements more manageable.

However, the introduction of machine learning might change this. Machine learning is a subset of artificial intelligence, and is considered revolutionary in terms of optimization. Every aspect of data center management — from planning and design to cooling operations — will become more efficient and reliable with the implementation of machine learning.

In fact, the cost of cooling data centers (in order to keep them running at full capacity) has been the focus of many companies in recent years, as it eats up considerable funds yet is absolutely necessary for the success and functionality of the center. The favored option right now is immersion cooling: by immersing servers in liquid, data centers can improve rack density, cooling capacity, and data center design and location options. Still, someone — a fallible and imperfect human being — needs to control the process.

With machine learning, the process becomes autonomous; taking a look at Google’s new AI cooling system shows the possibilities as it eliminates around 40% of the facility cooling system’s total energy use. But what effect would this have on the tax incentives?

Utilizing machine learning would certainly boost the popularity of data centers: potential new investors could see how much they’re saving on energy costs and travel from far and wide to establish themselves. On the flip side, however, the cost of artificial intelligence programming will be considerably more than the already extremely expensive technological equipment used in data centers.

The U.S. government will simply have to weight the benefits against the costs, as they have been doing since the mid-2000s. Only time will tell what decision they come to.

Leave a Reply

Your email address will not be published. Required fields are marked *